diff --git a/README.md b/README.md index f54b0346..2fc9b0be 100644 --- a/README.md +++ b/README.md @@ -26,7 +26,7 @@ Explore the world of **Artificial Intelligence** (AI) with our 12-week, 24-lesso #### Supported via GitHub Action (Automated & Always Up-to-Date) -[Arabic](./translations/ar/README.md) | [Bengali](./translations/bn/README.md) | [Bulgarian](./translations/bg/README.md) | [Burmese (Myanmar)](./translations/my/README.md) | [Chinese (Simplified)](./translations/zh/README.md) | [Chinese (Traditional, Hong Kong)](./translations/hk/README.md) | [Chinese (Traditional, Macau)](./translations/mo/README.md) | [Chinese (Traditional, Taiwan)](./translations/tw/README.md) | [Croatian](./translations/hr/README.md) | [Czech](./translations/cs/README.md) | [Danish](./translations/da/README.md) | [Dutch](./translations/nl/README.md) | [Estonian](./translations/et/README.md) | [Finnish](./translations/fi/README.md) | [French](./translations/fr/README.md) | [German](./translations/de/README.md) | [Greek](./translations/el/README.md) | [Hebrew](./translations/he/README.md) | [Hindi](./translations/hi/README.md) | [Hungarian](./translations/hu/README.md) | [Indonesian](./translations/id/README.md) | [Italian](./translations/it/README.md) | [Japanese](./translations/ja/README.md) | [Kannada](./translations/kn/README.md) | [Korean](./translations/ko/README.md) | [Lithuanian](./translations/lt/README.md) | [Malay](./translations/ms/README.md) | [Malayalam](./translations/ml/README.md) | [Marathi](./translations/mr/README.md) | [Nepali](./translations/ne/README.md) | [Nigerian Pidgin](./translations/pcm/README.md) | [Norwegian](./translations/no/README.md) | [Persian (Farsi)](./translations/fa/README.md) | [Polish](./translations/pl/README.md) | [Portuguese (Brazil)](./translations/br/README.md) | [Portuguese (Portugal)](./translations/pt/README.md) | [Punjabi (Gurmukhi)](./translations/pa/README.md) | [Romanian](./translations/ro/README.md) | [Russian](./translations/ru/README.md) | [Serbian (Cyrillic)](./translations/sr/README.md) | [Slovak](./translations/sk/README.md) | [Slovenian](./translations/sl/README.md) | [Spanish](./translations/es/README.md) | [Swahili](./translations/sw/README.md) | [Swedish](./translations/sv/README.md) | [Tagalog (Filipino)](./translations/tl/README.md) | [Tamil](./translations/ta/README.md) | [Telugu](./translations/te/README.md) | [Thai](./translations/th/README.md) | [Turkish](./translations/tr/README.md) | [Ukrainian](./translations/uk/README.md) | [Urdu](./translations/ur/README.md) | [Vietnamese](./translations/vi/README.md) +[Arabic](./translations/ar/README.md) | [Bengali](./translations/bn/README.md) | [Bulgarian](./translations/bg/README.md) | [Burmese (Myanmar)](./translations/my/README.md) | [Chinese (Simplified)](./translations/zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](./translations/zh-HK/README.md) | [Chinese (Traditional, Macau)](./translations/zh-MO/README.md) | [Chinese (Traditional, Taiwan)](./translations/zh-TW/README.md) | [Croatian](./translations/hr/README.md) | [Czech](./translations/cs/README.md) | [Danish](./translations/da/README.md) | [Dutch](./translations/nl/README.md) | [Estonian](./translations/et/README.md) | [Finnish](./translations/fi/README.md) | [French](./translations/fr/README.md) | [German](./translations/de/README.md) | [Greek](./translations/el/README.md) | [Hebrew](./translations/he/README.md) | [Hindi](./translations/hi/README.md) | [Hungarian](./translations/hu/README.md) | [Indonesian](./translations/id/README.md) | [Italian](./translations/it/README.md) | [Japanese](./translations/ja/README.md) | [Kannada](./translations/kn/README.md) | [Korean](./translations/ko/README.md) | [Lithuanian](./translations/lt/README.md) | [Malay](./translations/ms/README.md) | [Malayalam](./translations/ml/README.md) | [Marathi](./translations/mr/README.md) | [Nepali](./translations/ne/README.md) | [Nigerian Pidgin](./translations/pcm/README.md) | [Norwegian](./translations/no/README.md) | [Persian (Farsi)](./translations/fa/README.md) | [Polish](./translations/pl/README.md) | [Portuguese (Brazil)](./translations/pt-BR/README.md) | [Portuguese (Portugal)](./translations/pt-PT/README.md) | [Punjabi (Gurmukhi)](./translations/pa/README.md) | [Romanian](./translations/ro/README.md) | [Russian](./translations/ru/README.md) | [Serbian (Cyrillic)](./translations/sr/README.md) | [Slovak](./translations/sk/README.md) | [Slovenian](./translations/sl/README.md) | [Spanish](./translations/es/README.md) | [Swahili](./translations/sw/README.md) | [Swedish](./translations/sv/README.md) | [Tagalog (Filipino)](./translations/tl/README.md) | [Tamil](./translations/ta/README.md) | [Telugu](./translations/te/README.md) | [Thai](./translations/th/README.md) | [Turkish](./translations/tr/README.md) | [Ukrainian](./translations/uk/README.md) | [Urdu](./translations/ur/README.md) | [Vietnamese](./translations/vi/README.md) > **Prefer to Clone Locally?** diff --git a/translations/br/README.md b/translations/br/README.md index c22c79ad..94635f62 100644 --- a/translations/br/README.md +++ b/translations/br/README.md @@ -23,7 +23,7 @@ CO_OP_TRANSLATOR_METADATA: # Inteligência Artificial para Iniciantes - Um Currículo -|![Sketchnote por @girlie_mac https://twitter.com/girlie_mac](../../../../translated_images/br/ai-overview.0857791951d19500.webp)| +|![Sketchnote por @girlie_mac https://twitter.com/girlie_mac](../../../../translated_images/pt-BR/ai-overview.0857791951d19500.webp)| |:---:| | AI For Beginners - _Sketchnote por [@girlie_mac](https://twitter.com/girlie_mac)_ | diff --git a/translations/br/lessons/1-Intro/README.md b/translations/br/lessons/1-Intro/README.md index 5db483cf..251e50b9 100644 --- a/translations/br/lessons/1-Intro/README.md +++ b/translations/br/lessons/1-Intro/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # Introdução à IA -![Resumo do conteúdo de Introdução à IA em um doodle](../../../../translated_images/br/ai-intro.bf28d1ac4235881c.webp) +![Resumo do conteúdo de Introdução à IA em um doodle](../../../../translated_images/pt-BR/ai-intro.bf28d1ac4235881c.webp) > Sketchnote por [Tomomi Imura](https://twitter.com/girlie_mac) @@ -19,7 +19,7 @@ CO_OP_TRANSLATOR_METADATA: Originalmente, os computadores foram inventados por [Charles Babbage](https://en.wikipedia.org/wiki/Charles_Babbage) para operar com números seguindo um procedimento bem definido - um algoritmo. Os computadores modernos, embora significativamente mais avançados que o modelo original proposto no século XIX, ainda seguem a mesma ideia de cálculos controlados. Assim, é possível programar um computador para fazer algo se soubermos a sequência exata de passos necessários para alcançar o objetivo. -![Foto de uma pessoa](../../../../translated_images/br/dsh_age.d212a30d4e54fb5f.webp) +![Foto de uma pessoa](../../../../translated_images/pt-BR/dsh_age.d212a30d4e54fb5f.webp) > Foto por [Vickie Soshnikova](http://twitter.com/vickievalerie) @@ -46,7 +46,7 @@ Para mais informações, consulte **[Inteligência Artificial Geral](https://en. Um dos problemas ao lidar com o termo **[Inteligência](https://en.wikipedia.org/wiki/Intelligence)** é que não há uma definição clara para esse termo. Pode-se argumentar que inteligência está conectada ao **pensamento abstrato** ou à **autoconsciência**, mas não conseguimos defini-la adequadamente. -![Foto de um gato](../../../../translated_images/br/photo-cat.8c8e8fb760ffe457.webp) +![Foto de um gato](../../../../translated_images/pt-BR/photo-cat.8c8e8fb760ffe457.webp) > [Foto](https://unsplash.com/photos/75715CVEJhI) por [Amber Kipp](https://unsplash.com/@sadmax) do Unsplash @@ -98,13 +98,13 @@ Alternativamente, podemos tentar modelar os elementos mais simples dentro de nos > | E o ML? | | > |--------------|-----------| -> | Parte da Inteligência Artificial que se baseia no aprendizado do computador para resolver um problema com base em alguns dados é chamada de **Machine Learning**. Não consideraremos o aprendizado de máquina clássico neste curso - recomendamos o currículo separado [Machine Learning para Iniciantes](http://aka.ms/ml-beginners). | ![ML para Iniciantes](../../../../translated_images/br/ml-for-beginners.9e4fed176fd5817d.webp) | +> | Parte da Inteligência Artificial que se baseia no aprendizado do computador para resolver um problema com base em alguns dados é chamada de **Machine Learning**. Não consideraremos o aprendizado de máquina clássico neste curso - recomendamos o currículo separado [Machine Learning para Iniciantes](http://aka.ms/ml-beginners). | ![ML para Iniciantes](../../../../translated_images/pt-BR/ml-for-beginners.9e4fed176fd5817d.webp) | ## Um Breve Histórico da IA A Inteligência Artificial começou como um campo no meio do século XX. Inicialmente, o raciocínio simbólico era a abordagem predominante, e isso levou a uma série de sucessos importantes, como sistemas especialistas – programas de computador capazes de atuar como especialistas em alguns domínios de problemas limitados. No entanto, logo ficou claro que essa abordagem não escala bem. Extrair o conhecimento de um especialista, representá-lo em um computador e manter essa base de conhecimento precisa acaba sendo uma tarefa muito complexa e cara demais para ser prática em muitos casos. Isso levou ao chamado [Inverno da IA](https://en.wikipedia.org/wiki/AI_winter) na década de 1970. -Breve Histórico da IA +Breve Histórico da IA > Imagem por [Dmitry Soshnikov](http://soshnikov.com) @@ -124,7 +124,7 @@ Da mesma forma, podemos ver como a abordagem para criar “programas que falam * Assistentes modernos, como Cortana, Siri ou Google Assistant, são todos sistemas híbridos que usam redes neurais para converter fala em texto e reconhecer nossa intenção, e então empregam algum raciocínio ou algoritmos explícitos para realizar as ações necessárias. * No futuro, podemos esperar um modelo completamente baseado em redes neurais para lidar com diálogos por conta própria. As recentes redes neurais da família GPT e [Turing-NLG](https://www.microsoft.com/research/blog/turing-nlg-a-17-billion-parameter-language-model-by-microsoft) mostram grande sucesso nisso. -a evolução do teste de Turing +a evolução do teste de Turing > Imagem de Dmitry Soshnikov, [foto](https://unsplash.com/photos/r8LmVbUKgns) por [Marina Abrosimova](https://unsplash.com/@abrosimova_marina_foto), Unsplash ## Pesquisas Recentes em IA diff --git a/translations/br/lessons/2-Symbolic/Animals.ipynb b/translations/br/lessons/2-Symbolic/Animals.ipynb index dcc9230e..7eff832a 100644 --- a/translations/br/lessons/2-Symbolic/Animals.ipynb +++ b/translations/br/lessons/2-Symbolic/Animals.ipynb @@ -12,7 +12,7 @@ "\n", "Neste exemplo, implementaremos um sistema baseado em conhecimento simples para determinar um animal com base em algumas características físicas. O sistema pode ser representado pela seguinte árvore AND-OR (esta é uma parte da árvore inteira, podemos facilmente adicionar mais algumas regras):\n", "\n", - "![](../../../../../../translated_images/br/AND-OR-Tree.5592d2c70187f283.webp)\n" + "![](../../../../../../translated_images/pt-BR/AND-OR-Tree.5592d2c70187f283.webp)\n" ] }, { diff --git a/translations/br/lessons/2-Symbolic/README.md b/translations/br/lessons/2-Symbolic/README.md index fbe088e8..ec7fd045 100644 --- a/translations/br/lessons/2-Symbolic/README.md +++ b/translations/br/lessons/2-Symbolic/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # Representação do Conhecimento e Sistemas Especialistas -![Resumo do conteúdo de IA Simbólica](../../../../../../translated_images/br/ai-symbolic.715a30cb610411a6.webp) +![Resumo do conteúdo de IA Simbólica](../../../../../../translated_images/pt-BR/ai-symbolic.715a30cb610411a6.webp) > Sketchnote por [Tomomi Imura](https://twitter.com/girlie_mac) @@ -35,13 +35,13 @@ Na maioria das vezes, não definimos conhecimento de forma estrita, mas o alinha * **Conhecimento** é a informação integrada em nosso modelo do mundo. Por exemplo, uma vez que aprendemos o que é um computador, começamos a ter algumas ideias sobre como ele funciona, quanto custa e para que pode ser usado. Essa rede de conceitos inter-relacionados forma nosso conhecimento. * **Sabedoria** é ainda um nível a mais da nossa compreensão do mundo, e representa *meta-conhecimento*, por exemplo, alguma noção sobre como e quando o conhecimento deve ser usado. - + *Imagem [da Wikipedia](https://commons.wikimedia.org/w/index.php?curid=37705247), Por Longlivetheux - Trabalho próprio, CC BY-SA 4.0* Assim, o problema da **representação do conhecimento** é encontrar algum meio eficaz para representar conhecimento dentro de um computador na forma de dados, para torná-lo automaticamente utilizável. Isso pode ser visto como um espectro: -![Espectro de representação do conhecimento](../../../../../../translated_images/br/knowledge-spectrum.b60df631852c0217.webp) +![Espectro de representação do conhecimento](../../../../../../translated_images/pt-BR/knowledge-spectrum.b60df631852c0217.webp) > Imagem por [Dmitry Soshnikov](http://soshnikov.com) @@ -94,7 +94,7 @@ Sintaxe de Bloco | Indentação | | | Um dos primeiros sucessos da IA simbólica foram os chamados **sistemas especialistas** - sistemas computacionais desenhados para atuar como um especialista em algum domínio de problema limitado. Eles foram baseados em uma **base de conhecimento** extraída de um ou mais especialistas humanos, e continham um **motor de inferência** que realizava algum raciocínio sobre ela. -![Arquitetura Humana](../../../../../../translated_images/br/arch-human.5d4d35f1bba3ab1c.webp) | ![Sistema Baseado em Conhecimento](../../../../../../translated_images/br/arch-kbs.3ec5c150b09fa8da.webp) +![Arquitetura Humana](../../../../../../translated_images/pt-BR/arch-human.5d4d35f1bba3ab1c.webp) | ![Sistema Baseado em Conhecimento](../../../../../../translated_images/pt-BR/arch-kbs.3ec5c150b09fa8da.webp) ---------------------------------------------|------------------------------------------------ Estrutura simplificada do sistema neural humano | Arquitetura de um sistema baseado em conhecimento @@ -106,7 +106,7 @@ Sistemas especialistas são construídos como o sistema de raciocínio humano, q Como exemplo, vamos considerar o seguinte sistema especialista para determinar um animal com base em suas características físicas: -![Árvore AND-OR](../../../../../../translated_images/br/AND-OR-Tree.5592d2c70187f283.webp) +![Árvore AND-OR](../../../../../../translated_images/pt-BR/AND-OR-Tree.5592d2c70187f283.webp) > Imagem por [Dmitry Soshnikov](http://soshnikov.com) @@ -175,7 +175,7 @@ Um conceito central na Web Semântica é o conceito de **Ontologia**. Refere-se Na web semântica, todas as representações são baseadas em tripletas. Cada objeto e cada relação são identificados unicamente pela URI. Por exemplo, se quisermos afirmar o fato de que este Currículo de IA foi desenvolvido por Dmitry Soshnikov em 1º de janeiro de 2022 - aqui estão as tripletas que podemos usar: - + ``` http://github.com/microsoft/ai-for-beginners http://www.example.com/terms/creation-date “Jan 1, 2022” @@ -186,7 +186,7 @@ http://github.com/microsoft/ai-for-beginners http://purl.org/dc/elements/1.1/cre Em um caso mais complexo, se quisermos definir uma lista de criadores, podemos usar algumas estruturas de dados definidas em RDF. - + > Diagramas acima por [Dmitry Soshnikov](http://soshnikov.com) @@ -210,7 +210,7 @@ GROUP BY ?eyeColorLabel > ✅ Se você quiser experimentar criar suas próprias ontologias, ou abrir as existentes, existe um ótimo editor visual de ontologias chamado [Protégé](https://protege.stanford.edu/). Faça o download ou use online. - + *Editor Web Protégé aberto com a ontologia da Família Romanov. Captura de tela por Dmitry Soshnikov* diff --git a/translations/br/lessons/3-NeuralNetworks/04-OwnFramework/OwnFramework.ipynb b/translations/br/lessons/3-NeuralNetworks/04-OwnFramework/OwnFramework.ipynb index 60ba16e6..26f4d374 100644 --- a/translations/br/lessons/3-NeuralNetworks/04-OwnFramework/OwnFramework.ipynb +++ b/translations/br/lessons/3-NeuralNetworks/04-OwnFramework/OwnFramework.ipynb @@ -487,7 +487,7 @@ "\n", "No caso de termos mais de 2 classes, o softmax irá normalizar as probabilidades entre todas elas. Aqui está um diagrama da arquitetura de rede que realiza a classificação de dígitos do MNIST:\n", "\n", - "![Classificador MNIST](../../../../../translated_images/br/Cross-Entropy-Loss.7acff482d48cc41b2d3ef85ff0b263fbab299da128a462fa9b0640c5b7bc12ef.png)\n" + "![Classificador MNIST](../../../../../translated_images/pt-BR/Cross-Entropy-Loss.7acff482d48cc41b2d3ef85ff0b263fbab299da128a462fa9b0640c5b7bc12ef.png)\n" ] }, { @@ -1256,7 +1256,7 @@ "* Baixa perda de treinamento - o modelo consegue aproximar bem os dados de treinamento, pois tem poder expressivo suficiente.\n", "* A perda de validação pode ser muito maior do que a perda de treinamento e pode começar a aumentar durante o treinamento - isso ocorre porque o modelo \"memoriza\" os pontos de treinamento e perde a \"visão geral\".\n", "\n", - "![Overfitting](../../../../../translated_images/br/overfit.a0bd57f717c15769.webp)\n", + "![Overfitting](../../../../../translated_images/pt-BR/overfit.a0bd57f717c15769.webp)\n", "\n", "> Nesta imagem, `x` representa os dados de treinamento, e `o` os dados de validação. À esquerda - modelo linear (uma camada), que aproxima bem a natureza dos dados. À direita - modelo com overfitting, que aproxima perfeitamente os dados de treinamento, mas deixa de fazer sentido com qualquer outro dado (o erro de validação é muito alto).\n" ] diff --git a/translations/br/lessons/3-NeuralNetworks/05-Frameworks/README.md b/translations/br/lessons/3-NeuralNetworks/05-Frameworks/README.md index 83bef40d..c5cec31d 100644 --- a/translations/br/lessons/3-NeuralNetworks/05-Frameworks/README.md +++ b/translations/br/lessons/3-NeuralNetworks/05-Frameworks/README.md @@ -58,7 +58,7 @@ Overfitting é um conceito extremamente importante em aprendizado de máquina, e Considere o seguinte problema de aproximar 5 pontos (representados por `x` nos gráficos abaixo): -![linear](../../../../../translated_images/br/overfit1.f24b71c6f652e59e.webp) | ![overfit](../../../../../translated_images/br/overfit2.131f5800ae10ca5e.webp) +![linear](../../../../../translated_images/pt-BR/overfit1.f24b71c6f652e59e.webp) | ![overfit](../../../../../translated_images/pt-BR/overfit2.131f5800ae10ca5e.webp) -------------------------|-------------------------- **Modelo linear, 2 parâmetros** | **Modelo não-linear, 7 parâmetros** Erro de treinamento = 5.3 | Erro de treinamento = 0 @@ -79,7 +79,7 @@ Erro de validação = 5.1 | Erro de validação = 20 Como você pode ver no gráfico acima, o overfitting pode ser detectado por um erro de treinamento muito baixo e um erro de validação alto. Normalmente, durante o treinamento, veremos tanto o erro de treinamento quanto o de validação começarem a diminuir, e então, em algum momento, o erro de validação pode parar de diminuir e começar a aumentar. Isso será um sinal de overfitting e um indicador de que provavelmente devemos parar o treinamento nesse ponto (ou pelo menos salvar um snapshot do modelo). -![overfitting](../../../../../translated_images/br/Overfitting.408ad91cd90b4371.webp) +![overfitting](../../../../../translated_images/pt-BR/Overfitting.408ad91cd90b4371.webp) ## Como prevenir o overfitting diff --git a/translations/br/lessons/3-NeuralNetworks/README.md b/translations/br/lessons/3-NeuralNetworks/README.md index 0695d4df..e2915557 100644 --- a/translations/br/lessons/3-NeuralNetworks/README.md +++ b/translations/br/lessons/3-NeuralNetworks/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # Introdução às Redes Neurais -![Resumo do conteúdo de Introdução às Redes Neurais em um desenho](../../../../translated_images/br/ai-neuralnetworks.1c687ae40bc86e83.webp) +![Resumo do conteúdo de Introdução às Redes Neurais em um desenho](../../../../translated_images/pt-BR/ai-neuralnetworks.1c687ae40bc86e83.webp) Como discutimos na introdução, uma das formas de alcançar inteligência é treinar um **modelo computacional** ou um **cérebro artificial**. Desde meados do século XX, pesquisadores experimentaram diferentes modelos matemáticos, até que, nos últimos anos, essa abordagem se mostrou extremamente bem-sucedida. Esses modelos matemáticos do cérebro são chamados de **redes neurais**. @@ -36,13 +36,13 @@ Neste currículo, focaremos apenas em modelos de redes neurais. Na biologia, sabemos que nosso cérebro é composto por células neurais (neurônios), cada uma delas tendo múltiplas "entradas" (dendritos) e uma única "saída" (axônio). Tanto os dendritos quanto os axônios podem conduzir sinais elétricos, e as conexões entre eles — conhecidas como sinapses — podem apresentar diferentes graus de condutividade, que são regulados por neurotransmissores. -![Modelo de um Neurônio](../../../../translated_images/br/synapse-wikipedia.ed20a9e4726ea1c6.webp) | ![Modelo de um Neurônio](../../../../translated_images/br/artneuron.1a5daa88d20ebe6f.webp) +![Modelo de um Neurônio](../../../../translated_images/pt-BR/synapse-wikipedia.ed20a9e4726ea1c6.webp) | ![Modelo de um Neurônio](../../../../translated_images/pt-BR/artneuron.1a5daa88d20ebe6f.webp) ----|---- Neurônio Real *([Imagem](https://en.wikipedia.org/wiki/Synapse#/media/File:SynapseSchematic_lines.svg) da Wikipedia)* | Neurônio Artificial *(Imagem do Autor)* Assim, o modelo matemático mais simples de um neurônio contém várias entradas X1, ..., XN e uma saída Y, e uma série de pesos W1, ..., WN. A saída é calculada como: -Y = f\left(\sum_{i=1}^N X_iW_i\right) +Y = f\left(\sum_{i=1}^N X_iW_i\right) onde f é alguma **função de ativação** não linear. diff --git a/translations/br/lessons/4-ComputerVision/06-IntroCV/README.md b/translations/br/lessons/4-ComputerVision/06-IntroCV/README.md index 61a024dc..29716941 100644 --- a/translations/br/lessons/4-ComputerVision/06-IntroCV/README.md +++ b/translations/br/lessons/4-ComputerVision/06-IntroCV/README.md @@ -73,14 +73,14 @@ Em nosso [OpenCV Notebook](OpenCV.ipynb), damos alguns exemplos de quando a vis * **Pré-processamento de uma fotografia de um livro em Braille**. Focamos em como podemos usar limiarização, detecção de características, transformação de perspectiva e manipulações NumPy para separar símbolos individuais em Braille para posterior classificação por uma rede neural. -![Imagem Braille](../../../../../translated_images/br/braille.341962ff76b1bd70.webp) | ![Imagem Braille Pré-processada](../../../../../translated_images/br/braille-result.46530fea020b03c7.webp) | ![Símbolos Braille](../../../../../translated_images/br/braille-symbols.0159185ab69d5339.webp) +![Imagem Braille](../../../../../translated_images/pt-BR/braille.341962ff76b1bd70.webp) | ![Imagem Braille Pré-processada](../../../../../translated_images/pt-BR/braille-result.46530fea020b03c7.webp) | ![Símbolos Braille](../../../../../translated_images/pt-BR/braille-symbols.0159185ab69d5339.webp) ----|-----|----- > Imagem de [OpenCV.ipynb](OpenCV.ipynb) * **Detectando movimento em vídeo usando diferença de quadros**. Se a câmera estiver fixa, os quadros do feed da câmera devem ser bastante semelhantes entre si. Como os quadros são representados como arrays, apenas subtraindo esses arrays de dois quadros subsequentes obteremos a diferença de pixels, que deve ser baixa para quadros estáticos e se tornar maior quando houver movimento substancial na imagem. -![Imagem de quadros de vídeo e diferenças de quadros](../../../../../translated_images/br/frame-difference.706f805491a0883c.webp) +![Imagem de quadros de vídeo e diferenças de quadros](../../../../../translated_images/pt-BR/frame-difference.706f805491a0883c.webp) > Imagem de [OpenCV.ipynb](OpenCV.ipynb) @@ -89,7 +89,7 @@ Em nosso [OpenCV Notebook](OpenCV.ipynb), damos alguns exemplos de quando a vis - **Fluxo Óptico Denso** calcula o campo vetorial que mostra para cada pixel onde ele está se movendo. - **Fluxo Óptico Esparso** é baseado em pegar algumas características distintivas na imagem (por exemplo, bordas) e construir sua trajetória de quadro a quadro. -![Imagem de Fluxo Óptico](../../../../../translated_images/br/optical.1f4a94464579a83a.webp) +![Imagem de Fluxo Óptico](../../../../../translated_images/pt-BR/optical.1f4a94464579a83a.webp) > Imagem de [OpenCV.ipynb](OpenCV.ipynb) diff --git a/translations/br/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md b/translations/br/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md index 33d4c25b..c8686988 100644 --- a/translations/br/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md +++ b/translations/br/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md @@ -13,11 +13,11 @@ CO_OP_TRANSLATOR_METADATA: VGG-16 é uma rede que alcançou 92,7% de precisão na classificação top-5 do ImageNet em 2014. Ela possui a seguinte estrutura de camadas: -![Camadas do ImageNet](../../../../../translated_images/br/vgg-16-arch1.d901a5583b3a51ba.webp) +![Camadas do ImageNet](../../../../../translated_images/pt-BR/vgg-16-arch1.d901a5583b3a51ba.webp) Como você pode ver, a VGG segue uma arquitetura tradicional em forma de pirâmide, que é uma sequência de camadas de convolução e pooling. -![Pirâmide do ImageNet](../../../../../translated_images/br/vgg-16-arch.64ff2137f50dd49f.webp) +![Pirâmide do ImageNet](../../../../../translated_images/pt-BR/vgg-16-arch.64ff2137f50dd49f.webp) > Imagem de [Researchgate](https://www.researchgate.net/figure/Vgg16-model-structure-To-get-the-VGG-NIN-model-we-replace-the-2-nd-4-th-6-th-7-th_fig2_335194493) diff --git a/translations/br/lessons/4-ComputerVision/07-ConvNets/ConvNetsPyTorch.ipynb b/translations/br/lessons/4-ComputerVision/07-ConvNets/ConvNetsPyTorch.ipynb index 56ff46a0..80d4f6f4 100644 --- a/translations/br/lessons/4-ComputerVision/07-ConvNets/ConvNetsPyTorch.ipynb +++ b/translations/br/lessons/4-ComputerVision/07-ConvNets/ConvNetsPyTorch.ipynb @@ -262,7 +262,7 @@ "\n", "Assim, em uma CNN típica, haveria várias camadas de convolução, com camadas de pooling entre elas para diminuir as dimensões da imagem. Também aumentaríamos o número de filtros, porque à medida que os padrões se tornam mais avançados, há mais combinações interessantes possíveis que precisamos buscar.\n", "\n", - "![Uma imagem mostrando várias camadas de convolução com camadas de pooling.](../../../../../translated_images/br/cnn-pyramid.85915455759ef0ce.webp)\n", + "![Uma imagem mostrando várias camadas de convolução com camadas de pooling.](../../../../../translated_images/pt-BR/cnn-pyramid.85915455759ef0ce.webp)\n", "\n", "Devido à diminuição das dimensões espaciais e ao aumento das dimensões de características/filtros, essa arquitetura também é chamada de **arquitetura piramidal**.\n" ] diff --git a/translations/br/lessons/4-ComputerVision/07-ConvNets/ConvNetsTF.ipynb b/translations/br/lessons/4-ComputerVision/07-ConvNets/ConvNetsTF.ipynb index 63aa1100..354b5716 100644 --- a/translations/br/lessons/4-ComputerVision/07-ConvNets/ConvNetsTF.ipynb +++ b/translations/br/lessons/4-ComputerVision/07-ConvNets/ConvNetsTF.ipynb @@ -358,7 +358,7 @@ "\n", "Assim, em uma CNN típica, haveria várias camadas de convolução, com camadas de pooling entre elas para diminuir as dimensões da imagem. Também aumentaríamos o número de filtros, porque, à medida que os padrões se tornam mais avançados, há mais combinações interessantes que precisamos procurar.\n", "\n", - "![Uma imagem mostrando várias camadas de convolução com camadas de pooling.](../../../../../translated_images/br/cnn-pyramid.85915455759ef0ce.webp)\n", + "![Uma imagem mostrando várias camadas de convolução com camadas de pooling.](../../../../../translated_images/pt-BR/cnn-pyramid.85915455759ef0ce.webp)\n", "\n", "Devido à diminuição das dimensões espaciais e ao aumento das dimensões de características/filtros, essa arquitetura também é chamada de **arquitetura piramidal**.\n" ] diff --git a/translations/br/lessons/4-ComputerVision/07-ConvNets/README.md b/translations/br/lessons/4-ComputerVision/07-ConvNets/README.md index f12866df..b5f90cae 100644 --- a/translations/br/lessons/4-ComputerVision/07-ConvNets/README.md +++ b/translations/br/lessons/4-ComputerVision/07-ConvNets/README.md @@ -17,7 +17,7 @@ Na vida real, queremos ser capazes de reconhecer objetos em uma imagem independe Para extrair padrões, utilizaremos o conceito de **filtros convolucionais**. Como você sabe, uma imagem é representada por uma matriz 2D ou um tensor 3D com profundidade de cor. Aplicar um filtro significa que pegamos uma matriz relativamente pequena chamada **kernel do filtro**, e para cada pixel na imagem original calculamos a média ponderada com os pontos vizinhos. Podemos imaginar isso como uma pequena janela deslizando sobre toda a imagem e calculando a média de todos os pixels de acordo com os pesos na matriz do kernel do filtro. -![Filtro de Borda Vertical](../../../../../translated_images/br/filter-vert.b7148390ca0bc356.webp) | ![Filtro de Borda Horizontal](../../../../../translated_images/br/filter-horiz.59b80ed4feb946ef.webp) +![Filtro de Borda Vertical](../../../../../translated_images/pt-BR/filter-vert.b7148390ca0bc356.webp) | ![Filtro de Borda Horizontal](../../../../../translated_images/pt-BR/filter-horiz.59b80ed4feb946ef.webp) ----|---- > Imagem por Dmitry Soshnikov @@ -38,7 +38,7 @@ O funcionamento das CNNs é baseado nas seguintes ideias importantes: * Podemos projetar a rede de forma que os filtros sejam treinados automaticamente * Podemos usar a mesma abordagem para encontrar padrões em características de alto nível, não apenas na imagem original. Assim, a extração de características pelas CNNs funciona em uma hierarquia de características, começando com combinações de pixels de baixo nível até combinações de alto nível de partes da imagem. -![Extração Hierárquica de Características](../../../../../translated_images/br/FeatureExtractionCNN.d9b456cbdae7cb64.webp) +![Extração Hierárquica de Características](../../../../../translated_images/pt-BR/FeatureExtractionCNN.d9b456cbdae7cb64.webp) > Imagem de [um artigo de Hislop-Lynch](https://www.semanticscholar.org/paper/Computer-vision-based-pedestrian-trajectory-Hislop-Lynch/26e6f74853fc9bbb7487b06dc2cf095d36c9021d), baseado em [sua pesquisa](https://dl.acm.org/doi/abs/10.1145/1553374.1553453) @@ -55,9 +55,9 @@ A maioria das CNNs usadas para processamento de imagens segue a chamada arquitet Como exemplo, vejamos a arquitetura do VGG-16, uma rede que alcançou 92,7% de precisão na classificação top-5 do ImageNet em 2014: -![Camadas do ImageNet](../../../../../translated_images/br/vgg-16-arch1.d901a5583b3a51ba.webp) +![Camadas do ImageNet](../../../../../translated_images/pt-BR/vgg-16-arch1.d901a5583b3a51ba.webp) -![Pirâmide do ImageNet](../../../../../translated_images/br/vgg-16-arch.64ff2137f50dd49f.webp) +![Pirâmide do ImageNet](../../../../../translated_images/pt-BR/vgg-16-arch.64ff2137f50dd49f.webp) > Imagem de [Researchgate](https://www.researchgate.net/figure/Vgg16-model-structure-To-get-the-VGG-NIN-model-we-replace-the-2-nd-4-th-6-th-7-th_fig2_335194493) diff --git a/translations/br/lessons/4-ComputerVision/07-ConvNets/lab/README.md b/translations/br/lessons/4-ComputerVision/07-ConvNets/lab/README.md index 91cbb1f6..8a4fea5f 100644 --- a/translations/br/lessons/4-ComputerVision/07-ConvNets/lab/README.md +++ b/translations/br/lessons/4-ComputerVision/07-ConvNets/lab/README.md @@ -21,7 +21,7 @@ Você precisa treinar uma rede neural convolucional para classificar diferentes Usaremos o [Oxford-IIIT Pet Dataset](https://www.robots.ox.ac.uk/~vgg/data/pets/), que contém imagens de 37 diferentes raças de cães e gatos. -![Conjunto de dados com o qual trabalharemos](../../../../../../translated_images/br/data.50b2a9d5484bdbf0.webp) +![Conjunto de dados com o qual trabalharemos](../../../../../../translated_images/pt-BR/data.50b2a9d5484bdbf0.webp) Para baixar o conjunto de dados, use este trecho de código: diff --git a/translations/br/lessons/4-ComputerVision/08-TransferLearning/AdversarialCat_TF.ipynb b/translations/br/lessons/4-ComputerVision/08-TransferLearning/AdversarialCat_TF.ipynb index c3ab2259..8347fb60 100644 --- a/translations/br/lessons/4-ComputerVision/08-TransferLearning/AdversarialCat_TF.ipynb +++ b/translations/br/lessons/4-ComputerVision/08-TransferLearning/AdversarialCat_TF.ipynb @@ -50,7 +50,7 @@ "\n", "Para visualizar o gato ideal, começaremos com uma imagem de ruído aleatório e tentaremos usar a técnica de otimização por descida de gradiente para ajustar a imagem de forma que a rede reconheça um gato.\n", "\n", - "![Loop de Otimização](../../../../../translated_images/br/ideal-cat-loop.999fbb8ff306e044.webp)\n", + "![Loop de Otimização](../../../../../translated_images/pt-BR/ideal-cat-loop.999fbb8ff306e044.webp)\n", "\n", "Aqui está nossa imagem inicial:\n" ] diff --git a/translations/br/lessons/4-ComputerVision/08-TransferLearning/README.md b/translations/br/lessons/4-ComputerVision/08-TransferLearning/README.md index 3373fb51..dab61dd5 100644 --- a/translations/br/lessons/4-ComputerVision/08-TransferLearning/README.md +++ b/translations/br/lessons/4-ComputerVision/08-TransferLearning/README.md @@ -29,7 +29,7 @@ Tanto o Keras quanto o PyTorch possuem funções para carregar facilmente pesos Aqui estão características extraídas de uma imagem de um gato pela rede VGG-16: -![Características extraídas pelo VGG-16](../../../../../translated_images/br/features.6291f9c7ba3a0b95.webp) +![Características extraídas pelo VGG-16](../../../../../translated_images/pt-BR/features.6291f9c7ba3a0b95.webp) ## Conjunto de Dados de Gatos vs. Cachorros @@ -48,19 +48,19 @@ Uma rede neural pré-treinada contém diferentes padrões em seu *cérebro*, inc Uma abordagem que podemos adotar é começar com uma imagem aleatória e, em seguida, tentar usar a técnica de **otimização por descida de gradiente** para ajustar essa imagem de forma que a rede comece a pensar que é um gato. -![Loop de Otimização de Imagem](../../../../../translated_images/br/ideal-cat-loop.999fbb8ff306e044.webp) +![Loop de Otimização de Imagem](../../../../../translated_images/pt-BR/ideal-cat-loop.999fbb8ff306e044.webp) No entanto, se fizermos isso, receberemos algo muito semelhante a um ruído aleatório. Isso ocorre porque *existem muitas maneiras de fazer a rede pensar que a imagem de entrada é um gato*, incluindo algumas que não fazem sentido visualmente. Embora essas imagens contenham muitos padrões típicos de um gato, não há nada que as restrinja a serem visualmente distintas. Para melhorar o resultado, podemos adicionar outro termo à função de perda, chamado de **perda de variação**. É uma métrica que mostra quão semelhantes são os pixels vizinhos da imagem. Minimizar a perda de variação torna a imagem mais suave e elimina o ruído, revelando padrões mais visualmente atraentes. Aqui está um exemplo de tais imagens "ideais", classificadas como gato e zebra com alta probabilidade: -![Gato Ideal](../../../../../translated_images/br/ideal-cat.203dd4597643d6b0.webp) | ![Zebra Ideal](../../../../../translated_images/br/ideal-zebra.7f70e8b54ee15a7a.webp) +![Gato Ideal](../../../../../translated_images/pt-BR/ideal-cat.203dd4597643d6b0.webp) | ![Zebra Ideal](../../../../../translated_images/pt-BR/ideal-zebra.7f70e8b54ee15a7a.webp) -----|----- *Gato Ideal* | *Zebra Ideal* Uma abordagem semelhante pode ser usada para realizar os chamados **ataques adversariais** em uma rede neural. Suponha que queremos enganar uma rede neural e fazer um cachorro parecer um gato. Se pegarmos a imagem de um cachorro, que é reconhecida pela rede como um cachorro, podemos ajustá-la um pouco usando otimização por descida de gradiente até que a rede comece a classificá-la como um gato: -![Imagem de um Cachorro](../../../../../translated_images/br/original-dog.8f68a67d2fe0911f.webp) | ![Imagem de um cachorro classificada como gato](../../../../../translated_images/br/adversarial-dog.d9fc7773b0142b89.webp) +![Imagem de um Cachorro](../../../../../translated_images/pt-BR/original-dog.8f68a67d2fe0911f.webp) | ![Imagem de um cachorro classificada como gato](../../../../../translated_images/pt-BR/adversarial-dog.d9fc7773b0142b89.webp) -----|----- *Imagem original de um cachorro* | *Imagem de um cachorro classificada como gato* diff --git a/translations/br/lessons/4-ComputerVision/09-Autoencoders/AutoEncodersPyTorch.ipynb b/translations/br/lessons/4-ComputerVision/09-Autoencoders/AutoEncodersPyTorch.ipynb index eea3aa66..81e397d1 100644 --- a/translations/br/lessons/4-ComputerVision/09-Autoencoders/AutoEncodersPyTorch.ipynb +++ b/translations/br/lessons/4-ComputerVision/09-Autoencoders/AutoEncodersPyTorch.ipynb @@ -21,7 +21,7 @@ "\n", "Como estamos treinando o autoencoder para capturar o máximo de informações da imagem original possível para uma reconstrução precisa, a rede tenta encontrar a melhor **representação** das imagens de entrada para capturar o significado.\n", "\n", - "![Diagrama do AutoEncoder](../../../../../translated_images/br/autoencoder_schema.5e6fc9ad98a5eb61.webp)\n", + "![Diagrama do AutoEncoder](../../../../../translated_images/pt-BR/autoencoder_schema.5e6fc9ad98a5eb61.webp)\n", "\n", "> Imagem do [blog do Keras](https://blog.keras.io/building-autoencoders-in-keras.html)\n", "\n", diff --git a/translations/br/lessons/4-ComputerVision/09-Autoencoders/AutoencodersTF.ipynb b/translations/br/lessons/4-ComputerVision/09-Autoencoders/AutoencodersTF.ipynb index bd60fc03..b2fe4841 100644 --- a/translations/br/lessons/4-ComputerVision/09-Autoencoders/AutoencodersTF.ipynb +++ b/translations/br/lessons/4-ComputerVision/09-Autoencoders/AutoencodersTF.ipynb @@ -19,7 +19,7 @@ "\n", "Como estamos treinando o autoencoder para capturar o máximo de informações possíveis da imagem original para uma reconstrução precisa, a rede tenta encontrar o melhor **embedding** das imagens de entrada para capturar o significado.\n", "\n", - "![Diagrama do AutoEncoder](../../../../../translated_images/br/autoencoder_schema.5e6fc9ad98a5eb61.webp)\n", + "![Diagrama do AutoEncoder](../../../../../translated_images/pt-BR/autoencoder_schema.5e6fc9ad98a5eb61.webp)\n", "\n", "*Imagem do [blog do Keras](https://blog.keras.io/building-autoencoders-in-keras.html)*\n", "\n", diff --git a/translations/br/lessons/4-ComputerVision/09-Autoencoders/README.md b/translations/br/lessons/4-ComputerVision/09-Autoencoders/README.md index 9961c0cf..0cb83485 100644 --- a/translations/br/lessons/4-ComputerVision/09-Autoencoders/README.md +++ b/translations/br/lessons/4-ComputerVision/09-Autoencoders/README.md @@ -19,7 +19,7 @@ No entanto, podemos querer usar dados brutos (não rotulados) para treinar extra Como estamos treinando um autoencoder para capturar o máximo de informações da imagem original possível para uma reconstrução precisa, a rede tenta encontrar a melhor **representação** das imagens de entrada para capturar seu significado. -![Diagrama de AutoEncoder](../../../../../translated_images/br/autoencoder_schema.5e6fc9ad98a5eb61.webp) +![Diagrama de AutoEncoder](../../../../../translated_images/pt-BR/autoencoder_schema.5e6fc9ad98a5eb61.webp) > Imagem do [blog do Keras](https://blog.keras.io/building-autoencoders-in-keras.html) diff --git a/translations/br/lessons/4-ComputerVision/11-ObjectDetection/README.md b/translations/br/lessons/4-ComputerVision/11-ObjectDetection/README.md index 0b28ddc9..2bddc410 100644 --- a/translations/br/lessons/4-ComputerVision/11-ObjectDetection/README.md +++ b/translations/br/lessons/4-ComputerVision/11-ObjectDetection/README.md @@ -13,7 +13,7 @@ Os modelos de classificação de imagens que abordamos até agora tomavam uma im ## [Quiz pré-aula](https://ff-quizzes.netlify.app/en/ai/quiz/21) -![Detecção de Objetos](../../../../../translated_images/br/Screen_Shot_2016-11-17_at_11.14.54_AM.b4bb3769353287be.webp) +![Detecção de Objetos](../../../../../translated_images/pt-BR/Screen_Shot_2016-11-17_at_11.14.54_AM.b4bb3769353287be.webp) > Imagem do [site YOLO v2](https://pjreddie.com/darknet/yolov2/) @@ -25,7 +25,7 @@ Suponha que queremos encontrar um gato em uma imagem. Uma abordagem muito ingên 2. Executar a classificação de imagem em cada bloco. 3. Os blocos que resultarem em uma ativação suficientemente alta podem ser considerados como contendo o objeto em questão. -![Detecção Ingênua de Objetos](../../../../../translated_images/br/naive-detection.e7f1ba220ccd08c6.webp) +![Detecção Ingênua de Objetos](../../../../../translated_images/pt-BR/naive-detection.e7f1ba220ccd08c6.webp) > *Imagem do [Notebook de Exercícios](ObjectDetection-TF.ipynb)* @@ -42,7 +42,7 @@ Você pode encontrar os seguintes conjuntos de dados para essa tarefa: * [PASCAL VOC](http://host.robots.ox.ac.uk/pascal/VOC/) - 20 classes * [COCO](http://cocodataset.org/#home) - Objetos Comuns em Contexto. 80 classes, caixas delimitadoras e máscaras de segmentação -![COCO](../../../../../translated_images/br/coco-examples.71bc60380fa6cceb.webp) +![COCO](../../../../../translated_images/pt-BR/coco-examples.71bc60380fa6cceb.webp) ## Métricas de Detecção de Objetos @@ -50,7 +50,7 @@ Você pode encontrar os seguintes conjuntos de dados para essa tarefa: Enquanto na classificação de imagens é fácil medir o desempenho do algoritmo, na detecção de objetos precisamos medir tanto a correção da classe quanto a precisão da localização inferida da caixa delimitadora. Para este último, usamos a chamada **Interseção sobre União** (IoU), que mede o quão bem duas caixas (ou duas áreas arbitrárias) se sobrepõem. -![IoU](../../../../../translated_images/br/iou_equation.9a4751d40fff4e11.webp) +![IoU](../../../../../translated_images/pt-BR/iou_equation.9a4751d40fff4e11.webp) > *Figura 2 de [este excelente post sobre IoU](https://pyimagesearch.com/2016/11/07/intersection-over-union-iou-for-object-detection/)* @@ -98,11 +98,11 @@ Existem duas grandes classes de algoritmos de detecção de objetos: [R-CNN](http://islab.ulsan.ac.kr/files/announcement/513/rcnn_pami.pdf) usa [Selective Search](http://www.huppelen.nl/publications/selectiveSearchDraft.pdf) para gerar uma estrutura hierárquica de regiões ROI, que são então passadas por extratores de características CNN e classificadores SVM para determinar a classe do objeto, e regressão linear para determinar as coordenadas da *caixa delimitadora*. [Artigo Oficial](https://arxiv.org/pdf/1506.01497v1.pdf) -![RCNN](../../../../../translated_images/br/rcnn1.cae407020dfb1d1f.webp) +![RCNN](../../../../../translated_images/pt-BR/rcnn1.cae407020dfb1d1f.webp) > *Imagem de van de Sande et al. ICCV’11* -![RCNN-1](../../../../../translated_images/br/rcnn2.2d9530bb83516484.webp) +![RCNN-1](../../../../../translated_images/pt-BR/rcnn2.2d9530bb83516484.webp) > *Imagens de [este blog](https://towardsdatascience.com/r-cnn-fast-r-cnn-faster-r-cnn-yolo-object-detection-algorithms-36d53571365e)* @@ -110,7 +110,7 @@ Existem duas grandes classes de algoritmos de detecção de objetos: Essa abordagem é semelhante à R-CNN, mas as regiões são definidas após as camadas de convolução terem sido aplicadas. -![FRCNN](../../../../../translated_images/br/f-rcnn.3cda6d9bb4188875.webp) +![FRCNN](../../../../../translated_images/pt-BR/f-rcnn.3cda6d9bb4188875.webp) > Imagem do [Artigo Oficial](https://www.cv-foundation.org/openaccess/content_iccv_2015/papers/Girshick_Fast_R-CNN_ICCV_2015_paper.pdf), [arXiv](https://arxiv.org/pdf/1504.08083.pdf), 2015 @@ -118,7 +118,7 @@ Essa abordagem é semelhante à R-CNN, mas as regiões são definidas após as c A ideia principal dessa abordagem é usar uma rede neural para prever ROIs - a chamada *Rede de Proposta de Região*. [Artigo](https://arxiv.org/pdf/1506.01497.pdf), 2016 -![FasterRCNN](../../../../../translated_images/br/faster-rcnn.8d46c099b87ef30a.webp) +![FasterRCNN](../../../../../translated_images/pt-BR/faster-rcnn.8d46c099b87ef30a.webp) > Imagem do [Artigo Oficial](https://arxiv.org/pdf/1506.01497.pdf) @@ -130,7 +130,7 @@ Este algoritmo é ainda mais rápido que o Faster R-CNN. A ideia principal é a 1. As características são processadas por **Position-Sensitive Score Map**. Cada objeto de $C$ classes é dividido em regiões $k\times k$, e treinamos para prever partes dos objetos. 1. Para cada parte das regiões $k\times k$, todas as redes votam pelas classes de objetos, e a classe de objeto com o maior número de votos é selecionada. -![r-fcn image](../../../../../translated_images/br/r-fcn.13eb88158b99a3da.webp) +![r-fcn image](../../../../../translated_images/pt-BR/r-fcn.13eb88158b99a3da.webp) > Imagem do [Artigo Oficial](https://arxiv.org/abs/1605.06409) @@ -141,7 +141,7 @@ YOLO é um algoritmo de uma única passagem em tempo real. A ideia principal é * A imagem é dividida em regiões $S\times S$. * Para cada região, **CNN** prevê $n$ objetos possíveis, coordenadas da *caixa delimitadora* e *confiança*=*probabilidade* * IoU. - ![YOLO](../../../../../translated_images/br/yolo.a2648ec82ee8bb4e.webp) + ![YOLO](../../../../../translated_images/pt-BR/yolo.a2648ec82ee8bb4e.webp) > Imagem do [Artigo Oficial](https://arxiv.org/abs/1506.02640) diff --git a/translations/br/lessons/4-ComputerVision/README.md b/translations/br/lessons/4-ComputerVision/README.md index 06d0728f..1200029b 100644 --- a/translations/br/lessons/4-ComputerVision/README.md +++ b/translations/br/lessons/4-ComputerVision/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # Visão Computacional -![Resumo do conteúdo de Visão Computacional em um desenho](../../../../translated_images/br/ai-computervision.6506ebebac3fbf76.webp) +![Resumo do conteúdo de Visão Computacional em um desenho](../../../../translated_images/pt-BR/ai-computervision.6506ebebac3fbf76.webp) Nesta seção, vamos aprender sobre: diff --git a/translations/br/lessons/5-NLP/13-TextRep/TextRepresentationPyTorch.ipynb b/translations/br/lessons/5-NLP/13-TextRep/TextRepresentationPyTorch.ipynb index 39c28817..280122fc 100644 --- a/translations/br/lessons/5-NLP/13-TextRep/TextRepresentationPyTorch.ipynb +++ b/translations/br/lessons/5-NLP/13-TextRep/TextRepresentationPyTorch.ipynb @@ -199,7 +199,7 @@ "\n", "A representação vetorial **Bag of Words** (BoW) é a representação vetorial tradicional mais comumente usada. Cada palavra está vinculada a um índice do vetor, e o elemento do vetor contém o número de ocorrências de uma palavra em um determinado documento.\n", "\n", - "![Imagem mostrando como uma representação vetorial Bag of Words é representada na memória.](../../../../../translated_images/br/bag-of-words-example.606fc1738f1d7ba9.webp) \n", + "![Imagem mostrando como uma representação vetorial Bag of Words é representada na memória.](../../../../../translated_images/pt-BR/bag-of-words-example.606fc1738f1d7ba9.webp) \n", "\n", "> **Note**: Você também pode pensar no BoW como a soma de todos os vetores one-hot-encoded para palavras individuais no texto.\n", "\n", diff --git a/translations/br/lessons/5-NLP/13-TextRep/TextRepresentationTF.ipynb b/translations/br/lessons/5-NLP/13-TextRep/TextRepresentationTF.ipynb index 2ec103e7..f1d71c22 100644 --- a/translations/br/lessons/5-NLP/13-TextRep/TextRepresentationTF.ipynb +++ b/translations/br/lessons/5-NLP/13-TextRep/TextRepresentationTF.ipynb @@ -188,7 +188,7 @@ "\n", "A representação vetorial **Bag-of-words** (BoW) é a forma tradicional mais simples de entender uma representação vetorial. Cada palavra é vinculada a um índice no vetor, e um elemento do vetor contém o número de ocorrências de cada palavra em um determinado documento.\n", "\n", - "![Imagem mostrando como uma representação vetorial Bag-of-words é armazenada na memória.](../../../../../translated_images/br/bag-of-words-example.606fc1738f1d7ba9.webp) \n", + "![Imagem mostrando como uma representação vetorial Bag-of-words é armazenada na memória.](../../../../../translated_images/pt-BR/bag-of-words-example.606fc1738f1d7ba9.webp) \n", "\n", "> **Note**: Você também pode pensar no BoW como a soma de todos os vetores one-hot codificados para as palavras individuais no texto.\n", "\n", diff --git a/translations/br/lessons/5-NLP/14-Embeddings/EmbeddingsPyTorch.ipynb b/translations/br/lessons/5-NLP/14-Embeddings/EmbeddingsPyTorch.ipynb index 244795f0..5ad47a68 100644 --- a/translations/br/lessons/5-NLP/14-Embeddings/EmbeddingsPyTorch.ipynb +++ b/translations/br/lessons/5-NLP/14-Embeddings/EmbeddingsPyTorch.ipynb @@ -62,7 +62,7 @@ "\n", "Ao usar a camada de embedding como a primeira camada em nossa rede, podemos mudar do modelo bag-of-words para o modelo **embedding bag**, onde primeiro convertemos cada palavra em nosso texto no embedding correspondente e, em seguida, calculamos alguma função de agregação sobre todos esses embeddings, como `sum`, `average` ou `max`.\n", "\n", - "![Imagem mostrando um classificador de embedding para cinco palavras em sequência.](../../../../../translated_images/br/embedding-classifier-example.b77f021a7ee67eee.webp)\n", + "![Imagem mostrando um classificador de embedding para cinco palavras em sequência.](../../../../../translated_images/pt-BR/embedding-classifier-example.b77f021a7ee67eee.webp)\n", "\n", "Nossa rede neural classificadora começará com uma camada de embedding, seguida por uma camada de agregação e, por fim, um classificador linear no topo:\n" ] @@ -176,7 +176,7 @@ "\n", "Na arquitetura anterior, precisávamos preencher todas as sequências para que tivessem o mesmo comprimento, a fim de ajustá-las em um minibatch. Essa não é a maneira mais eficiente de representar sequências de comprimento variável - outra abordagem seria usar um vetor de **offset**, que armazenaria os deslocamentos de todas as sequências em um único vetor grande.\n", "\n", - "![Imagem mostrando uma representação de sequência com offset](../../../../../translated_images/br/offset-sequence-representation.eb73fcefb29b46ee.webp)\n", + "![Imagem mostrando uma representação de sequência com offset](../../../../../translated_images/pt-BR/offset-sequence-representation.eb73fcefb29b46ee.webp)\n", "\n", "> **Note**: Na imagem acima, mostramos uma sequência de caracteres, mas em nosso exemplo estamos trabalhando com sequências de palavras. No entanto, o princípio geral de representar sequências com um vetor de offset permanece o mesmo.\n", "\n", @@ -311,7 +311,7 @@ "\n", "CBoW é mais rápido, enquanto skip-gram é mais lento, mas faz um trabalho melhor ao representar palavras menos frequentes.\n", "\n", - "![Imagem mostrando os algoritmos CBoW e Skip-Gram para converter palavras em vetores.](../../../../../translated_images/br/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp)\n", + "![Imagem mostrando os algoritmos CBoW e Skip-Gram para converter palavras em vetores.](../../../../../translated_images/pt-BR/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp)\n", "\n", "Para experimentar com embeddings word2vec pré-treinados no conjunto de dados Google News, podemos usar a biblioteca **gensim**. Abaixo, encontramos as palavras mais semelhantes a 'neural'.\n", "\n", diff --git a/translations/br/lessons/5-NLP/14-Embeddings/EmbeddingsTF.ipynb b/translations/br/lessons/5-NLP/14-Embeddings/EmbeddingsTF.ipynb index 0e871d1f..2c590de2 100644 --- a/translations/br/lessons/5-NLP/14-Embeddings/EmbeddingsTF.ipynb +++ b/translations/br/lessons/5-NLP/14-Embeddings/EmbeddingsTF.ipynb @@ -37,7 +37,7 @@ "\n", "Ao usar uma camada de embedding como a primeira camada da nossa rede, podemos mudar de um modelo de bag-of-words para um modelo de **embedding bag**, onde primeiro convertemos cada palavra do nosso texto no embedding correspondente e, em seguida, calculamos alguma função de agregação sobre todos esses embeddings, como `sum`, `average` ou `max`.\n", "\n", - "![Imagem mostrando um classificador com embedding para cinco palavras em sequência.](../../../../../translated_images/br/embedding-classifier-example.b77f021a7ee67eee.webp)\n", + "![Imagem mostrando um classificador com embedding para cinco palavras em sequência.](../../../../../translated_images/pt-BR/embedding-classifier-example.b77f021a7ee67eee.webp)\n", "\n", "Nossa rede neural classificadora consiste nas seguintes camadas:\n", "\n", @@ -283,7 +283,7 @@ "\n", "CBoW é mais rápido, enquanto o skip-gram, embora mais lento, faz um trabalho melhor ao representar palavras menos frequentes.\n", "\n", - "![Imagem mostrando os algoritmos CBoW e Skip-Gram para converter palavras em vetores.](../../../../../translated_images/br/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp)\n", + "![Imagem mostrando os algoritmos CBoW e Skip-Gram para converter palavras em vetores.](../../../../../translated_images/pt-BR/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp)\n", "\n", "Para experimentar o embedding Word2Vec pré-treinado no conjunto de dados do Google News, podemos usar a biblioteca **gensim**. Abaixo, encontramos as palavras mais semelhantes a 'neural'.\n", "\n", diff --git a/translations/br/lessons/5-NLP/14-Embeddings/README.md b/translations/br/lessons/5-NLP/14-Embeddings/README.md index 53035ff0..bd234136 100644 --- a/translations/br/lessons/5-NLP/14-Embeddings/README.md +++ b/translations/br/lessons/5-NLP/14-Embeddings/README.md @@ -19,7 +19,7 @@ Assim, a camada de embedding receberia uma palavra como entrada e produziria um Ao usar uma camada de embedding como a primeira camada em nossa rede de classificação, podemos mudar de um modelo bag-of-words para um modelo **embedding bag**, onde primeiro convertemos cada palavra em nosso texto no embedding correspondente e, em seguida, calculamos alguma função agregada sobre todos esses embeddings, como `sum`, `average` ou `max`. -![Imagem mostrando um classificador de embedding para cinco palavras de sequência.](../../../../../translated_images/br/embedding-classifier-example.b77f021a7ee67eee.webp) +![Imagem mostrando um classificador de embedding para cinco palavras de sequência.](../../../../../translated_images/pt-BR/embedding-classifier-example.b77f021a7ee67eee.webp) > Imagem do autor @@ -40,7 +40,7 @@ Para isso, precisamos pré-treinar nosso modelo de embedding em uma grande cole CBoW é mais rápido, enquanto skip-gram é mais lento, mas faz um trabalho melhor ao representar palavras menos frequentes. -![Imagem mostrando os algoritmos CBoW e Skip-Gram para converter palavras em vetores.](../../../../../translated_images/br/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) +![Imagem mostrando os algoritmos CBoW e Skip-Gram para converter palavras em vetores.](../../../../../translated_images/pt-BR/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) > Imagem retirada [deste artigo](https://arxiv.org/pdf/1301.3781.pdf) diff --git a/translations/br/lessons/5-NLP/15-LanguageModeling/README.md b/translations/br/lessons/5-NLP/15-LanguageModeling/README.md index 65639386..d7f6bab2 100644 --- a/translations/br/lessons/5-NLP/15-LanguageModeling/README.md +++ b/translations/br/lessons/5-NLP/15-LanguageModeling/README.md @@ -23,7 +23,7 @@ Nos nossos exemplos anteriores, usamos embeddings semânticos pré-treinados, ma * **Continuous Bag-of-Words** (CBoW), onde prevemos o token do meio $W_0$ em uma sequência de tokens $W_{-N}$, ..., $W_N$. * **Skip-gram**, onde prevemos um conjunto de tokens vizinhos {$W_{-N},\dots, W_{-1}, W_1,\dots, W_N$} a partir do token do meio $W_0$. -![imagem do artigo sobre conversão de palavras em vetores](../../../../../translated_images/br/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) +![imagem do artigo sobre conversão de palavras em vetores](../../../../../translated_images/pt-BR/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) > Imagem retirada [deste artigo](https://arxiv.org/pdf/1301.3781.pdf) diff --git a/translations/br/lessons/5-NLP/16-RNN/README.md b/translations/br/lessons/5-NLP/16-RNN/README.md index d19ef8eb..2b86cebf 100644 --- a/translations/br/lessons/5-NLP/16-RNN/README.md +++ b/translations/br/lessons/5-NLP/16-RNN/README.md @@ -15,7 +15,7 @@ Nas seções anteriores, utilizamos representações semânticas ricas de texto Para capturar o significado de uma sequência de texto, precisamos usar outra arquitetura de rede neural, chamada de **rede neural recorrente**, ou RNN. Na RNN, passamos nossa frase pela rede um símbolo de cada vez, e a rede produz algum **estado**, que então passamos novamente para a rede junto com o próximo símbolo. -![RNN](../../../../../translated_images/br/rnn.27f5c29c53d727b5.webp) +![RNN](../../../../../translated_images/pt-BR/rnn.27f5c29c53d727b5.webp) > Imagem do autor @@ -61,7 +61,7 @@ Discutimos redes recorrentes que operam em uma direção, do início de uma sequ Uma rede recorrente, seja unidirecional ou bidirecional, captura certos padrões dentro de uma sequência e pode armazená-los em um vetor de estado ou passá-los para a saída. Assim como nas redes convolucionais, podemos construir outra camada recorrente sobre a primeira para capturar padrões de nível superior e construir a partir dos padrões de baixo nível extraídos pela primeira camada. Isso nos leva à noção de uma **RNN multicamada**, que consiste em duas ou mais redes recorrentes, onde a saída da camada anterior é passada para a próxima camada como entrada. -![Imagem mostrando uma RNN LSTM multicamada](../../../../../translated_images/br/multi-layer-lstm.dd975e29bb2a59fe.webp) +![Imagem mostrando uma RNN LSTM multicamada](../../../../../translated_images/pt-BR/multi-layer-lstm.dd975e29bb2a59fe.webp) *Imagem retirada [deste post maravilhoso](https://towardsdatascience.com/from-a-lstm-cell-to-a-multilayer-lstm-network-with-pytorch-2899eb5696f3) de Fernando López* diff --git a/translations/br/lessons/5-NLP/16-RNN/RNNPyTorch.ipynb b/translations/br/lessons/5-NLP/16-RNN/RNNPyTorch.ipynb index a3ca811b..f00fff11 100644 --- a/translations/br/lessons/5-NLP/16-RNN/RNNPyTorch.ipynb +++ b/translations/br/lessons/5-NLP/16-RNN/RNNPyTorch.ipynb @@ -422,7 +422,7 @@ "\n", "Uma rede recorrente, seja unidirecional ou bidirecional, captura certos padrões dentro de uma sequência e pode armazená-los no vetor de estado ou passá-los para a saída. Assim como nas redes convolucionais, podemos construir outra camada recorrente sobre a primeira para capturar padrões de nível mais alto, construídos a partir de padrões de baixo nível extraídos pela primeira camada. Isso nos leva à noção de **RNN multicamada**, que consiste em duas ou mais redes recorrentes, onde a saída da camada anterior é passada para a próxima camada como entrada.\n", "\n", - "![Imagem mostrando uma RNN LSTM multicamada](../../../../../translated_images/br/multi-layer-lstm.dd975e29bb2a59fe.webp)\n", + "![Imagem mostrando uma RNN LSTM multicamada](../../../../../translated_images/pt-BR/multi-layer-lstm.dd975e29bb2a59fe.webp)\n", "\n", "*Imagem retirada [deste post maravilhoso](https://towardsdatascience.com/from-a-lstm-cell-to-a-multilayer-lstm-network-with-pytorch-2899eb5696f3) por Fernando López*\n", "\n", diff --git a/translations/br/lessons/5-NLP/16-RNN/RNNTF.ipynb b/translations/br/lessons/5-NLP/16-RNN/RNNTF.ipynb index 0ad61fde..01210e71 100644 --- a/translations/br/lessons/5-NLP/16-RNN/RNNTF.ipynb +++ b/translations/br/lessons/5-NLP/16-RNN/RNNTF.ipynb @@ -10,7 +10,7 @@ "\n", "Para capturar o significado de uma sequência de texto, usaremos uma arquitetura de rede neural chamada **rede neural recorrente**, ou RNN. Ao usar uma RNN, passamos nossa sentença pela rede um token de cada vez, e a rede produz algum **estado**, que então passamos novamente para a rede junto com o próximo token.\n", "\n", - "![Imagem mostrando um exemplo de geração de rede neural recorrente.](../../../../../translated_images/br/rnn.27f5c29c53d727b5.webp)\n", + "![Imagem mostrando um exemplo de geração de rede neural recorrente.](../../../../../translated_images/pt-BR/rnn.27f5c29c53d727b5.webp)\n", "\n", "Dada a sequência de entrada de tokens $X_0,\\dots,X_n$, a RNN cria uma sequência de blocos de rede neural e treina essa sequência de ponta a ponta usando retropropagação. Cada bloco de rede recebe um par $(X_i,S_i)$ como entrada e produz $S_{i+1}$ como resultado. O estado final $S_n$ ou a saída $Y_n$ é enviado para um classificador linear para produzir o resultado. Todos os blocos de rede compartilham os mesmos pesos e são treinados de ponta a ponta usando uma única passagem de retropropagação.\n", "\n", @@ -369,7 +369,7 @@ "\n", "Redes recorrentes, sejam unidirecionais ou bidirecionais, capturam padrões dentro de uma sequência e os armazenam em vetores de estado ou os retornam como saída. Assim como nas redes convolucionais, podemos construir outra camada recorrente após a primeira para capturar padrões de nível mais alto, construídos a partir de padrões de nível mais baixo extraídos pela primeira camada. Isso nos leva à noção de uma **RNN multicamada**, que consiste em duas ou mais redes recorrentes, onde a saída da camada anterior é passada para a próxima camada como entrada.\n", "\n", - "![Imagem mostrando uma RNN multicamada de memória de longo e curto prazo](../../../../../translated_images/br/multi-layer-lstm.dd975e29bb2a59fe.webp)\n", + "![Imagem mostrando uma RNN multicamada de memória de longo e curto prazo](../../../../../translated_images/pt-BR/multi-layer-lstm.dd975e29bb2a59fe.webp)\n", "\n", "*Imagem retirada [deste post incrível](https://towardsdatascience.com/from-a-lstm-cell-to-a-multilayer-lstm-network-with-pytorch-2899eb5696f3) por Fernando López.*\n", "\n", diff --git a/translations/br/lessons/5-NLP/17-GenerativeNetworks/GenerativePyTorch.ipynb b/translations/br/lessons/5-NLP/17-GenerativeNetworks/GenerativePyTorch.ipynb index af3e7725..fe01e849 100644 --- a/translations/br/lessons/5-NLP/17-GenerativeNetworks/GenerativePyTorch.ipynb +++ b/translations/br/lessons/5-NLP/17-GenerativeNetworks/GenerativePyTorch.ipynb @@ -119,7 +119,7 @@ "\n", "A maneira como treinaremos a RNN para gerar texto é a seguinte. A cada etapa, pegaremos uma sequência de caracteres de comprimento `nchars` e pediremos à rede que gere o próximo caractere de saída para cada caractere de entrada:\n", "\n", - "![Imagem mostrando um exemplo de geração de RNN com a palavra 'HELLO'.](../../../../../translated_images/br/rnn-generate.56c54afb52f9781d.webp)\n", + "![Imagem mostrando um exemplo de geração de RNN com a palavra 'HELLO'.](../../../../../translated_images/pt-BR/rnn-generate.56c54afb52f9781d.webp)\n", "\n", "Dependendo do cenário real, também podemos querer incluir alguns caracteres especiais, como *fim de sequência* ``. No nosso caso, queremos apenas treinar a rede para geração contínua de texto, então fixaremos o tamanho de cada sequência para ser igual a `nchars` tokens. Consequentemente, cada exemplo de treinamento consistirá em `nchars` entradas e `nchars` saídas (que são a sequência de entrada deslocada um símbolo para a esquerda). O minibatch consistirá de várias dessas sequências.\n", "\n", diff --git a/translations/br/lessons/5-NLP/17-GenerativeNetworks/GenerativeTF.ipynb b/translations/br/lessons/5-NLP/17-GenerativeNetworks/GenerativeTF.ipynb index 54a63bd7..ee5594d6 100644 --- a/translations/br/lessons/5-NLP/17-GenerativeNetworks/GenerativeTF.ipynb +++ b/translations/br/lessons/5-NLP/17-GenerativeNetworks/GenerativeTF.ipynb @@ -111,7 +111,7 @@ "\n", "A maneira como treinaremos a RNN para gerar títulos de notícias é a seguinte. A cada etapa, pegaremos um título, que será alimentado em uma RNN, e para cada caractere de entrada pediremos à rede que gere o próximo caractere de saída:\n", "\n", - "![Imagem mostrando um exemplo de geração de RNN da palavra 'HELLO'.](../../../../../translated_images/br/rnn-generate.56c54afb52f9781d.webp)\n", + "![Imagem mostrando um exemplo de geração de RNN da palavra 'HELLO'.](../../../../../translated_images/pt-BR/rnn-generate.56c54afb52f9781d.webp)\n", "\n", "Para o último caractere da nossa sequência, pediremos à rede que gere o token ``.\n", "\n", diff --git a/translations/br/lessons/5-NLP/17-GenerativeNetworks/README.md b/translations/br/lessons/5-NLP/17-GenerativeNetworks/README.md index 856b4e34..b6c8a70f 100644 --- a/translations/br/lessons/5-NLP/17-GenerativeNetworks/README.md +++ b/translations/br/lessons/5-NLP/17-GenerativeNetworks/README.md @@ -19,7 +19,7 @@ Na arquitetura de RNN que discutimos na unidade anterior, cada unidade RNN produ Isso permite diferentes arquiteturas neurais, como mostrado na imagem abaixo: -![Imagem mostrando padrões comuns de redes neurais recorrentes.](../../../../../translated_images/br/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) +![Imagem mostrando padrões comuns de redes neurais recorrentes.](../../../../../translated_images/pt-BR/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) > Imagem do post no blog [Unreasonable Effectiveness of Recurrent Neural Networks](http://karpathy.github.io/2015/05/21/rnn-effectiveness/) por [Andrej Karpaty](http://karpathy.github.io/) @@ -32,7 +32,7 @@ Nesta unidade, focaremos em modelos generativos simples que nos ajudam a gerar t Treinaremos esta RNN para gerar texto passo a passo. Em cada etapa, pegaremos uma sequência de caracteres de comprimento `nchars` e pediremos à rede que gere o próximo caractere de saída para cada caractere de entrada: -![Imagem mostrando um exemplo de geração de RNN da palavra 'HELLO'.](../../../../../translated_images/br/rnn-generate.56c54afb52f9781d.webp) +![Imagem mostrando um exemplo de geração de RNN da palavra 'HELLO'.](../../../../../translated_images/pt-BR/rnn-generate.56c54afb52f9781d.webp) Ao gerar texto (durante a inferência), começamos com algum **prompt**, que é passado pelas células RNN para gerar seu estado intermediário, e então a geração começa a partir desse estado. Geramos um caractere por vez e passamos o estado e o caractere gerado para outra célula RNN para gerar o próximo, até que tenhamos gerado caracteres suficientes. diff --git a/translations/br/lessons/5-NLP/18-Transformers/README.md b/translations/br/lessons/5-NLP/18-Transformers/README.md index 8dc8f499..5fe3d65d 100644 --- a/translations/br/lessons/5-NLP/18-Transformers/README.md +++ b/translations/br/lessons/5-NLP/18-Transformers/README.md @@ -20,13 +20,13 @@ Com RNNs, a tarefa de sequência para sequência é implementada por duas redes Os **Mecanismos de Atenção** fornecem um meio de ponderar o impacto contextual de cada vetor de entrada em cada previsão de saída da RNN. Isso é implementado criando atalhos entre os estados intermediários da RNN de entrada e a RNN de saída. Dessa forma, ao gerar o símbolo de saída yt, levamos em conta todos os estados ocultos de entrada hi, com diferentes coeficientes de peso αt,i. -![Imagem mostrando um modelo codificador/decodificador com uma camada de atenção aditiva](../../../../../translated_images/br/encoder-decoder-attention.7a726296894fb567.webp) +![Imagem mostrando um modelo codificador/decodificador com uma camada de atenção aditiva](../../../../../translated_images/pt-BR/encoder-decoder-attention.7a726296894fb567.webp) > O modelo codificador-decodificador com mecanismo de atenção aditiva em [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf), citado deste [post no blog](https://lilianweng.github.io/lil-log/2018/06/24/attention-attention.html) A matriz de atenção {αi,j} representaria o grau em que certas palavras de entrada influenciam a geração de uma palavra específica na sequência de saída. Abaixo está um exemplo de tal matriz: -![Imagem mostrando um alinhamento de exemplo encontrado pelo RNNsearch-50, retirada de Bahdanau - arviz.org](../../../../../translated_images/br/bahdanau-fig3.09ba2d37f202a6af.webp) +![Imagem mostrando um alinhamento de exemplo encontrado pelo RNNsearch-50, retirada de Bahdanau - arviz.org](../../../../../translated_images/pt-BR/bahdanau-fig3.09ba2d37f202a6af.webp) > Figura de [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf) (Fig.3) @@ -66,7 +66,7 @@ O resultado que obtemos com o embutimento posicional incorpora tanto o token ori Em seguida, precisamos capturar alguns padrões dentro de nossa sequência. Para isso, os transformers usam um mecanismo de **autoatenção**, que é essencialmente atenção aplicada à mesma sequência como entrada e saída. Aplicar autoatenção nos permite levar em conta o **contexto** dentro da sentença e ver quais palavras estão inter-relacionadas. Por exemplo, isso nos permite identificar quais palavras são referidas por correferências, como *it* (ele/ela), e também considerar o contexto: -![](../../../../../translated_images/br/CoreferenceResolution.861924d6d384a7d6.webp) +![](../../../../../translated_images/pt-BR/CoreferenceResolution.861924d6d384a7d6.webp) > Imagem do [Blog do Google](https://research.googleblog.com/2017/08/transformer-novel-neural-network.html) @@ -91,7 +91,7 @@ Como cada posição de entrada é mapeada independentemente para cada posição **BERT** (Bidirectional Encoder Representations from Transformers) é uma rede transformer muito grande com várias camadas: 12 camadas para o *BERT-base* e 24 para o *BERT-large*. O modelo é primeiro pré-treinado em um grande corpus de dados textuais (Wikipedia + livros) usando treinamento não supervisionado (prevendo palavras mascaradas em uma sentença). Durante o pré-treinamento, o modelo absorve níveis significativos de compreensão da linguagem, que podem ser aproveitados com outros conjuntos de dados usando ajuste fino. Esse processo é chamado de **aprendizado por transferência**. -![imagem de http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/br/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp) +![imagem de http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/pt-BR/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp) > Imagem [fonte](http://jalammar.github.io/illustrated-bert/) diff --git a/translations/br/lessons/5-NLP/18-Transformers/TransformersPyTorch.ipynb b/translations/br/lessons/5-NLP/18-Transformers/TransformersPyTorch.ipynb index 6030dcd4..ec281be7 100644 --- a/translations/br/lessons/5-NLP/18-Transformers/TransformersPyTorch.ipynb +++ b/translations/br/lessons/5-NLP/18-Transformers/TransformersPyTorch.ipynb @@ -12,12 +12,12 @@ "\n", "**Mecanismos de Atenção** fornecem um meio de ponderar o impacto contextual de cada vetor de entrada em cada previsão de saída da RNN. Isso é implementado criando atalhos entre estados intermediários da RNN de entrada e da RNN de saída. Dessa forma, ao gerar o símbolo de saída $y_t$, levaremos em conta todos os estados ocultos de entrada $h_i$, com diferentes coeficientes de peso $\\alpha_{t,i}$.\n", "\n", - "![Imagem mostrando um modelo codificador/decodificador com uma camada de atenção aditiva](../../../../../translated_images/br/encoder-decoder-attention.7a726296894fb567.webp)\n", + "![Imagem mostrando um modelo codificador/decodificador com uma camada de atenção aditiva](../../../../../translated_images/pt-BR/encoder-decoder-attention.7a726296894fb567.webp)\n", "*O modelo codificador-decodificador com mecanismo de atenção aditiva em [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf), citado deste [post de blog](https://lilianweng.github.io/lil-log/2018/06/24/attention-attention.html)*\n", "\n", "A matriz de atenção $\\{\\alpha_{i,j}\\}$ representaria o grau em que certas palavras de entrada influenciam a geração de uma palavra específica na sequência de saída. Abaixo está um exemplo de tal matriz:\n", "\n", - "![Imagem mostrando um alinhamento de exemplo encontrado pelo RNNsearch-50, retirada de Bahdanau - arviz.org](../../../../../translated_images/br/bahdanau-fig3.09ba2d37f202a6af.webp)\n", + "![Imagem mostrando um alinhamento de exemplo encontrado pelo RNNsearch-50, retirada de Bahdanau - arviz.org](../../../../../translated_images/pt-BR/bahdanau-fig3.09ba2d37f202a6af.webp)\n", "\n", "*Figura retirada de [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf) (Fig.3)*\n", "\n", @@ -35,7 +35,7 @@ "\n", "**BERT** (Representações de Codificador Bidirecional de Transformadores) é uma rede transformadora muito grande e multilayer com 12 camadas para *BERT-base* e 24 para *BERT-large*. O modelo é primeiro pré-treinado em um grande corpus de dados de texto (WikiPedia + livros) usando treinamento não supervisionado (prevendo palavras mascaradas em uma frase). Durante o pré-treinamento, o modelo absorve um nível significativo de compreensão da linguagem, que pode ser aproveitado com outros conjuntos de dados usando ajuste fino. Esse processo é chamado de **aprendizado por transferência**.\n", "\n", - "![Imagem de http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/br/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp)\n", + "![Imagem de http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/pt-BR/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp)\n", "\n", "Existem muitas variações de arquiteturas de transformadores, incluindo BERT, DistilBERT, BigBird, OpenGPT3 e outras que podem ser ajustadas. O pacote [HuggingFace](https://github.com/huggingface/) fornece um repositório para treinar muitas dessas arquiteturas com PyTorch.\n", "\n", diff --git a/translations/br/lessons/5-NLP/18-Transformers/TransformersTF.ipynb b/translations/br/lessons/5-NLP/18-Transformers/TransformersTF.ipynb index 4f6024b7..31beafe1 100644 --- a/translations/br/lessons/5-NLP/18-Transformers/TransformersTF.ipynb +++ b/translations/br/lessons/5-NLP/18-Transformers/TransformersTF.ipynb @@ -12,12 +12,12 @@ "\n", "**Mecanismos de Atenção** fornecem um meio de ponderar o impacto contextual de cada vetor de entrada em cada previsão de saída da RNN. Isso é implementado criando atalhos entre os estados intermediários da RNN de entrada e a RNN de saída. Dessa forma, ao gerar o símbolo de saída $y_t$, levaremos em conta todos os estados ocultos de entrada $h_i$, com diferentes coeficientes de peso $\\alpha_{t,i}$.\n", "\n", - "![Imagem mostrando um modelo codificador/decodificador com uma camada de atenção aditiva](../../../../../translated_images/br/encoder-decoder-attention.7a726296894fb567.webp)\n", + "![Imagem mostrando um modelo codificador/decodificador com uma camada de atenção aditiva](../../../../../translated_images/pt-BR/encoder-decoder-attention.7a726296894fb567.webp)\n", "*O modelo codificador-decodificador com mecanismo de atenção aditiva em [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf), citado deste [post de blog](https://lilianweng.github.io/lil-log/2018/06/24/attention-attention.html)*\n", "\n", "A matriz de atenção $\\{\\alpha_{i,j}\\}$ representaria o grau em que certas palavras de entrada influenciam a geração de uma determinada palavra na sequência de saída. Abaixo está um exemplo de tal matriz:\n", "\n", - "![Imagem mostrando um alinhamento de exemplo encontrado pelo RNNsearch-50, retirada de Bahdanau - arviz.org](../../../../../translated_images/br/bahdanau-fig3.09ba2d37f202a6af.webp)\n", + "![Imagem mostrando um alinhamento de exemplo encontrado pelo RNNsearch-50, retirada de Bahdanau - arviz.org](../../../../../translated_images/pt-BR/bahdanau-fig3.09ba2d37f202a6af.webp)\n", "\n", "*Figura retirada de [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf) (Fig.3)*\n", "\n", @@ -225,7 +225,7 @@ "\n", "**BERT** (Representações de Codificadores Bidirecionais de Transformers) é uma rede transformer muito grande e multilayer, com 12 camadas para o *BERT-base* e 24 para o *BERT-large*. O modelo é inicialmente pré-treinado em um grande corpus de dados textuais (WikiPedia + livros) usando treinamento não supervisionado (prevendo palavras mascaradas em uma sentença). Durante o pré-treinamento, o modelo adquire um nível significativo de compreensão da linguagem, que pode ser aproveitado com outros conjuntos de dados por meio de ajuste fino. Esse processo é chamado de **aprendizado por transferência**.\n", "\n", - "![imagem de http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/br/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp)\n", + "![imagem de http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/pt-BR/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp)\n", "\n", "Existem muitas variações de arquiteturas Transformer, incluindo BERT, DistilBERT, BigBird, OpenGPT3 e outras, que podem ser ajustadas.\n", "\n", diff --git a/translations/br/lessons/5-NLP/19-NER/README.md b/translations/br/lessons/5-NLP/19-NER/README.md index 911f3e8d..861e2bfc 100644 --- a/translations/br/lessons/5-NLP/19-NER/README.md +++ b/translations/br/lessons/5-NLP/19-NER/README.md @@ -58,7 +58,7 @@ infant | O Como precisamos construir uma correspondência um-para-um entre tokens e classes, podemos treinar um modelo neural **muitos-para-muitos** da seguinte forma: -![Imagem mostrando padrões comuns de redes neurais recorrentes.](../../../../../translated_images/br/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) +![Imagem mostrando padrões comuns de redes neurais recorrentes.](../../../../../translated_images/pt-BR/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) > *Imagem do [post no blog](http://karpathy.github.io/2015/05/21/rnn-effectiveness/) de [Andrej Karpathy](http://karpathy.github.io/). Os modelos de classificação de tokens NER correspondem à arquitetura de rede mais à direita nesta imagem.* diff --git a/translations/br/lessons/5-NLP/README.md b/translations/br/lessons/5-NLP/README.md index 8e044804..88ab21cd 100644 --- a/translations/br/lessons/5-NLP/README.md +++ b/translations/br/lessons/5-NLP/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # Processamento de Linguagem Natural -![Resumo das tarefas de PLN em um desenho](../../../../translated_images/br/ai-nlp.b22dcb8ca4707cea.webp) +![Resumo das tarefas de PLN em um desenho](../../../../translated_images/pt-BR/ai-nlp.b22dcb8ca4707cea.webp) Nesta seção, vamos focar no uso de Redes Neurais para lidar com tarefas relacionadas ao **Processamento de Linguagem Natural (PLN)**. Existem muitos problemas de PLN que queremos que os computadores sejam capazes de resolver: diff --git a/translations/br/lessons/6-Other/23-MultiagentSystems/README.md b/translations/br/lessons/6-Other/23-MultiagentSystems/README.md index b54718f7..017230e6 100644 --- a/translations/br/lessons/6-Other/23-MultiagentSystems/README.md +++ b/translations/br/lessons/6-Other/23-MultiagentSystems/README.md @@ -70,7 +70,7 @@ Você pode abrir um dos modelos, por exemplo **Biology → Flocking**. Após abrir o modelo, você será levado à tela principal do NetLogo. Aqui está um modelo de exemplo que descreve a população de lobos e ovelhas, considerando recursos finitos (grama). -![Tela Principal do NetLogo](../../../../../translated_images/br/NetLogo-Main.32653711ec1a01b3.webp) +![Tela Principal do NetLogo](../../../../../translated_images/pt-BR/NetLogo-Main.32653711ec1a01b3.webp) > Captura de tela por Dmitry Soshnikov diff --git a/translations/br/lessons/README.md b/translations/br/lessons/README.md index 5fb6dff0..f0b6aeba 100644 --- a/translations/br/lessons/README.md +++ b/translations/br/lessons/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # Visão Geral -![Visão Geral em um rabisco](../../../translated_images/br/ai-overview.0857791951d19500.webp) +![Visão Geral em um rabisco](../../../translated_images/pt-BR/ai-overview.0857791951d19500.webp) > Rabisco por [Tomomi Imura](https://twitter.com/girlie_mac) diff --git a/translations/br/lessons/X-Extras/X1-MultiModal/README.md b/translations/br/lessons/X-Extras/X1-MultiModal/README.md index 1549f17f..067533cf 100644 --- a/translations/br/lessons/X-Extras/X1-MultiModal/README.md +++ b/translations/br/lessons/X-Extras/X1-MultiModal/README.md @@ -15,7 +15,7 @@ Após o sucesso dos modelos transformers na resolução de tarefas de PLN, as me A ideia principal do CLIP é ser capaz de comparar descrições textuais com uma imagem e determinar o quão bem a imagem corresponde à descrição. -![Arquitetura do CLIP](../../../../../translated_images/br/clip-arch.b3dbf20b4e8ed8be.webp) +![Arquitetura do CLIP](../../../../../translated_images/pt-BR/clip-arch.b3dbf20b4e8ed8be.webp) > *Imagem retirada [deste post no blog](https://openai.com/blog/clip/)* @@ -29,7 +29,7 @@ Uma vez que este modelo é pré-treinado, podemos fornecer a ele um lote de imag Suponha que precisamos classificar imagens entre, por exemplo, gatos, cachorros e humanos. Nesse caso, podemos fornecer ao modelo uma imagem e uma série de descrições textuais: "*uma foto de um gato*", "*uma foto de um cachorro*", "*uma foto de um humano*". No vetor resultante de 3 probabilidades, basta selecionar o índice com o maior valor. -![CLIP para Classificação de Imagens](../../../../../translated_images/br/clip-class.3af42ef0b2b19369.webp) +![CLIP para Classificação de Imagens](../../../../../translated_images/pt-BR/clip-class.3af42ef0b2b19369.webp) > *Imagem retirada [deste post no blog](https://openai.com/blog/clip/)* @@ -53,13 +53,13 @@ Saiba mais sobre o VQGAN no site [Taming Transformers](https://compvis.github.io Uma das diferenças importantes entre o VQGAN e um GAN tradicional é que o último pode produzir uma imagem decente a partir de qualquer vetor de entrada, enquanto o VQGAN provavelmente produzirá uma imagem incoerente. Assim, precisamos orientar ainda mais o processo de criação da imagem, e isso pode ser feito usando o CLIP. -![Arquitetura VQGAN+CLIP](../../../../../translated_images/br/vqgan.5027fe05051dfa31.webp) +![Arquitetura VQGAN+CLIP](../../../../../translated_images/pt-BR/vqgan.5027fe05051dfa31.webp) Para gerar uma imagem correspondente a uma descrição textual, começamos com algum vetor de codificação aleatório que é passado pelo VQGAN para produzir uma imagem. Em seguida, o CLIP é usado para produzir uma função de perda que mostra o quão bem a imagem corresponde à descrição textual. O objetivo, então, é minimizar essa perda, usando retropropagação para ajustar os parâmetros do vetor de entrada. Uma ótima biblioteca que implementa o VQGAN+CLIP é o [Pixray](http://github.com/pixray/pixray). -![Imagem gerada pelo Pixray](../../../../../translated_images/br/a_closeup_watercolor_portrait_of_young_male_teacher_of_literature_with_a_book.2384968e9db8a0d0.webp) | ![Imagem gerada pelo Pixray](../../../../../translated_images/br/a_closeup_oil_portrait_of_young_female_teacher_of_computer_science_with_a_computer.e0b6495f210a4390.webp) | ![Imagem gerada pelo Pixray](../../../../../translated_images/br/a_closeup_oil_portrait_of_old_male_teacher_of_math.5362e67aa7fc2683.webp) +![Imagem gerada pelo Pixray](../../../../../translated_images/pt-BR/a_closeup_watercolor_portrait_of_young_male_teacher_of_literature_with_a_book.2384968e9db8a0d0.webp) | ![Imagem gerada pelo Pixray](../../../../../translated_images/pt-BR/a_closeup_oil_portrait_of_young_female_teacher_of_computer_science_with_a_computer.e0b6495f210a4390.webp) | ![Imagem gerada pelo Pixray](../../../../../translated_images/pt-BR/a_closeup_oil_portrait_of_old_male_teacher_of_math.5362e67aa7fc2683.webp) ----|----|---- Imagem gerada a partir da descrição *um retrato em aquarela de um jovem professor de literatura com um livro* | Imagem gerada a partir da descrição *um retrato a óleo de uma jovem professora de ciência da computação com um computador* | Imagem gerada a partir da descrição *um retrato a óleo de um velho professor de matemática em frente a um quadro-negro* @@ -75,7 +75,7 @@ Diferentemente do CLIP, o DALL-E recebe tanto texto quanto imagem como um único A principal diferença entre o DALL-E 1 e o 2 é que o último gera imagens e artes mais realistas. Exemplos de imagens geradas com o DALL-E: -![Imagem gerada pelo DALL-E](../../../../../translated_images/br/DALL·E%202023-06-20%2015.56.56%20-%20a%20closeup%20watercolor%20portrait%20of%20young%20male%20teacher%20of%20literature%20with%20a%20book.6c235e8271d9ed10ce985d86aeb241a58518958647973af136912116b9518fce.png) | ![Imagem gerada pelo DALL-E](../../../../../translated_images/br/DALL·E%202023-06-20%2015.57.43%20-%20a%20closeup%20oil%20portrait%20of%20young%20female%20teacher%20of%20computer%20science%20with%20a%20computer.f21dc4166340b6c8b4d1cb57efd1e22127407f9b28c9ac7afe11344065369e64.png) | ![Imagem gerada pelo DALL-E](../../../../../translated_images/br/DALL·E%202023-06-20%2015.58.42%20-%20%20a%20closeup%20oil%20portrait%20of%20old%20male%20teacher%20of%20mathematics%20in%20front%20of%20blackboard.d331c2dfbdc3f7c46aa65c0809066f5e7ed4b49609cd259852e760df21051e4a.png) +![Imagem gerada pelo DALL-E](../../../../../translated_images/pt-BR/DALL·E%202023-06-20%2015.56.56%20-%20a%20closeup%20watercolor%20portrait%20of%20young%20male%20teacher%20of%20literature%20with%20a%20book.6c235e8271d9ed10ce985d86aeb241a58518958647973af136912116b9518fce.png) | ![Imagem gerada pelo DALL-E](../../../../../translated_images/pt-BR/DALL·E%202023-06-20%2015.57.43%20-%20a%20closeup%20oil%20portrait%20of%20young%20female%20teacher%20of%20computer%20science%20with%20a%20computer.f21dc4166340b6c8b4d1cb57efd1e22127407f9b28c9ac7afe11344065369e64.png) | ![Imagem gerada pelo DALL-E](../../../../../translated_images/pt-BR/DALL·E%202023-06-20%2015.58.42%20-%20%20a%20closeup%20oil%20portrait%20of%20old%20male%20teacher%20of%20mathematics%20in%20front%20of%20blackboard.d331c2dfbdc3f7c46aa65c0809066f5e7ed4b49609cd259852e760df21051e4a.png) ----|----|---- Imagem gerada a partir da descrição *um retrato em aquarela de um jovem professor de literatura com um livro* | Imagem gerada a partir da descrição *um retrato a óleo de uma jovem professora de ciência da computação com um computador* | Imagem gerada a partir da descrição *um retrato a óleo de um velho professor de matemática em frente a um quadro-negro* diff --git a/translations/de/.co-op-translator.json b/translations/de/.co-op-translator.json new file mode 100644 index 00000000..066f0692 --- /dev/null +++ b/translations/de/.co-op-translator.json @@ -0,0 +1,398 @@ +{ + "AGENTS.md": { + "original_hash": "6b11a37115944252ab3ed04e358d830d", + "translation_date": "2025-10-03T09:11:45+00:00", + "source_file": "AGENTS.md", + "language_code": "de" + }, + "README.md": { + "original_hash": "1984fc89dd304a8a33ab5584691a99aa", + "translation_date": "2026-01-30T01:02:51+00:00", + "source_file": "README.md", + "language_code": "de" + }, + "SECURITY.md": { + "original_hash": "a583f49d359c7ebba61433e4dfcd05a9", + "translation_date": "2025-08-24T09:28:59+00:00", + "source_file": "SECURITY.md", + "language_code": "de" + }, + "etc/CODE_OF_CONDUCT.md": { + "original_hash": "c06b12caf3c901eb3156e3dd5b0aea56", + "translation_date": "2025-08-24T09:44:37+00:00", + "source_file": "etc/CODE_OF_CONDUCT.md", + "language_code": "de" + }, + "etc/CONTRIBUTING.md": { + "original_hash": "847a587aa1b83f4d00858183ff3ed18a", + "translation_date": "2025-08-24T09:44:51+00:00", + "source_file": "etc/CONTRIBUTING.md", + "language_code": "de" + }, + "etc/Mindmap.md": { + "original_hash": "f2f88dbd2debd38e26149b27b1fd272d", + "translation_date": "2025-08-24T09:44:59+00:00", + "source_file": "etc/Mindmap.md", + "language_code": "de" + }, + "etc/SUPPORT.md": { + "original_hash": "fdfc08baee91e402938a2b1f94fe0949", + "translation_date": "2025-08-24T09:44:32+00:00", + "source_file": "etc/SUPPORT.md", + "language_code": "de" + }, + "etc/TRANSLATIONS.md": { + "original_hash": "62b3e3ad5182edb905eec649a87eeeb4", + "translation_date": "2025-08-24T09:44:42+00:00", + "source_file": "etc/TRANSLATIONS.md", + "language_code": "de" + }, + "etc/quiz-app/README.md": { + "original_hash": "d699cf8509f74baa5b0b838de5cf0662", + "translation_date": "2025-08-24T09:45:20+00:00", + "source_file": "etc/quiz-app/README.md", + "language_code": "de" + }, + "examples/README.md": { + "original_hash": "0d1babfdcbeb46525f2db3fbaaa54cd7", + "translation_date": "2025-10-03T11:26:30+00:00", + "source_file": "examples/README.md", + "language_code": "de" + }, + "lessons/0-course-setup/for-teachers.md": { + "original_hash": "a094ef9927883de1cfcee51dbd143381", + "translation_date": "2025-08-24T09:44:12+00:00", + "source_file": "lessons/0-course-setup/for-teachers.md", + "language_code": "de" + }, + "lessons/0-course-setup/how-to-run.md": { + "original_hash": "a4717bd9103b9f6cd84d534b83534689", + "translation_date": "2026-01-15T10:52:10+00:00", + "source_file": "lessons/0-course-setup/how-to-run.md", + "language_code": "de" + }, + "lessons/0-course-setup/setup.md": { + "original_hash": "7b4e5b8956915870d0a0ed3cc5890042", + "translation_date": "2025-12-12T19:08:49+00:00", + "source_file": "lessons/0-course-setup/setup.md", + "language_code": "de" + }, + "lessons/1-Intro/README.md": { + "original_hash": "f57e8aa46141fd220b16ffed8f11aec7", + "translation_date": "2025-11-18T20:34:32+00:00", + "source_file": "lessons/1-Intro/README.md", + "language_code": "de" + }, + "lessons/1-Intro/assignment.md": { + "original_hash": "a334df77a82aaaf2a29c77065d3e481e", + "translation_date": "2025-11-18T20:35:38+00:00", + "source_file": "lessons/1-Intro/assignment.md", + "language_code": "de" + }, + "lessons/2-Symbolic/README.md": { + "original_hash": "f9f06b266b8b2bfc6b8792ff2bb1bea4", + "translation_date": "2026-01-15T10:52:48+00:00", + "source_file": "lessons/2-Symbolic/README.md", + "language_code": "de" + }, + "lessons/2-Symbolic/assignment.md": { + "original_hash": "a057a8604f3976c3e309884453f1fad0", + "translation_date": "2025-08-24T09:43:46+00:00", + "source_file": "lessons/2-Symbolic/assignment.md", + "language_code": "de" + }, + "lessons/3-NeuralNetworks/03-Perceptron/README.md": { + "original_hash": "c34cbba802058b6fa267e1a294d4e510", + "translation_date": "2025-09-23T12:21:52+00:00", + "source_file": "lessons/3-NeuralNetworks/03-Perceptron/README.md", + "language_code": "de" + }, + "lessons/3-NeuralNetworks/03-Perceptron/lab/README.md": { + "original_hash": "ba5d1eb353d20d3e7181066b3c424b99", + "translation_date": "2025-08-31T15:37:43+00:00", + "source_file": "lessons/3-NeuralNetworks/03-Perceptron/lab/README.md", + "language_code": "de" + }, + "lessons/3-NeuralNetworks/04-OwnFramework/README.md": { + "original_hash": "789d6c3fb6fc7948a470b33078a5983a", + "translation_date": "2025-09-23T12:21:32+00:00", + "source_file": "lessons/3-NeuralNetworks/04-OwnFramework/README.md", + "language_code": "de" + }, + "lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md": { + "original_hash": "48fdd704d483e19bc3d7464074c9fcbe", + "translation_date": "2025-08-24T09:40:10+00:00", + "source_file": "lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md", + "language_code": "de" + }, + "lessons/3-NeuralNetworks/05-Frameworks/README.md": { + "original_hash": "ddd216f558a255260a9374008002c971", + "translation_date": "2025-09-23T12:22:07+00:00", + "source_file": "lessons/3-NeuralNetworks/05-Frameworks/README.md", + "language_code": "de" + }, + "lessons/3-NeuralNetworks/05-Frameworks/lab/README.md": { + "original_hash": "e452d897efb9a89700f41021834cf6e5", + "translation_date": "2025-08-24T09:40:45+00:00", + "source_file": "lessons/3-NeuralNetworks/05-Frameworks/lab/README.md", + "language_code": "de" + }, + "lessons/3-NeuralNetworks/README.md": { + "original_hash": "f862a99d88088163df12270e2f2ad6c3", + "translation_date": "2025-10-03T12:41:36+00:00", + "source_file": "lessons/3-NeuralNetworks/README.md", + "language_code": "de" + }, + "lessons/4-ComputerVision/06-IntroCV/README.md": { + "original_hash": "feeca98225cb420afc89415f24f63d92", + "translation_date": "2025-09-23T12:17:33+00:00", + "source_file": "lessons/4-ComputerVision/06-IntroCV/README.md", + "language_code": "de" + }, + "lessons/4-ComputerVision/06-IntroCV/lab/README.md": { + "original_hash": "3d53d6409f80970f7281a45dee35328a", + "translation_date": "2025-08-24T09:36:49+00:00", + "source_file": "lessons/4-ComputerVision/06-IntroCV/lab/README.md", + "language_code": "de" + }, + "lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md": { + "original_hash": "53faab85adfcebd8c10bcd71dc2fa557", + "translation_date": "2025-09-23T12:17:02+00:00", + "source_file": "lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md", + "language_code": "de" + }, + "lessons/4-ComputerVision/07-ConvNets/README.md": { + "original_hash": "a560d5b845962cf33dc102266e409568", + "translation_date": "2025-09-23T12:16:35+00:00", + "source_file": "lessons/4-ComputerVision/07-ConvNets/README.md", + "language_code": "de" + }, + "lessons/4-ComputerVision/07-ConvNets/lab/README.md": { + "original_hash": "b70fcf7fcee862990f848c679090943f", + "translation_date": "2025-10-03T14:52:02+00:00", + "source_file": "lessons/4-ComputerVision/07-ConvNets/lab/README.md", + "language_code": "de" + }, + "lessons/4-ComputerVision/08-TransferLearning/README.md": { + "original_hash": "178c0b5ee5395733eb18aec51e71a0a9", + "translation_date": "2025-09-23T12:17:16+00:00", + "source_file": "lessons/4-ComputerVision/08-TransferLearning/README.md", + "language_code": "de" + }, + "lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md": { + "original_hash": "ae074cd940fc2f4dc24fc07b66ccbd99", + "translation_date": "2025-08-24T09:37:15+00:00", + "source_file": "lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md", + "language_code": "de" + }, + "lessons/4-ComputerVision/08-TransferLearning/lab/README.md": { + "original_hash": "7765935c35fcee69b9fe2d0cfd6963e2", + "translation_date": "2025-08-24T09:37:45+00:00", + "source_file": "lessons/4-ComputerVision/08-TransferLearning/lab/README.md", + "language_code": "de" + }, + "lessons/4-ComputerVision/09-Autoencoders/README.md": { + "original_hash": "1b8d9e1b3a6f1daa864b1ff3dfc3076d", + "translation_date": "2025-09-23T12:18:32+00:00", + "source_file": "lessons/4-ComputerVision/09-Autoencoders/README.md", + "language_code": "de" + }, + "lessons/4-ComputerVision/10-GANs/README.md": { + "original_hash": "0ff65b4da07b23697235de2beb2a3c25", + "translation_date": "2025-09-23T12:19:09+00:00", + "source_file": "lessons/4-ComputerVision/10-GANs/README.md", + "language_code": "de" + }, + "lessons/4-ComputerVision/11-ObjectDetection/README.md": { + "original_hash": "d76a7eda28de5210c8b1ba50a6216c69", + "translation_date": "2025-09-23T12:17:58+00:00", + "source_file": "lessons/4-ComputerVision/11-ObjectDetection/README.md", + "language_code": "de" + }, + "lessons/4-ComputerVision/11-ObjectDetection/lab/README.md": { + "original_hash": "ad568d55ae65c856fe929fc2b278510a", + "translation_date": "2025-08-24T09:35:32+00:00", + "source_file": "lessons/4-ComputerVision/11-ObjectDetection/lab/README.md", + "language_code": "de" + }, + "lessons/4-ComputerVision/12-Segmentation/README.md": { + "original_hash": "6568aaae7e0e4afed4b5d74b5b223700", + "translation_date": "2025-09-23T12:18:54+00:00", + "source_file": "lessons/4-ComputerVision/12-Segmentation/README.md", + "language_code": "de" + }, + "lessons/4-ComputerVision/12-Segmentation/lab/README.md": { + "original_hash": "365f0decfe0f47b460bbde8227c5009d", + "translation_date": "2025-08-24T09:34:29+00:00", + "source_file": "lessons/4-ComputerVision/12-Segmentation/lab/README.md", + "language_code": "de" + }, + "lessons/4-ComputerVision/README.md": { + "original_hash": "58a52f000089c1d8906a4daa4ab1169b", + "translation_date": "2025-08-24T09:33:41+00:00", + "source_file": "lessons/4-ComputerVision/README.md", + "language_code": "de" + }, + "lessons/5-NLP/13-TextRep/README.md": { + "original_hash": "dbd3f73e4139f030ecb2e20387d70fee", + "translation_date": "2025-09-23T12:24:58+00:00", + "source_file": "lessons/5-NLP/13-TextRep/README.md", + "language_code": "de" + }, + "lessons/5-NLP/13-TextRep/assignment.md": { + "original_hash": "cdc1f2e631f055f3473b36d18e4760b3", + "translation_date": "2025-08-24T09:31:50+00:00", + "source_file": "lessons/5-NLP/13-TextRep/assignment.md", + "language_code": "de" + }, + "lessons/5-NLP/14-Embeddings/README.md": { + "original_hash": "b708c9b85b833864c73c6281f1e6b96e", + "translation_date": "2025-09-23T12:24:39+00:00", + "source_file": "lessons/5-NLP/14-Embeddings/README.md", + "language_code": "de" + }, + "lessons/5-NLP/14-Embeddings/assignment.md": { + "original_hash": "bc690ecf68b38d311cc9e12f3144a28c", + "translation_date": "2025-08-24T09:30:53+00:00", + "source_file": "lessons/5-NLP/14-Embeddings/assignment.md", + "language_code": "de" + }, + "lessons/5-NLP/15-LanguageModeling/README.md": { + "original_hash": "7ba20f54a5bfcd6521018cdfb17c7c57", + "translation_date": "2025-09-23T12:23:03+00:00", + "source_file": "lessons/5-NLP/15-LanguageModeling/README.md", + "language_code": "de" + }, + "lessons/5-NLP/15-LanguageModeling/lab/README.md": { + "original_hash": "5130f01fdc5ebb83032b23d489027aac", + "translation_date": "2025-08-24T09:32:08+00:00", + "source_file": "lessons/5-NLP/15-LanguageModeling/lab/README.md", + "language_code": "de" + }, + "lessons/5-NLP/16-RNN/README.md": { + "original_hash": "e2273cc150380a5e191903cea858f021", + "translation_date": "2025-09-23T12:24:13+00:00", + "source_file": "lessons/5-NLP/16-RNN/README.md", + "language_code": "de" + }, + "lessons/5-NLP/16-RNN/assignment.md": { + "original_hash": "47f7d3c6a5373543e051e4d1140ce898", + "translation_date": "2025-08-24T09:30:25+00:00", + "source_file": "lessons/5-NLP/16-RNN/assignment.md", + "language_code": "de" + }, + "lessons/5-NLP/17-GenerativeNetworks/README.md": { + "original_hash": "51be6057374d01d70e07dd5ec88ebc0d", + "translation_date": "2025-09-23T12:22:45+00:00", + "source_file": "lessons/5-NLP/17-GenerativeNetworks/README.md", + "language_code": "de" + }, + "lessons/5-NLP/17-GenerativeNetworks/lab/README.md": { + "original_hash": "439e12796197a90e7623d4c9c057b9c2", + "translation_date": "2025-08-24T09:31:22+00:00", + "source_file": "lessons/5-NLP/17-GenerativeNetworks/lab/README.md", + "language_code": "de" + }, + "lessons/5-NLP/18-Transformers/README.md": { + "original_hash": "f335dfcb4a993920504c387973a36957", + "translation_date": "2025-09-23T12:23:12+00:00", + "source_file": "lessons/5-NLP/18-Transformers/README.md", + "language_code": "de" + }, + "lessons/5-NLP/18-Transformers/assignment.md": { + "original_hash": "177f3ea3995d725e6f9f5c66af16edcd", + "translation_date": "2025-08-24T09:32:15+00:00", + "source_file": "lessons/5-NLP/18-Transformers/assignment.md", + "language_code": "de" + }, + "lessons/5-NLP/19-NER/README.md": { + "original_hash": "6522312ff835796ca34136a9462fafb2", + "translation_date": "2025-09-23T12:23:56+00:00", + "source_file": "lessons/5-NLP/19-NER/README.md", + "language_code": "de" + }, + "lessons/5-NLP/19-NER/lab/README.md": { + "original_hash": "032bda5068f543d6c1fcb30c34231461", + "translation_date": "2025-08-24T09:32:40+00:00", + "source_file": "lessons/5-NLP/19-NER/lab/README.md", + "language_code": "de" + }, + "lessons/5-NLP/20-LangModels/README.md": { + "original_hash": "97836d30a6bec736f8e3b4411c572bc2", + "translation_date": "2025-09-23T12:23:42+00:00", + "source_file": "lessons/5-NLP/20-LangModels/README.md", + "language_code": "de" + }, + "lessons/5-NLP/README.md": { + "original_hash": "8ef02a9318257ea140ed3ed74442096d", + "translation_date": "2025-08-24T09:29:31+00:00", + "source_file": "lessons/5-NLP/README.md", + "language_code": "de" + }, + "lessons/6-Other/21-GeneticAlgorithms/README.md": { + "original_hash": "6bbd632dfe6c62e5f66bb51fd78c174a", + "translation_date": "2025-09-23T12:15:15+00:00", + "source_file": "lessons/6-Other/21-GeneticAlgorithms/README.md", + "language_code": "de" + }, + "lessons/6-Other/22-DeepRL/README.md": { + "original_hash": "04395657fc01648f8f70484d0e55ab67", + "translation_date": "2025-09-23T12:16:09+00:00", + "source_file": "lessons/6-Other/22-DeepRL/README.md", + "language_code": "de" + }, + "lessons/6-Other/22-DeepRL/lab/README.md": { + "original_hash": "7bd8dc72040e98e35e7225e34058cd4e", + "translation_date": "2025-08-24T09:39:10+00:00", + "source_file": "lessons/6-Other/22-DeepRL/lab/README.md", + "language_code": "de" + }, + "lessons/6-Other/23-MultiagentSystems/README.md": { + "original_hash": "38a1185ae3d54b180378bbd71ae3ef16", + "translation_date": "2025-09-23T12:15:28+00:00", + "source_file": "lessons/6-Other/23-MultiagentSystems/README.md", + "language_code": "de" + }, + "lessons/6-Other/23-MultiagentSystems/assignment.md": { + "original_hash": "cf654ca60c7f86c8dad28596fb42994b", + "translation_date": "2025-08-24T09:38:40+00:00", + "source_file": "lessons/6-Other/23-MultiagentSystems/assignment.md", + "language_code": "de" + }, + "lessons/7-Ethics/README.md": { + "original_hash": "437c988596e751072e41a5aad3fcc5d9", + "translation_date": "2025-08-24T09:29:15+00:00", + "source_file": "lessons/7-Ethics/README.md", + "language_code": "de" + }, + "lessons/README.md": { + "original_hash": "5fef1a0b22498d7188959e2a2cb08af7", + "translation_date": "2025-08-24T09:29:11+00:00", + "source_file": "lessons/README.md", + "language_code": "de" + }, + "lessons/X-Extras/X1-MultiModal/README.md": { + "original_hash": "9c592c26aca16ca085d268c732284187", + "translation_date": "2025-08-24T09:39:16+00:00", + "source_file": "lessons/X-Extras/X1-MultiModal/README.md", + "language_code": "de" + }, + "lessons/sketchnotes/LICENSE.md": { + "original_hash": "45ab63a2cd8f5faef6c9b150618837a4", + "translation_date": "2025-08-24T09:42:06+00:00", + "source_file": "lessons/sketchnotes/LICENSE.md", + "language_code": "de" + }, + "lessons/sketchnotes/README.md": { + "original_hash": "050b8bddebafba55b129414e6ab096ab", + "translation_date": "2025-08-24T09:41:18+00:00", + "source_file": "lessons/sketchnotes/README.md", + "language_code": "de" + }, + "troubleshoot.md": { + "original_hash": "8d9c5a4a7c7798d699672a22cb7fea86", + "translation_date": "2025-10-03T09:34:33+00:00", + "source_file": "troubleshoot.md", + "language_code": "de" + } +} \ No newline at end of file diff --git a/translations/de/AGENTS.md b/translations/de/AGENTS.md index c2866d63..514e3758 100644 --- a/translations/de/AGENTS.md +++ b/translations/de/AGENTS.md @@ -1,12 +1,3 @@ - # AGENTS.md ## Projektübersicht diff --git a/translations/de/README.md b/translations/de/README.md index 3f2dd232..a32a091e 100644 --- a/translations/de/README.md +++ b/translations/de/README.md @@ -1,12 +1,3 @@ - [![GitHub license](https://img.shields.io/github/license/microsoft/AI-For-Beginners.svg)](https://github.com/microsoft/AI-For-Beginners/blob/main/LICENSE) [![GitHub contributors](https://img.shields.io/github/contributors/microsoft/AI-For-Beginners.svg)](https://GitHub.com/microsoft/AI-For-Beginners/graphs/contributors/) [![GitHub issues](https://img.shields.io/github/issues/microsoft/AI-For-Beginners.svg)](https://GitHub.com/microsoft/AI-For-Beginners/issues/) @@ -21,207 +12,208 @@ CO_OP_TRANSLATOR_METADATA: [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/nTYy5BXMWG)](https://discord.gg/nTYy5BXMWG) -# Künstliche Intelligenz für Anfänger - Ein Lehrplan +# Künstliche Intelligenz für Anfänger – Ein Lehrplan -|![Sketchnote von @girlie_mac https://twitter.com/girlie_mac](../../../../translated_images/de/ai-overview.0857791951d19500.webp)| +|![Sketchnote von @girlie_mac https://twitter.com/girlie_mac](../../translated_images/de/ai-overview.0857791951d19500.webp)| |:---:| -| AI For Beginners - _Sketchnote von [@girlie_mac](https://twitter.com/girlie_mac)_ | +| KI für Anfänger – _Sketchnote von [@girlie_mac](https://twitter.com/girlie_mac)_ | -Erkunde die Welt der **Künstlichen Intelligenz** (KI) mit unserem 12-wöchigen, 24-teiligen Lehrplan! Er enthält praktische Lektionen, Quizze und Labore. Der Lehrplan ist anfängerfreundlich und behandelt Werkzeuge wie TensorFlow und PyTorch sowie ethische Aspekte in der KI +Entdecke die Welt der **Künstlichen Intelligenz** (KI) mit unserem 12-wöchigen Lehrplan mit 24 Lektionen! Er beinhaltet praktische Lektionen, Quizfragen und Labs. Der Lehrplan ist anfängerfreundlich und behandelt Werkzeuge wie TensorFlow und PyTorch sowie Ethik in der KI. ### 🌐 Mehrsprachige Unterstützung -#### Unterstützt über GitHub Action (Automatisiert & Immer Aktuell) +#### Unterstützt durch GitHub Action (Automatisiert & Immer Aktuell) -[Arabisch](../ar/README.md) | [Bengalisch](../bn/README.md) | [Bulgarisch](../bg/README.md) | [Birmanisch (Myanmar)](../my/README.md) | [Chinesisch (Vereinfacht)](../zh/README.md) | [Chinesisch (Traditionell, Hongkong)](../hk/README.md) | [Chinesisch (Traditionell, Macau)](../mo/README.md) | [Chinesisch (Traditionell, Taiwan)](../tw/README.md) | [Kroatisch](../hr/README.md) | [Tschechisch](../cs/README.md) | [Dänisch](../da/README.md) | [Niederländisch](../nl/README.md) | [Estnisch](../et/README.md) | [Finnisch](../fi/README.md) | [Französisch](../fr/README.md) | [Deutsch](./README.md) | [Griechisch](../el/README.md) | [Hebräisch](../he/README.md) | [Hindi](../hi/README.md) | [Ungarisch](../hu/README.md) | [Indonesisch](../id/README.md) | [Italienisch](../it/README.md) | [Japanisch](../ja/README.md) | [Kannada](../kn/README.md) | [Koreanisch](../ko/README.md) | [Litauisch](../lt/README.md) | [Malaiisch](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepalesisch](../ne/README.md) | [Nigerianisches Pidgin](../pcm/README.md) | [Norwegisch](../no/README.md) | [Persisch (Farsi)](../fa/README.md) | [Polnisch](../pl/README.md) | [Portugiesisch (Brasilien)](../br/README.md) | [Portugiesisch (Portugal)](../pt/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Rumänisch](../ro/README.md) | [Russisch](../ru/README.md) | [Serbisch (Kyrillisch)](../sr/README.md) | [Slowakisch](../sk/README.md) | [Slowenisch](../sl/README.md) | [Spanisch](../es/README.md) | [Swahili](../sw/README.md) | [Schwedisch](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thailändisch](../th/README.md) | [Türkisch](../tr/README.md) | [Ukrainisch](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamesisch](../vi/README.md) +[Arabisch](../ar/README.md) | [Bengalisch](../bn/README.md) | [Bulgarisch](../bg/README.md) | [Birmanisch (Myanmar)](../my/README.md) | [Chinesisch (Vereinfacht)](../zh-CN/README.md) | [Chinesisch (Traditionell, Hongkong)](../zh-HK/README.md) | [Chinesisch (Traditionell, Macau)](../zh-MO/README.md) | [Chinesisch (Traditionell, Taiwan)](../zh-TW/README.md) | [Kroatisch](../hr/README.md) | [Tschechisch](../cs/README.md) | [Dänisch](../da/README.md) | [Niederländisch](../nl/README.md) | [Estnisch](../et/README.md) | [Finnisch](../fi/README.md) | [Französisch](../fr/README.md) | [Deutsch](./README.md) | [Griechisch](../el/README.md) | [Hebräisch](../he/README.md) | [Hindi](../hi/README.md) | [Ungarisch](../hu/README.md) | [Indonesisch](../id/README.md) | [Italienisch](../it/README.md) | [Japanisch](../ja/README.md) | [Kannada](../kn/README.md) | [Koreanisch](../ko/README.md) | [Litauisch](../lt/README.md) | [Malaiisch](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepalesisch](../ne/README.md) | [Nigerianisches Pidgin](../pcm/README.md) | [Norwegisch](../no/README.md) | [Persisch (Farsi)](../fa/README.md) | [Polnisch](../pl/README.md) | [Portugiesisch (Brasilien)](../pt-BR/README.md) | [Portugiesisch (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Rumänisch](../ro/README.md) | [Russisch](../ru/README.md) | [Serbisch (Kyrillisch)](../sr/README.md) | [Slowakisch](../sk/README.md) | [Slowenisch](../sl/README.md) | [Spanisch](../es/README.md) | [Suaheli](../sw/README.md) | [Schwedisch](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thailändisch](../th/README.md) | [Türkisch](../tr/README.md) | [Ukrainisch](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamesisch](../vi/README.md) -> **Möchten Sie lokal klonen?** +> **Lieber lokal klonen?** -> Dieses Repository enthält über 50 Sprachübersetzungen, was die Downloadgröße erheblich erhöht. Um ohne Übersetzungen zu klonen, verwenden Sie Sparse Checkout: +> Dieses Repository enthält über 50 Sprachübersetzungen, was die Downloadgröße erheblich erhöht. Zum Klonen ohne Übersetzungen, verwenden Sie Sparse Checkout: > ```bash > git clone --filter=blob:none --sparse https://github.com/microsoft/AI-For-Beginners.git > cd AI-For-Beginners > git sparse-checkout set --no-cone '/*' '!translations' '!translated_images' > ``` -> Das gibt Ihnen alles, was Sie zur Teilnahme am Kurs benötigen, mit einem viel schnelleren Download. +> Dies liefert Ihnen alles, was Sie für den Kurs brauchen, bei deutlich schnellerem Download. -**Falls Sie weitere unterstützte Übersetzungssprachen wünschen, finden Sie diese [hier](https://github.com/Azure/co-op-translator/blob/main/getting_started/supported-languages.md)** +**Wenn Sie wünschen, dass weitere Übersetzungssprachen unterstützt werden, sind diese [hier](https://github.com/Azure/co-op-translator/blob/main/getting_started/supported-languages.md) aufgelistet** -## Werden Sie Teil der Community +## Trete der Community bei [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/nTYy5BXMWG)](https://discord.gg/nTYy5BXMWG) ## Was Sie lernen werden **[Mindmap des Kurses](http://soshnikov.com/courses/ai-for-beginners/mindmap.html)** -In diesem Lehrplan lernen Sie: +In diesem Lehrplan lernst du: -* Verschiedene Ansätze der Künstlichen Intelligenz, einschließlich des „guten alten“ symbolischen Ansatzes mit **Wissensrepräsentation** und Schlussfolgerungen ([GOFAI](https://en.wikipedia.org/wiki/Symbolic_artificial_intelligence)). -* **Neuronale Netzwerke** und **Deep Learning**, die im Kern der modernen KI stehen. Wir werden die Konzepte hinter diesen wichtigen Themen anhand von Code in zwei der beliebtesten Frameworks veranschaulichen - [TensorFlow](http://Tensorflow.org) und [PyTorch](http://pytorch.org). -* **Neuronale Architekturen** für die Arbeit mit Bildern und Text. Wir behandeln aktuelle Modelle, könnten aber leicht hinter dem neuesten Stand sein. -* Weniger populäre KI-Ansätze, wie **Genetische Algorithmen** und **Multiagentensysteme**. +* Verschiedene Ansätze der künstlichen Intelligenz, einschließlich des "guten alten" symbolischen Ansatzes mit **Wissensrepräsentation** und Schlussfolgerungen ([GOFAI](https://en.wikipedia.org/wiki/Symbolic_artificial_intelligence)). +* **Neuronale Netze** und **Deep Learning**, die im Kern der modernen KI stehen. Wir veranschaulichen die Konzepte dieser wichtigen Themen anhand von Code in zwei der beliebtesten Frameworks – [TensorFlow](http://Tensorflow.org) und [PyTorch](http://pytorch.org). +* **Neuronale Architekturen** zur Arbeit mit Bildern und Text. Wir behandeln neuere Modelle, die jedoch etwas weniger den neuesten Stand der Technik widerspiegeln könnten. +* Weniger populäre KI-Ansätze, wie **Genetische Algorithmen** und **Multi-Agenten-Systeme**. Was wir in diesem Lehrplan nicht behandeln: -> [Finden Sie alle zusätzlichen Ressourcen zu diesem Kurs in unserer Microsoft Learn Sammlung](https://learn.microsoft.com/en-us/collections/7w28iy2xrqzdj0?WT.mc_id=academic-77998-bethanycheum) +> [Finde alle zusätzlichen Ressourcen für diesen Kurs in unserer Microsoft Learn Kollektion](https://learn.microsoft.com/en-us/collections/7w28iy2xrqzdj0?WT.mc_id=academic-77998-bethanycheum) -* Geschäftsfälle für den Einsatz von **KI im Geschäftsleben**. Ziehen Sie in Erwägung, den Lernpfad [Einführung in KI für Geschäftsanwender](https://docs.microsoft.com/learn/paths/introduction-ai-for-business-users/?WT.mc_id=academic-77998-bethanycheum) bei Microsoft Learn oder die [AI Business School](https://www.microsoft.com/ai/ai-business-school/?WT.mc_id=academic-77998-bethanycheum), entwickelt in Zusammenarbeit mit [INSEAD](https://www.insead.edu/), zu besuchen. -* **Klassisches Machine Learning**, das in unserem [Lehrplan für Machine Learning für Anfänger](http://github.com/Microsoft/ML-for-Beginners) gut beschrieben ist. -* Praktische KI-Anwendungen, die mit **[Cognitive Services](https://azure.microsoft.com/services/cognitive-services/?WT.mc_id=academic-77998-bethanycheum)** erstellt wurden. Dafür empfehlen wir Ihnen, mit den Modulen bei Microsoft Learn für [Vision](https://docs.microsoft.com/learn/paths/create-computer-vision-solutions-azure-cognitive-services/?WT.mc_id=academic-77998-bethanycheum), [natürliche Sprachverarbeitung](https://docs.microsoft.com/learn/paths/explore-natural-language-processing/?WT.mc_id=academic-77998-bethanycheum), **[Generative AI mit Azure OpenAI Service](https://learn.microsoft.com/en-us/training/paths/develop-ai-solutions-azure-openai/?WT.mc_id=academic-77998-bethanycheum)** und andere zu beginnen. -* Spezifische ML-**Cloud-Frameworks**, wie [Azure Machine Learning](https://azure.microsoft.com/services/machine-learning/?WT.mc_id=academic-77998-bethanycheum), [Microsoft Fabric](https://learn.microsoft.com/en-us/training/paths/get-started-fabric/?WT.mc_id=academic-77998-bethanycheum) oder [Azure Databricks](https://docs.microsoft.com/learn/paths/data-engineer-azure-databricks?WT.mc_id=academic-77998-bethanycheum). Ziehen Sie die Lernpfade [Maschinelles Lernen mit Azure Machine Learning entwickeln und betreiben](https://docs.microsoft.com/learn/paths/build-ai-solutions-with-azure-ml-service/?WT.mc_id=academic-77998-bethanycheum) und [Maschinelles Lernen mit Azure Databricks entwickeln und betreiben](https://docs.microsoft.com/learn/paths/build-operate-machine-learning-solutions-azure-databricks/?WT.mc_id=academic-77998-bethanycheum) in Betracht. -* **Konversationelle KI** und **Chatbots**. Es gibt einen separaten Lernpfad [Konversationelle KI-Lösungen erstellen](https://docs.microsoft.com/learn/paths/create-conversational-ai-solutions/?WT.mc_id=academic-77998-bethanycheum), und Sie können auch [diesen Blogbeitrag](https://soshnikov.com/azure/hello-bot-conversational-ai-on-microsoft-platform/) für weitere Details lesen. -* **Tiefgehende Mathematik** hinter Deep Learning. Dafür empfehlen wir [Deep Learning](https://www.amazon.com/Deep-Learning-Adaptive-Computation-Machine/dp/0262035618) von Ian Goodfellow, Yoshua Bengio und Aaron Courville, das auch online unter [https://www.deeplearningbook.org/](https://www.deeplearningbook.org/) verfügbar ist. +* Geschäftsbeispiele für die Nutzung von **KI im Business**. Ziehe in Betracht, den Lernpfad [Einführung in KI für Business-Anwender](https://docs.microsoft.com/learn/paths/introduction-ai-for-business-users/?WT.mc_id=academic-77998-bethanycheum) auf Microsoft Learn oder die [AI Business School](https://www.microsoft.com/ai/ai-business-school/?WT.mc_id=academic-77998-bethanycheum) zu nutzen, entwickelt in Zusammenarbeit mit [INSEAD](https://www.insead.edu/). +* **Klassisches Machine Learning**, das gut in unserem [Machine Learning für Anfänger Lehrplan](http://github.com/Microsoft/ML-for-Beginners) beschrieben wird. +* Praktische KI-Anwendungen, die mit **[Cognitive Services](https://azure.microsoft.com/services/cognitive-services/?WT.mc_id=academic-77998-bethanycheum)** gebaut werden. Hierfür empfehlen wir Module auf Microsoft Learn zu [Vision](https://docs.microsoft.com/learn/paths/create-computer-vision-solutions-azure-cognitive-services/?WT.mc_id=academic-77998-bethanycheum), [natürlicher Sprachverarbeitung](https://docs.microsoft.com/learn/paths/explore-natural-language-processing/?WT.mc_id=academic-77998-bethanycheum), **[Generative KI mit Azure OpenAI Service](https://learn.microsoft.com/en-us/training/paths/develop-ai-solutions-azure-openai/?WT.mc_id=academic-77998-bethanycheum)** und weiteren. +* Spezifische ML-**Cloud-Frameworks** wie [Azure Machine Learning](https://azure.microsoft.com/services/machine-learning/?WT.mc_id=academic-77998-bethanycheum), [Microsoft Fabric](https://learn.microsoft.com/en-us/training/paths/get-started-fabric/?WT.mc_id=academic-77998-bethanycheum) oder [Azure Databricks](https://docs.microsoft.com/learn/paths/data-engineer-azure-databricks?WT.mc_id=academic-77998-bethanycheum). Ziehe Lernpfade wie [Maschinelles Lernen mit Azure Machine Learning erstellen und betreiben](https://docs.microsoft.com/learn/paths/build-ai-solutions-with-azure-ml-service/?WT.mc_id=academic-77998-bethanycheum) und [Maschinelle Lern-Lösungen mit Azure Databricks aufbauen und betreiben](https://docs.microsoft.com/learn/paths/build-operate-machine-learning-solutions-azure-databricks/?WT.mc_id=academic-77998-bethanycheum) in Betracht. +* **Konversations-KI** und **Chatbots**. Es gibt einen separaten Lernpfad [Konversations-KI-Lösungen erstellen](https://docs.microsoft.com/learn/paths/create-conversational-ai-solutions/?WT.mc_id=academic-77998-bethanycheum), und du kannst dich auch in [diesem Blogbeitrag](https://soshnikov.com/azure/hello-bot-conversational-ai-on-microsoft-platform/) genauer informieren. +* **Tiefe Mathematik** hinter Deep Learning. Hierfür empfehlen wir [Deep Learning](https://www.amazon.com/Deep-Learning-Adaptive-Computation-Machine/dp/0262035618) von Ian Goodfellow, Yoshua Bengio und Aaron Courville, das auch online unter [https://www.deeplearningbook.org/](https://www.deeplearningbook.org/) verfügbar ist. -Für eine sanfte Einführung in _KI in der Cloud_-Themen können Sie den Lernpfad [Erste Schritte mit Künstlicher Intelligenz auf Azure](https://docs.microsoft.com/learn/paths/get-started-with-artificial-intelligence-on-azure/?WT.mc_id=academic-77998-bethanycheum) in Betracht ziehen. +Für eine sanfte Einführung in _KI in der Cloud_ kannst du den Lernpfad [Mit Künstlicher Intelligenz auf Azure starten](https://docs.microsoft.com/learn/paths/get-started-with-artificial-intelligence-on-azure/?WT.mc_id=academic-77998-bethanycheum) in Betracht ziehen. # Inhalt -| | Link zur Lektion | PyTorch/Keras/TensorFlow | Labor | -| :-: | :------------------------------------------------------------------------------------------------------------------------------------------: | :---------------------------------------------------------------------------------------------: | ------------------------------------------------------------------------------ | -| 0 | [Kurs-Einrichtung](./lessons/0-course-setup/setup.md) | [Einrichten Ihrer Entwicklungsumgebung](./lessons/0-course-setup/how-to-run.md) | | +| | Link zur Lektion | PyTorch/Keras/TensorFlow | Labor | +| :-: | :-----------------------------------------------------------------------------------------------------------------------------------------: | :---------------------------------------------------------------------------------------------: | ------------------------------------------------------------------------------ | +| 0 | [Kurs Setup](./lessons/0-course-setup/setup.md) | [Richte deine Entwicklungsumgebung ein](./lessons/0-course-setup/how-to-run.md) | | | I | [**Einführung in KI**](./lessons/1-Intro/README.md) | | | | 01 | [Einführung und Geschichte der KI](./lessons/1-Intro/README.md) | - | - | | II | **Symbolische KI** | -| 02 | [Wissensrepräsentation und Expertensysteme](./lessons/2-Symbolic/README.md) | [Expertensysteme](./lessons/2-Symbolic/Animals.ipynb) / [Ontologie](./lessons/2-Symbolic/FamilyOntology.ipynb) /[Konzeptgraph](./lessons/2-Symbolic/MSConceptGraph.ipynb) | | -| III | [**Einführung in Neuronale Netzwerke**](./lessons/3-NeuralNetworks/README.md) ||| -| 03 | [Perzeptron](./lessons/3-NeuralNetworks/03-Perceptron/README.md) | [Notebook](./lessons/3-NeuralNetworks/03-Perceptron/Perceptron.ipynb) | [Lab](./lessons/3-NeuralNetworks/03-Perceptron/lab/README.md) | -| 04 | [Mehrschichtiger Perzeptron und Erstellung unseres eigenen Frameworks](./lessons/3-NeuralNetworks/04-OwnFramework/README.md) | [Notebook](./lessons/3-NeuralNetworks/04-OwnFramework/OwnFramework.ipynb) | [Lab](./lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md) | -| 05 | [Einführung in Frameworks (PyTorch/TensorFlow) und Overfitting](./lessons/3-NeuralNetworks/05-Frameworks/README.md) | [PyTorch](./lessons/3-NeuralNetworks/05-Frameworks/IntroPyTorch.ipynb) / [Keras](./lessons/3-NeuralNetworks/05-Frameworks/IntroKeras.ipynb) / [TensorFlow](./lessons/3-NeuralNetworks/05-Frameworks/IntroKerasTF.ipynb) | [Lab](./lessons/3-NeuralNetworks/05-Frameworks/lab/README.md) | -| IV | [**Computer Vision**](./lessons/4-ComputerVision/README.md) | [PyTorch](https://docs.microsoft.com/learn/modules/intro-computer-vision-pytorch/?WT.mc_id=academic-77998-cacaste) / [TensorFlow](https://docs.microsoft.com/learn/modules/intro-computer-vision-TensorFlow/?WT.mc_id=academic-77998-cacaste)| [Erkunden Sie Computer Vision auf Microsoft Azure](https://learn.microsoft.com/en-us/collections/7w28iy2xrqzdj0?WT.mc_id=academic-77998-bethanycheum) | -| 06 | [Einführung in Computer Vision. OpenCV](./lessons/4-ComputerVision/06-IntroCV/README.md) | [Notebook](./lessons/4-ComputerVision/06-IntroCV/OpenCV.ipynb) | [Lab](./lessons/4-ComputerVision/06-IntroCV/lab/README.md) | -| 07 | [Convolutional Neural Networks](./lessons/4-ComputerVision/07-ConvNets/README.md) & [CNN-Architekturen](./lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md) | [PyTorch](./lessons/4-ComputerVision/07-ConvNets/ConvNetsPyTorch.ipynb) /[TensorFlow](./lessons/4-ComputerVision/07-ConvNets/ConvNetsTF.ipynb) | [Lab](./lessons/4-ComputerVision/07-ConvNets/lab/README.md) | -| 08 | [Vortrainierte Netzwerke und Transferlernen](./lessons/4-ComputerVision/08-TransferLearning/README.md) und [Training Tricks](./lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md) | [PyTorch](./lessons/4-ComputerVision/08-TransferLearning/TransferLearningPyTorch.ipynb) / [TensorFlow](./lessons/3-NeuralNetworks/05-Frameworks/IntroKerasTF.ipynb) | [Lab](./lessons/4-ComputerVision/08-TransferLearning/lab/README.md) | +| 02 | [Wissensrepräsentation und Expertensysteme](./lessons/2-Symbolic/README.md) | [Expertensysteme](./lessons/2-Symbolic/Animals.ipynb) / [Ontologie](./lessons/2-Symbolic/FamilyOntology.ipynb) /[Begriffsgrafik](./lessons/2-Symbolic/MSConceptGraph.ipynb) | | +| III | [**Einführung in neuronale Netze**](./lessons/3-NeuralNetworks/README.md) ||| +| 03 | [Perzeptron](./lessons/3-NeuralNetworks/03-Perceptron/README.md) | [Notebook](./lessons/3-NeuralNetworks/03-Perceptron/Perceptron.ipynb) | [Labor](./lessons/3-NeuralNetworks/03-Perceptron/lab/README.md) | +| 04 | [Mehrschichtiges Perzeptron und Erstellung unseres eigenen Frameworks](./lessons/3-NeuralNetworks/04-OwnFramework/README.md) | [Notebook](./lessons/3-NeuralNetworks/04-OwnFramework/OwnFramework.ipynb) | [Labor](./lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md) | +| 05 | [Einführung in Frameworks (PyTorch/TensorFlow) und Overfitting](./lessons/3-NeuralNetworks/05-Frameworks/README.md) | [PyTorch](./lessons/3-NeuralNetworks/05-Frameworks/IntroPyTorch.ipynb) / [Keras](./lessons/3-NeuralNetworks/05-Frameworks/IntroKeras.ipynb) / [TensorFlow](./lessons/3-NeuralNetworks/05-Frameworks/IntroKerasTF.ipynb) | [Labor](./lessons/3-NeuralNetworks/05-Frameworks/lab/README.md) | +| IV | [**Computer Vision**](./lessons/4-ComputerVision/README.md) | [PyTorch](https://docs.microsoft.com/learn/modules/intro-computer-vision-pytorch/?WT.mc_id=academic-77998-cacaste) / [TensorFlow](https://docs.microsoft.com/learn/modules/intro-computer-vision-TensorFlow/?WT.mc_id=academic-77998-cacaste)| [Untersuche Computer Vision auf Microsoft Azure](https://learn.microsoft.com/en-us/collections/7w28iy2xrqzdj0?WT.mc_id=academic-77998-bethanycheum) | +| 06 | [Einführung in Computer Vision. OpenCV](./lessons/4-ComputerVision/06-IntroCV/README.md) | [Notebook](./lessons/4-ComputerVision/06-IntroCV/OpenCV.ipynb) | [Labor](./lessons/4-ComputerVision/06-IntroCV/lab/README.md) | +| 07 | [Convolutional Neural Networks](./lessons/4-ComputerVision/07-ConvNets/README.md) & [CNN-Architekturen](./lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md) | [PyTorch](./lessons/4-ComputerVision/07-ConvNets/ConvNetsPyTorch.ipynb) /[TensorFlow](./lessons/4-ComputerVision/07-ConvNets/ConvNetsTF.ipynb) | [Labor](./lessons/4-ComputerVision/07-ConvNets/lab/README.md) | +| 08 | [Vortrainierte Netzwerke und Transferlernen](./lessons/4-ComputerVision/08-TransferLearning/README.md) und [Trainingstricks](./lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md) | [PyTorch](./lessons/4-ComputerVision/08-TransferLearning/TransferLearningPyTorch.ipynb) / [TensorFlow](./lessons/3-NeuralNetworks/05-Frameworks/IntroKerasTF.ipynb) | [Labor](./lessons/4-ComputerVision/08-TransferLearning/lab/README.md) | | 09 | [Autoencoder und VAEs](./lessons/4-ComputerVision/09-Autoencoders/README.md) | [PyTorch](./lessons/4-ComputerVision/09-Autoencoders/AutoEncodersPyTorch.ipynb) / [TensorFlow](./lessons/4-ComputerVision/09-Autoencoders/AutoencodersTF.ipynb) | | -| 10 | [Generative Adversarial Networks & Künstlerischer Stiltransfer](./lessons/4-ComputerVision/10-GANs/README.md) | [PyTorch](./lessons/4-ComputerVision/10-GANs/GANPyTorch.ipynb) / [TensorFlow](./lessons/4-ComputerVision/10-GANs/GANTF.ipynb) | | -| 11 | [Objekterkennung](./lessons/4-ComputerVision/11-ObjectDetection/README.md) | [TensorFlow](./lessons/4-ComputerVision/11-ObjectDetection/ObjectDetection.ipynb) | [Lab](./lessons/4-ComputerVision/11-ObjectDetection/lab/README.md) | +| 10 | [Generative adversariale Netzwerke & künstlerischer Stiltransfer](./lessons/4-ComputerVision/10-GANs/README.md) | [PyTorch](./lessons/4-ComputerVision/10-GANs/GANPyTorch.ipynb) / [TensorFlow](./lessons/4-ComputerVision/10-GANs/GANTF.ipynb) | | +| 11 | [Objekterkennung](./lessons/4-ComputerVision/11-ObjectDetection/README.md) | [TensorFlow](./lessons/4-ComputerVision/11-ObjectDetection/ObjectDetection.ipynb) | [Labor](./lessons/4-ComputerVision/11-ObjectDetection/lab/README.md) | | 12 | [Semantische Segmentierung. U-Net](./lessons/4-ComputerVision/12-Segmentation/README.md) | [PyTorch](./lessons/4-ComputerVision/12-Segmentation/SemanticSegmentationPytorch.ipynb) / [TensorFlow](./lessons/4-ComputerVision/12-Segmentation/SemanticSegmentationTF.ipynb) | | -| V | [**Natural Language Processing**](./lessons/5-NLP/README.md) | [PyTorch](https://docs.microsoft.com/learn/modules/intro-natural-language-processing-pytorch/?WT.mc_id=academic-77998-cacaste) /[TensorFlow](https://docs.microsoft.com/learn/modules/intro-natural-language-processing-TensorFlow/?WT.mc_id=academic-77998-cacaste) | [Erkunden Sie Natural Language Processing auf Microsoft Azure](https://learn.microsoft.com/en-us/collections/7w28iy2xrqzdj0?WT.mc_id=academic-77998-bethanycheum)| -| 13 | [Textrepräsentation. Bow/TF-IDF](./lessons/5-NLP/13-TextRep/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/13-TextRep/TextRepresentationPyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/13-TextRep/TextRepresentationTF.ipynb) | | -| 14 | [Semantische Wort-Einbettungen. Word2Vec und GloVe](./lessons/5-NLP/14-Embeddings/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/14-Embeddings/EmbeddingsPyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/14-Embeddings/EmbeddingsTF.ipynb) | | -| 15 | [Sprachmodellierung. Training eigener Einbettungen](./lessons/5-NLP/15-LanguageModeling/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/15-LanguageModeling/CBoW-PyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/15-LanguageModeling/CBoW-TF.ipynb) | [Lab](./lessons/5-NLP/15-LanguageModeling/lab/README.md) | -| 16 | [Rekurrente Neuronale Netze](./lessons/5-NLP/16-RNN/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/16-RNN/RNNPyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/16-RNN/RNNTF.ipynb) | | -| 17 | [Generative Rekurrente Netze](./lessons/5-NLP/17-GenerativeNetworks/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/17-GenerativeNetworks/GenerativePyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/17-GenerativeNetworks/GenerativeTF.ipynb) | [Lab](./lessons/5-NLP/17-GenerativeNetworks/lab/README.md) | +| V | [**Natürliche Sprachverarbeitung**](./lessons/5-NLP/README.md) | [PyTorch](https://docs.microsoft.com/learn/modules/intro-natural-language-processing-pytorch/?WT.mc_id=academic-77998-cacaste) /[TensorFlow](https://docs.microsoft.com/learn/modules/intro-natural-language-processing-TensorFlow/?WT.mc_id=academic-77998-cacaste) | [Untersuche natürliche Sprachverarbeitung auf Microsoft Azure](https://learn.microsoft.com/en-us/collections/7w28iy2xrqzdj0?WT.mc_id=academic-77998-bethanycheum)| +| 13 | [Textrepräsentation. BoW/TF-IDF](./lessons/5-NLP/13-TextRep/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/13-TextRep/TextRepresentationPyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/13-TextRep/TextRepresentationTF.ipynb) | | +| 14 | [Semantische Wort-Embeddings. Word2Vec und GloVe](./lessons/5-NLP/14-Embeddings/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/14-Embeddings/EmbeddingsPyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/14-Embeddings/EmbeddingsTF.ipynb) | | +| 15 | [Sprachmodellierung. Trainiere eigene Embeddings](./lessons/5-NLP/15-LanguageModeling/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/15-LanguageModeling/CBoW-PyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/15-LanguageModeling/CBoW-TF.ipynb) | [Labor](./lessons/5-NLP/15-LanguageModeling/lab/README.md) | +| 16 | [Rekurrente neuronale Netze](./lessons/5-NLP/16-RNN/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/16-RNN/RNNPyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/16-RNN/RNNTF.ipynb) | | +| 17 | [Generative rekurrente Netze](./lessons/5-NLP/17-GenerativeNetworks/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/17-GenerativeNetworks/GenerativePyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/17-GenerativeNetworks/GenerativeTF.ipynb) | [Labor](./lessons/5-NLP/17-GenerativeNetworks/lab/README.md) | | 18 | [Transformers. BERT.](./lessons/5-NLP/18-Transformers/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/18-Transformers/TransformersPyTorch.ipynb) /[TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/18-Transformers/TransformersTF.ipynb) | | -| 19 | [Erkennung benannter Entitäten](./lessons/5-NLP/19-NER/README.md) | [TensorFlow](https://microsoft.github.io/AI-For-Beginners/lessons/5-NLP/19-NER/NER-TF.ipynb) | [Lab](./lessons/5-NLP/19-NER/lab/README.md) | +| 19 | [Erkennung benannter Entitäten](./lessons/5-NLP/19-NER/README.md) | [TensorFlow](https://microsoft.github.io/AI-For-Beginners/lessons/5-NLP/19-NER/NER-TF.ipynb) | [Labor](./lessons/5-NLP/19-NER/lab/README.md) | | 20 | [Große Sprachmodelle, Prompt-Programmierung und Few-Shot-Aufgaben](./lessons/5-NLP/20-LangModels/README.md) | [PyTorch](https://microsoft.github.io/AI-For-Beginners/lessons/5-NLP/20-LangModels/GPT-PyTorch.ipynb) | | | VI | **Andere KI-Techniken** || | | 21 | [Genetische Algorithmen](./lessons/6-Other/21-GeneticAlgorithms/README.md) | [Notebook](./lessons/6-Other/21-GeneticAlgorithms/Genetic.ipynb) | | -| 22 | [Tiefes Reinforcement Learning](./lessons/6-Other/22-DeepRL/README.md) | [PyTorch](./lessons/6-Other/22-DeepRL/CartPole-RL-PyTorch.ipynb) /[TensorFlow](./lessons/6-Other/22-DeepRL/CartPole-RL-TF.ipynb) | [Lab](./lessons/6-Other/22-DeepRL/lab/README.md) | +| 22 | [Deep Reinforcement Learning](./lessons/6-Other/22-DeepRL/README.md) | [PyTorch](./lessons/6-Other/22-DeepRL/CartPole-RL-PyTorch.ipynb) /[TensorFlow](./lessons/6-Other/22-DeepRL/CartPole-RL-TF.ipynb) | [Labor](./lessons/6-Other/22-DeepRL/lab/README.md) | | 23 | [Multi-Agenten-Systeme](./lessons/6-Other/23-MultiagentSystems/README.md) | | | | VII | **KI-Ethik** | | | | 24 | [KI-Ethik und verantwortungsvolle KI](./lessons/7-Ethics/README.md) | [Microsoft Learn: Prinzipien verantwortungsvoller KI](https://docs.microsoft.com/learn/paths/responsible-ai-business-principles/?WT.mc_id=academic-77998-cacaste) | | | IX | **Extras** | | | -| 25 | [Multimodale Netzwerke, CLIP und VQGAN](./lessons/X-Extras/X1-MultiModal/README.md) | [Notebook](./lessons/X-Extras/X1-MultiModal/Clip.ipynb) | | +| 25 | [Multi-Modal-Netze, CLIP und VQGAN](./lessons/X-Extras/X1-MultiModal/README.md) | [Notebook](./lessons/X-Extras/X1-MultiModal/Clip.ipynb) | | ## Jede Lektion enthält -* Vorbereitendes Lesematerial -* Ausführbare Jupyter-Notebooks, die häufig framework-spezifisch sind (**PyTorch** oder **TensorFlow**). Das ausführbare Notebook enthält auch viel theoretisches Material, daher ist zum Verständnis des Themas mindestens eine Version des Notebooks durchzugehen (entweder PyTorch oder TensorFlow). -* **Labs** zu einigen Themen, die dir die Möglichkeit geben, das Gelernte an einem konkreten Problem anzuwenden. +* Vorlesematerial +* Ausführbare Jupyter Notebooks, die oft auf das jeweilige Framework (**PyTorch** oder **TensorFlow**) spezialisiert sind. Das ausführbare Notebook enthält auch viele theoretische Inhalte, daher solltest du mindestens eine Version des Notebooks (entweder PyTorch oder TensorFlow) durcharbeiten, um das Thema zu verstehen. +* **Labore** zu einzelnen Themen, die dir die Möglichkeit geben, das Gelernte an einem konkreten Problem anzuwenden. * Einige Abschnitte enthalten Links zu [**MS Learn**](https://learn.microsoft.com/en-us/collections/7w28iy2xrqzdj0?WT.mc_id=academic-77998-bethanycheum) Modulen, die verwandte Themen abdecken. ## Erste Schritte ### 🎯 Neu in KI? Hier starten! -Wenn du völlig neu in KI bist und schnelle, praxisorientierte Beispiele suchst, sieh dir unsere [**Anfängergerechten Beispiele**](./examples/README.md) an! Diese umfassen: +Wenn du komplett neu in KI bist und schnelle, praxisnahe Beispiele suchst, schau dir unsere [**Anfängergerechten Beispiele**](./examples/README.md) an! Diese enthalten: -- 🌟 **Hello AI World** - Dein erstes KI-Programm (Mustererkennung) -- 🧠 **Einfaches neuronales Netzwerk** - Erstellen eines neuronalen Netzwerks von Grund auf -- 🖼️ **Bildklassifikator** - Bilder klassifizieren mit detaillierten Kommentaren -- 💬 **Textsentiment** - Analysiere positiven/negativen Text +- 🌟 **Hallo KI Welt** - Dein erstes KI-Programm (Mustererkennung) +- 🧠 **Einfaches neuronales Netz** - Baue ein neuronales Netz von Grund auf auf -Diese Beispiele sollen dir helfen, KI-Konzepte zu verstehen, bevor du in den vollständigen Lehrplan eintauchst. +- 🖼️ **Bildklassifikator** - Klassifizieren Sie Bilder mit ausführlichen Kommentaren +- 💬 **Text-Stimmung** - Analysieren Sie positive/negative Texte -### 📚 Einrichten des vollständigen Lehrplans +Diese Beispiele sollen Ihnen helfen, KI-Konzepte zu verstehen, bevor Sie in den vollständigen Lehrplan eintauchen. -- Wir haben eine [Setup-Lektion](./lessons/0-course-setup/setup.md) erstellt, um dir bei der Einrichtung deiner Entwicklungsumgebung zu helfen. - Für Lehrkräfte haben wir ebenfalls eine [Lehrplan-Einrichtungslektion](./lessons/0-course-setup/for-teachers.md) erstellt! -- Wie man den Code in VSCode oder einem Codespace [ausführt](./lessons/0-course-setup/how-to-run.md) +### 📚 Vollständige Lehrplaneinrichtung -Folge diesen Schritten: +- Wir haben eine [Setup-Lektion](./lessons/0-course-setup/setup.md) erstellt, um Ihnen beim Einrichten Ihrer Entwicklungsumgebung zu helfen. - Für Lehrkräfte haben wir ebenfalls eine [Lehrplaneinrichtungslektion](./lessons/0-course-setup/for-teachers.md) erstellt! +- Wie man [den Code in VSCode oder einem Codespace ausführt](./lessons/0-course-setup/how-to-run.md) -Forke das Repository: Klicke auf die Schaltfläche „Fork“ oben rechts auf dieser Seite. +Folgen Sie diesen Schritten: -Klonen des Repositorys: `git clone https://github.com/microsoft/AI-For-Beginners.git` +Repository forkieren: Klicken Sie auf die Schaltfläche „Fork“ oben rechts auf dieser Seite. -Vergiss nicht, dieses Repo mit einem Stern (🌟) zu markieren, damit du es später leichter findest. +Repository klonen: `git clone https://github.com/microsoft/AI-For-Beginners.git` -## Andere Lernende treffen +Vergessen Sie nicht, dieses Repo mit einem Stern (🌟) zu versehen, damit Sie es später leichter finden. -Tritt unserem [offiziellen AI Discord-Server](https://aka.ms/genai-discord?WT.mc_id=academic-105485-bethanycheum) bei, um andere Lernende, die diesen Kurs machen, zu treffen und dich zu vernetzen und Unterstützung zu erhalten. +## Treffen Sie andere Lernende -Wenn du Produkt-Feedback oder Fragen beim Entwickeln hast, besuche unser [Azure AI Foundry Entwicklerforum](https://aka.ms/foundry/forum) +Treten Sie unserem [offiziellen AI Discord-Server](https://aka.ms/genai-discord?WT.mc_id=academic-105485-bethanycheum) bei, um andere Lernende, die diesen Kurs belegen, kennenzulernen und sich zu vernetzen sowie Unterstützung zu erhalten. -## Quizze +Wenn Sie Produktfeedback oder Fragen während des Buildens haben, besuchen Sie unser [Azure AI Foundry Developer Forum](https://aka.ms/foundry/forum). -> **Eine Anmerkung zu den Quizzen**: Alle Quizze befinden sich im Quiz-App-Ordner in etc\quiz-app oder [online hier](https://ff-quizzes.netlify.app/) Sie sind aus den Lektionen verlinkt. Die Quiz-App kann lokal ausgeführt oder auf Azure bereitgestellt werden; folge den Anweisungen im `quiz-app` Ordner. Sie werden schrittweise lokalisiert. +## Quizze -## Hilfe gesucht +> **Eine Anmerkung zu Quizzen**: Alle Quizze befinden sich im Ordner Quiz-app im Verzeichnis etc\quiz-app oder [Online Hier](https://ff-quizzes.netlify.app/). Sie sind in den Lektionen verlinkt. Die Quiz-App kann lokal ausgeführt oder auf Azure bereitgestellt werden; befolgen Sie die Anweisungen im Ordner `quiz-app`. Sie werden nach und nach lokalisiert. -Hast du Vorschläge oder Rechtschreib- oder Codefehler gefunden? Erstelle ein Issue oder einen Pull Request. +## Gesucht: Unterstützung + +Haben Sie Vorschläge oder Rechtschreib- bzw. Codefehler gefunden? Eröffnen Sie ein Issue oder erstellen Sie einen Pull Request. ## Besonderer Dank -* **✍️ Hauptautor:** [Dmitry Soshnikov](http://soshnikov.com), PhD -* **🔥 Herausgeber:** [Jen Looper](https://twitter.com/jenlooper), PhD -* **🎨 Sketchnote-Illustratorin:** [Tomomi Imura](https://twitter.com/girlie_mac) -* **✅ Quiz-Erstellerin:** [Lateefah Bello](https://github.com/CinnamonXI), [MLSA](https://studentambassadors.microsoft.com/) +* **✍️ Hauptautor:** [Dmitry Soshnikov](http://soshnikov.com), PhD +* **🔥 Herausgeber:** [Jen Looper](https://twitter.com/jenlooper), PhD +* **🎨 Sketchnote-Illustratorin:** [Tomomi Imura](https://twitter.com/girlie_mac) +* **✅ Quiz-Ersteller:** [Lateefah Bello](https://github.com/CinnamonXI), [MLSA](https://studentambassadors.microsoft.com/) * **🙏 Hauptbeitragende:** [Evgenii Pishchik](https://github.com/Pe4enIks) -## Andere Lehrpläne +## Weitere Lehrpläne -Unser Team entwickelt weitere Lehrpläne! Schau mal rein: +Unser Team erstellt weitere Lehrpläne! Schauen Sie sich an: -### LangChain -[![LangChain4j für Anfänger](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +### LangChain +[![LangChain4j für Anfänger](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) [![LangChain.js für Anfänger](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Agents -[![AZD für Anfänger](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI für Anfänger](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP für Anfänger](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +### Azure / Edge / MCP / Agents +[![AZD für Anfänger](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI für Anfänger](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP für Anfänger](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) [![AI Agents für Anfänger](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Generative KI-Serie -[![Generative KI für Anfänger](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Generative KI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![Generative KI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) + +### Generative KI Serie +[![Generative KI für Anfänger](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Generative KI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![Generative KI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) [![Generative KI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - -### Kernlernen -[![ML für Anfänger](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Datenwissenschaft für Anfänger](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![KI für Anfänger](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Cybersicherheit für Anfänger](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Webentwicklung für Anfänger](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT für Anfänger](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) + +### Kernlernen +[![ML für Anfänger](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Datenwissenschaft für Anfänger](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![KI für Anfänger](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![Cybersicherheit für Anfänger](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Web-Entwicklung für Anfänger](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT für Anfänger](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) [![XR-Entwicklung für Anfänger](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Copilot-Serie -[![Copilot für KI-Paired Programming](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot für C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) + +### Copilot Serie +[![Copilot für KI-Paarprogrammierung](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot für C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![Copilot Abenteuer](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) ## Hilfe erhalten -Wenn du feststeckst oder Fragen zum Erstellen von KI-Apps hast. Tritt anderen Lernenden und erfahrenen Entwicklern bei Diskussionen über MCP bei. Es ist eine unterstützende Gemeinschaft, in der Fragen willkommen sind und Wissen frei geteilt wird. +Falls Sie stecken bleiben oder Fragen zum Erstellen von KI-Apps haben, schließen Sie sich anderen Lernenden und erfahrenen Entwicklern zu Diskussionen über MCP an. Es ist eine unterstützende Gemeinschaft, in der Fragen willkommen sind und Wissen frei geteilt wird. [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/nTYy5BXMWG)](https://discord.gg/nTYy5BXMWG) -Wenn du Produkt-Feedback oder Fehler beim Entwickeln hast, besuche: +Wenn Sie Produktfeedback oder Fehler beim Entwickeln haben, besuchen Sie: [![Microsoft Foundry Entwicklerforum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) @@ -229,5 +221,5 @@ Wenn du Produkt-Feedback oder Fehler beim Entwickeln hast, besuche: **Haftungsausschluss**: -Dieses Dokument wurde mithilfe des KI-Übersetzungsdienstes [Co-op Translator](https://github.com/Azure/co-op-translator) übersetzt. Obwohl wir auf Genauigkeit achten, sind automatisierte Übersetzungen möglicherweise fehlerhaft oder ungenau. Das Originaldokument in seiner Ursprungssprache ist die maßgebliche Quelle. Für wichtige Informationen wird eine professionelle menschliche Übersetzung empfohlen. Wir übernehmen keine Haftung für Missverständnisse oder Fehlinterpretationen, die durch die Nutzung dieser Übersetzung entstehen. +Dieses Dokument wurde mithilfe des KI-Übersetzungsdienstes [Co-op Translator](https://github.com/Azure/co-op-translator) übersetzt. Obwohl wir auf Genauigkeit achten, beachten Sie bitte, dass automatisierte Übersetzungen Fehler oder Ungenauigkeiten enthalten können. Das Originaldokument in seiner Ursprungssprache gilt als maßgebliche Quelle. Für wichtige Informationen wird eine professionelle menschliche Übersetzung empfohlen. Wir übernehmen keine Haftung für Missverständnisse oder Fehlinterpretationen, die durch die Nutzung dieser Übersetzung entstehen. \ No newline at end of file diff --git a/translations/de/SECURITY.md b/translations/de/SECURITY.md index 023e8110..bf57c832 100644 --- a/translations/de/SECURITY.md +++ b/translations/de/SECURITY.md @@ -1,12 +1,3 @@ - ## Sicherheit Microsoft nimmt die Sicherheit seiner Softwareprodukte und -dienste sehr ernst, einschließlich aller Quellcode-Repositories, die über unsere GitHub-Organisationen verwaltet werden. Dazu gehören [Microsoft](https://github.com/Microsoft), [Azure](https://github.com/Azure), [DotNet](https://github.com/dotnet), [AspNet](https://github.com/aspnet), [Xamarin](https://github.com/xamarin) und [unsere GitHub-Organisationen](https://opensource.microsoft.com/). diff --git a/translations/de/etc/CODE_OF_CONDUCT.md b/translations/de/etc/CODE_OF_CONDUCT.md index 9311c9c1..403680e9 100644 --- a/translations/de/etc/CODE_OF_CONDUCT.md +++ b/translations/de/etc/CODE_OF_CONDUCT.md @@ -1,12 +1,3 @@ - # Microsoft Open Source Verhaltenskodex Dieses Projekt hat den [Microsoft Open Source Verhaltenskodex](https://opensource.microsoft.com/codeofconduct/) übernommen. diff --git a/translations/de/etc/CONTRIBUTING.md b/translations/de/etc/CONTRIBUTING.md index ae9ec3b1..82425af2 100644 --- a/translations/de/etc/CONTRIBUTING.md +++ b/translations/de/etc/CONTRIBUTING.md @@ -1,12 +1,3 @@ - # Beiträge Dieses Projekt begrüßt Beiträge und Vorschläge. Die meisten Beiträge erfordern, dass Sie einer Contributor License Agreement (CLA) zustimmen, die erklärt, dass Sie das Recht haben und tatsächlich gewähren, uns die Rechte zur Nutzung Ihres Beitrags zu übertragen. Weitere Details finden Sie unter https://cla.microsoft.com. diff --git a/translations/de/etc/Mindmap.md b/translations/de/etc/Mindmap.md index a69d8289..1413b4ea 100644 --- a/translations/de/etc/Mindmap.md +++ b/translations/de/etc/Mindmap.md @@ -1,12 +1,3 @@ - # KI ## [Einführung in KI](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/1-Intro/README.md) diff --git a/translations/de/etc/SUPPORT.md b/translations/de/etc/SUPPORT.md index 6cc0a58a..893d0937 100644 --- a/translations/de/etc/SUPPORT.md +++ b/translations/de/etc/SUPPORT.md @@ -1,12 +1,3 @@ - # Unterstützung ## So melden Sie Probleme und erhalten Hilfe diff --git a/translations/de/etc/TRANSLATIONS.md b/translations/de/etc/TRANSLATIONS.md index a57be792..c0761d94 100644 --- a/translations/de/etc/TRANSLATIONS.md +++ b/translations/de/etc/TRANSLATIONS.md @@ -1,12 +1,3 @@ - # Beiträge durch Übersetzung von Lektionen Wir freuen uns über Übersetzungen der Lektionen in diesem Lehrplan! diff --git a/translations/de/etc/quiz-app/README.md b/translations/de/etc/quiz-app/README.md index 0c36d298..15aa051f 100644 --- a/translations/de/etc/quiz-app/README.md +++ b/translations/de/etc/quiz-app/README.md @@ -1,12 +1,3 @@ - # Quizfragen Diese Quizfragen sind die Vor- und Nachbereitungsquizfragen für den KI-Lehrplan unter https://aka.ms/ai-beginners diff --git a/translations/de/examples/README.md b/translations/de/examples/README.md index 0df6bbb1..4738b23a 100644 --- a/translations/de/examples/README.md +++ b/translations/de/examples/README.md @@ -1,12 +1,3 @@ - # Anfängerfreundliche KI-Beispiele Willkommen! Dieses Verzeichnis enthält einfache, eigenständige Beispiele, die Ihnen den Einstieg in KI und maschinelles Lernen erleichtern sollen. Jedes Beispiel ist anfängerfreundlich gestaltet und enthält ausführliche Kommentare sowie Schritt-für-Schritt-Erklärungen. diff --git a/translations/de/lessons/0-course-setup/for-teachers.md b/translations/de/lessons/0-course-setup/for-teachers.md index 11388455..e78019be 100644 --- a/translations/de/lessons/0-course-setup/for-teachers.md +++ b/translations/de/lessons/0-course-setup/for-teachers.md @@ -1,12 +1,3 @@ - # Für Lehrkräfte Möchten Sie dieses Curriculum in Ihrem Unterricht verwenden? Fühlen Sie sich frei, es zu nutzen! diff --git a/translations/de/lessons/0-course-setup/how-to-run.md b/translations/de/lessons/0-course-setup/how-to-run.md index c3457638..1016de8f 100644 --- a/translations/de/lessons/0-course-setup/how-to-run.md +++ b/translations/de/lessons/0-course-setup/how-to-run.md @@ -1,12 +1,3 @@ - # Wie man den Code ausführt Dieses Curriculum enthält viele ausführbare Beispiele und Labs, die Sie ausführen möchten. Um dies zu tun, benötigen Sie die Möglichkeit, Python-Code in Jupyter Notebooks auszuführen, die als Teil dieses Curriculums bereitgestellt werden. Sie haben mehrere Optionen, um den Code auszuführen: diff --git a/translations/de/lessons/0-course-setup/setup.md b/translations/de/lessons/0-course-setup/setup.md index cfd99915..7cf7997a 100644 --- a/translations/de/lessons/0-course-setup/setup.md +++ b/translations/de/lessons/0-course-setup/setup.md @@ -1,12 +1,3 @@ - # Einstieg in dieses Curriculum ## Bist du ein Student? diff --git a/translations/de/lessons/1-Intro/README.md b/translations/de/lessons/1-Intro/README.md index 76dc84f6..0a86d19b 100644 --- a/translations/de/lessons/1-Intro/README.md +++ b/translations/de/lessons/1-Intro/README.md @@ -1,12 +1,3 @@ - # Einführung in KI ![Zusammenfassung der Einführung in KI in einer Skizze](../../../../translated_images/de/ai-intro.bf28d1ac4235881c.webp) diff --git a/translations/de/lessons/1-Intro/assignment.md b/translations/de/lessons/1-Intro/assignment.md index 2cba61ff..e459f2a0 100644 --- a/translations/de/lessons/1-Intro/assignment.md +++ b/translations/de/lessons/1-Intro/assignment.md @@ -1,12 +1,3 @@ - # Game Jam Spiele sind ein Bereich, der stark von Entwicklungen in KI und ML beeinflusst wurde. In dieser Aufgabe sollst du ein kurzes Essay über ein Spiel schreiben, das dir gefällt und das von der Entwicklung der KI beeinflusst wurde. Es sollte ein Spiel sein, das alt genug ist, um von verschiedenen Arten von Computersystemen beeinflusst worden zu sein. Ein gutes Beispiel ist Schach oder Go, aber schaue dir auch Videospiele wie Pong oder Pac-Man an. Schreibe einen Aufsatz, der die Vergangenheit, Gegenwart und die KI-Zukunft des Spiels diskutiert. diff --git a/translations/de/lessons/2-Symbolic/README.md b/translations/de/lessons/2-Symbolic/README.md index 127fa435..c1b2870a 100644 --- a/translations/de/lessons/2-Symbolic/README.md +++ b/translations/de/lessons/2-Symbolic/README.md @@ -1,15 +1,6 @@ - # Wissensrepräsentation und Expertensysteme -![Zusammenfassung des symbolischen KI-Inhalts](../../../../../../translated_images/de/ai-symbolic.715a30cb610411a6.webp) +![Zusammenfassung des symbolischen KI-Inhalts](../../../../translated_images/de/ai-symbolic.715a30cb610411a6.webp) > Sketchnote von [Tomomi Imura](https://twitter.com/girlie_mac) @@ -41,7 +32,7 @@ Meistens definieren wir Wissen nicht strikt, sondern ordnen es anderen verwandte Somit besteht das Problem der **Wissensrepräsentation** darin, eine effektive Methode zu finden, Wissen innerhalb eines Computers in Form von Daten so darzustellen, dass es automatisch nutzbar ist. Dies kann als Spektrum gesehen werden: -![Spektrum der Wissensrepräsentation](../../../../../../translated_images/de/knowledge-spectrum.b60df631852c0217.webp) +![Spektrum der Wissensrepräsentation](../../../../translated_images/de/knowledge-spectrum.b60df631852c0217.webp) > Bild von [Dmitry Soshnikov](http://soshnikov.com) @@ -94,7 +85,7 @@ Blocksyntax | Einrückung | | | Einer der frühen Erfolge der symbolischen KI waren sogenannte **Expertensysteme** – Computersysteme, die entwickelt wurden, um als Experte in einem begrenzten Problembereich zu wirken. Sie basierten auf einer **Wissensbasis**, die von einem oder mehreren menschlichen Experten extrahiert wurde, und enthielten eine **Inferenzmaschine**, die darauf basierend Schlussfolgerungen zog. -![Menschliche Architektur](../../../../../../translated_images/de/arch-human.5d4d35f1bba3ab1c.webp) | ![Wissensbasiertes System](../../../../../../translated_images/de/arch-kbs.3ec5c150b09fa8da.webp) +![Menschliche Architektur](../../../../translated_images/de/arch-human.5d4d35f1bba3ab1c.webp) | ![Wissensbasiertes System](../../../../translated_images/de/arch-kbs.3ec5c150b09fa8da.webp) ---------------------------------------------|------------------------------------------------ Vereinfachte Struktur des menschlichen neuronalen Systems | Architektur eines wissensbasierten Systems @@ -106,7 +97,7 @@ Expertensysteme sind ähnlich wie das menschliche Schlusssystem aufgebaut, das * Als Beispiel betrachten wir folgendes Expertensystem zur Bestimmung eines Tieres anhand seiner physischen Eigenschaften: -![AND-OR-Baum](../../../../../../translated_images/de/AND-OR-Tree.5592d2c70187f283.webp) +![AND-OR-Baum](../../../../translated_images/de/AND-OR-Tree.5592d2c70187f283.webp) > Bild von [Dmitry Soshnikov](http://soshnikov.com) diff --git a/translations/de/lessons/2-Symbolic/assignment.md b/translations/de/lessons/2-Symbolic/assignment.md index aa23100e..6b44b8c5 100644 --- a/translations/de/lessons/2-Symbolic/assignment.md +++ b/translations/de/lessons/2-Symbolic/assignment.md @@ -1,12 +1,3 @@ - # Erstellen Sie eine Ontologie Eine Wissensbasis zu erstellen bedeutet, ein Modell zu kategorisieren, das Fakten zu einem bestimmten Thema darstellt. Wählen Sie ein Thema – wie eine Person, einen Ort oder ein Objekt – und erstellen Sie dann ein Modell zu diesem Thema. Verwenden Sie einige der in dieser Lektion beschriebenen Techniken und Strategien zur Modellerstellung. Ein Beispiel wäre das Erstellen einer Ontologie eines Wohnzimmers mit Möbeln, Lampen und so weiter. Worin unterscheidet sich das Wohnzimmer von der Küche? Dem Badezimmer? Woran erkennen Sie, dass es sich um ein Wohnzimmer und nicht um ein Esszimmer handelt? Verwenden Sie [Protégé](https://protege.stanford.edu/), um Ihre Ontologie zu erstellen. diff --git a/translations/de/lessons/3-NeuralNetworks/03-Perceptron/README.md b/translations/de/lessons/3-NeuralNetworks/03-Perceptron/README.md index 376b77bb..05459742 100644 --- a/translations/de/lessons/3-NeuralNetworks/03-Perceptron/README.md +++ b/translations/de/lessons/3-NeuralNetworks/03-Perceptron/README.md @@ -1,12 +1,3 @@ - # Einführung in Neuronale Netzwerke: Perzeptron ## [Quiz vor der Vorlesung](https://ff-quizzes.netlify.app/en/ai/quiz/5) @@ -15,7 +6,7 @@ Einer der ersten Versuche, etwas Ähnliches wie ein modernes neuronales Netzwerk | | | |--------------|-----------| -|Frank Rosenblatt | Das Mark 1 Perzeptron| +|Frank Rosenblatt | Das Mark 1 Perzeptron| > Bilder [von Wikipedia](https://en.wikipedia.org/wiki/Perceptron) @@ -34,7 +25,7 @@ y(x) = f(wTx) wobei f eine Stufenaktivierungsfunktion ist. - + ## Training des Perzeptrons diff --git a/translations/de/lessons/3-NeuralNetworks/03-Perceptron/lab/README.md b/translations/de/lessons/3-NeuralNetworks/03-Perceptron/lab/README.md index 128e73ad..8b548f43 100644 --- a/translations/de/lessons/3-NeuralNetworks/03-Perceptron/lab/README.md +++ b/translations/de/lessons/3-NeuralNetworks/03-Perceptron/lab/README.md @@ -1,12 +1,3 @@ - # Mehrklassenklassifikation mit Perzeptron Laboraufgabe aus dem [AI for Beginners Curriculum](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/de/lessons/3-NeuralNetworks/04-OwnFramework/README.md b/translations/de/lessons/3-NeuralNetworks/04-OwnFramework/README.md index 2e8e0ba2..7529c51c 100644 --- a/translations/de/lessons/3-NeuralNetworks/04-OwnFramework/README.md +++ b/translations/de/lessons/3-NeuralNetworks/04-OwnFramework/README.md @@ -1,12 +1,3 @@ - # Einführung in Neuronale Netze. Mehrschichtige Perzeptrons Im vorherigen Abschnitt hast du das einfachste Modell eines neuronalen Netzes kennengelernt – ein einlagiges Perzeptron, ein lineares Modell zur Klassifikation von zwei Klassen. @@ -65,7 +56,7 @@ Der Algorithmus des Gradientenabstiegs bleibt derselbe, aber die Berechnung der Beachte, dass der linke Teil all dieser Ausdrücke derselbe ist, und daher können wir die Ableitungen effektiv berechnen, indem wir von der Verlustfunktion ausgehend "rückwärts" durch den Berechnungsgraphen gehen. Daher wird die Methode zum Trainieren eines mehrschichtigen Perzeptrons als **Backpropagation** oder 'Backprop' bezeichnet. -Berechnungsgraph +Berechnungsgraph > TODO: Bildquelle angeben diff --git a/translations/de/lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md b/translations/de/lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md index 3361ee06..2a3cdff5 100644 --- a/translations/de/lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md +++ b/translations/de/lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md @@ -1,12 +1,3 @@ - # MNIST-Klassifikation mit unserem eigenen Framework Laboraufgabe aus dem [AI for Beginners Curriculum](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/de/lessons/3-NeuralNetworks/05-Frameworks/README.md b/translations/de/lessons/3-NeuralNetworks/05-Frameworks/README.md index 4945e419..bac9e40f 100644 --- a/translations/de/lessons/3-NeuralNetworks/05-Frameworks/README.md +++ b/translations/de/lessons/3-NeuralNetworks/05-Frameworks/README.md @@ -1,12 +1,3 @@ - # Neuronale Netzwerk-Frameworks Wie wir bereits gelernt haben, müssen wir zwei Dinge tun, um neuronale Netzwerke effizient trainieren zu können: diff --git a/translations/de/lessons/3-NeuralNetworks/05-Frameworks/lab/README.md b/translations/de/lessons/3-NeuralNetworks/05-Frameworks/lab/README.md index 5c8e2c9f..418b36f3 100644 --- a/translations/de/lessons/3-NeuralNetworks/05-Frameworks/lab/README.md +++ b/translations/de/lessons/3-NeuralNetworks/05-Frameworks/lab/README.md @@ -1,12 +1,3 @@ - # Klassifikation mit PyTorch/TensorFlow Laboraufgabe aus dem [AI for Beginners Curriculum](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/de/lessons/3-NeuralNetworks/README.md b/translations/de/lessons/3-NeuralNetworks/README.md index c0885544..5fb97a1b 100644 --- a/translations/de/lessons/3-NeuralNetworks/README.md +++ b/translations/de/lessons/3-NeuralNetworks/README.md @@ -1,12 +1,3 @@ - # Einführung in Neuronale Netzwerke ![Zusammenfassung des Inhalts zu neuronalen Netzwerken in einer Skizze](../../../../translated_images/de/ai-neuralnetworks.1c687ae40bc86e83.webp) diff --git a/translations/de/lessons/4-ComputerVision/06-IntroCV/README.md b/translations/de/lessons/4-ComputerVision/06-IntroCV/README.md index e69872f1..0d374d8c 100644 --- a/translations/de/lessons/4-ComputerVision/06-IntroCV/README.md +++ b/translations/de/lessons/4-ComputerVision/06-IntroCV/README.md @@ -1,12 +1,3 @@ - # Einführung in Computer Vision [Computer Vision](https://wikipedia.org/wiki/Computer_vision) ist ein Fachgebiet, dessen Ziel es ist, Computern ein hochentwickeltes Verständnis von digitalen Bildern zu ermöglichen. Dies ist eine recht breite Definition, da *Verständnis* viele verschiedene Dinge bedeuten kann, wie z. B. das Finden eines Objekts auf einem Bild (**Objekterkennung**), das Verstehen, was passiert (**Ereigniserkennung**), das Beschreiben eines Bildes in Text oder die Rekonstruktion einer Szene in 3D. Es gibt auch spezielle Aufgaben im Zusammenhang mit menschlichen Bildern: Alter- und Emotionserkennung, Gesichtserkennung und -identifikation sowie 3D-Pose-Schätzung, um nur einige zu nennen. @@ -115,7 +106,7 @@ Lesen Sie mehr über optischen Fluss [in diesem großartigen Tutorial](https://l In diesem Labor werden Sie ein Video mit einfachen Gesten aufnehmen, und Ihr Ziel ist es, Aufwärts-/Abwärts-/Links-/Rechts-Bewegungen mithilfe des optischen Flusses zu extrahieren. -Palm Movement Frame +Palm Movement Frame --- diff --git a/translations/de/lessons/4-ComputerVision/06-IntroCV/lab/README.md b/translations/de/lessons/4-ComputerVision/06-IntroCV/lab/README.md index ef71bedd..6ceaaa09 100644 --- a/translations/de/lessons/4-ComputerVision/06-IntroCV/lab/README.md +++ b/translations/de/lessons/4-ComputerVision/06-IntroCV/lab/README.md @@ -1,12 +1,3 @@ - # Bewegungen mit Optical Flow erkennen Laboraufgabe aus dem [AI for Beginners Curriculum](https://aka.ms/ai-beginners). diff --git a/translations/de/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md b/translations/de/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md index 0d831b95..985f6ea3 100644 --- a/translations/de/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md +++ b/translations/de/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md @@ -1,12 +1,3 @@ - # Bekannte CNN-Architekturen ### VGG-16 @@ -25,7 +16,7 @@ Wie man sehen kann, folgt VGG einer traditionellen Pyramidenarchitektur, die aus ResNet ist eine Modellfamilie, die 2015 von Microsoft Research vorgeschlagen wurde. Die Hauptidee von ResNet ist die Verwendung von **Residual Blocks**: - + > Bild aus [diesem Paper](https://arxiv.org/pdf/1512.03385.pdf) @@ -37,7 +28,7 @@ Man kann sich dieses Netzwerk auch so vorstellen, dass es seine Komplexität an Die Google Inception-Architektur geht einen Schritt weiter und baut jede Netzwerkschicht als Kombination aus mehreren verschiedenen Pfaden: - + > Bild von [Researchgate](https://www.researchgate.net/figure/Inception-module-with-dimension-reductions-left-and-schema-for-Inception-ResNet-v1_fig2_355547454) diff --git a/translations/de/lessons/4-ComputerVision/07-ConvNets/README.md b/translations/de/lessons/4-ComputerVision/07-ConvNets/README.md index 713a0641..6b750776 100644 --- a/translations/de/lessons/4-ComputerVision/07-ConvNets/README.md +++ b/translations/de/lessons/4-ComputerVision/07-ConvNets/README.md @@ -1,12 +1,3 @@ - # Convolutional Neural Networks Wir haben bereits gesehen, dass neuronale Netze recht gut mit Bildern umgehen können, und selbst ein Ein-Schicht-Perzeptron ist in der Lage, handgeschriebene Ziffern aus dem MNIST-Datensatz mit einer akzeptablen Genauigkeit zu erkennen. Der MNIST-Datensatz ist jedoch etwas Besonderes, da alle Ziffern zentriert im Bild dargestellt sind, was die Aufgabe vereinfacht. @@ -24,7 +15,7 @@ Um Muster zu extrahieren, verwenden wir den Begriff der **Faltung (Convolutional Wenn wir beispielsweise 3x3-Filter für vertikale und horizontale Kanten auf die MNIST-Ziffern anwenden, können wir Bereiche hervorheben (z. B. hohe Werte), in denen es vertikale und horizontale Kanten im Originalbild gibt. Diese beiden Filter können also verwendet werden, um "nach" Kanten zu suchen. Ebenso können wir verschiedene Filter entwerfen, um nach anderen grundlegenden Mustern zu suchen: - + > Bild des [Leung-Malik Filter Bank](https://www.robots.ox.ac.uk/~vgg/research/texclass/filters.html) diff --git a/translations/de/lessons/4-ComputerVision/07-ConvNets/lab/README.md b/translations/de/lessons/4-ComputerVision/07-ConvNets/lab/README.md index d0746fa7..1231dbbe 100644 --- a/translations/de/lessons/4-ComputerVision/07-ConvNets/lab/README.md +++ b/translations/de/lessons/4-ComputerVision/07-ConvNets/lab/README.md @@ -1,12 +1,3 @@ - # Klassifikation von Haustiergesichtern Laboraufgabe aus dem [AI for Beginners Curriculum](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/de/lessons/4-ComputerVision/08-TransferLearning/README.md b/translations/de/lessons/4-ComputerVision/08-TransferLearning/README.md index 4d014280..4cc8a387 100644 --- a/translations/de/lessons/4-ComputerVision/08-TransferLearning/README.md +++ b/translations/de/lessons/4-ComputerVision/08-TransferLearning/README.md @@ -1,12 +1,3 @@ - # Vorgefertigte Netzwerke und Transfer-Learning Das Training von CNNs kann viel Zeit in Anspruch nehmen, und es wird eine große Menge an Daten dafür benötigt. Ein Großteil der Zeit wird jedoch darauf verwendet, die besten Low-Level-Filter zu lernen, die ein Netzwerk nutzen kann, um Muster aus Bildern zu extrahieren. Eine natürliche Frage stellt sich: Können wir ein neuronales Netzwerk, das auf einem Datensatz trainiert wurde, verwenden und anpassen, um andere Bilder zu klassifizieren, ohne den gesamten Trainingsprozess erneut durchlaufen zu müssen? diff --git a/translations/de/lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md b/translations/de/lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md index 0bef203d..b866be4d 100644 --- a/translations/de/lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md +++ b/translations/de/lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md @@ -1,12 +1,3 @@ - # Deep Learning Training Tricks Je tiefer neuronale Netzwerke werden, desto herausfordernder wird ihr Training. Ein Hauptproblem sind die sogenannten [verschwindenden Gradienten](https://en.wikipedia.org/wiki/Vanishing_gradient_problem) oder [explodierenden Gradienten](https://deepai.org/machine-learning-glossary-and-terms/exploding-gradient-problem#:~:text=Exploding%20gradients%20are%20a%20problem,updates%20are%20small%20and%20controlled.). [Dieser Artikel](https://towardsdatascience.com/the-vanishing-exploding-gradient-problem-in-deep-neural-networks-191358470c11) bietet eine gute Einführung in diese Probleme. diff --git a/translations/de/lessons/4-ComputerVision/08-TransferLearning/lab/README.md b/translations/de/lessons/4-ComputerVision/08-TransferLearning/lab/README.md index 2d2ec247..a05e67d9 100644 --- a/translations/de/lessons/4-ComputerVision/08-TransferLearning/lab/README.md +++ b/translations/de/lessons/4-ComputerVision/08-TransferLearning/lab/README.md @@ -1,12 +1,3 @@ - # Klassifikation von Oxford-Haustieren mit Transfer-Learning Laboraufgabe aus dem [AI for Beginners Curriculum](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/de/lessons/4-ComputerVision/09-Autoencoders/README.md b/translations/de/lessons/4-ComputerVision/09-Autoencoders/README.md index 19b449f7..fc4da168 100644 --- a/translations/de/lessons/4-ComputerVision/09-Autoencoders/README.md +++ b/translations/de/lessons/4-ComputerVision/09-Autoencoders/README.md @@ -1,12 +1,3 @@ - # Autoencoder Beim Training von CNNs besteht eines der Probleme darin, dass wir eine große Menge an gelabelten Daten benötigen. Im Fall der Bildklassifikation müssen wir Bilder in verschiedene Klassen unterteilen, was manuell erfolgen muss. @@ -46,7 +37,7 @@ Zusammengefasst: * Wir entnehmen einen Vektor `sample` aus der Verteilung N(zmean,exp(zlog\_sigma)). * Der Decoder versucht, das ursprüngliche Bild mithilfe von `sample` als Eingabevektor zu dekodieren. - + > Bild aus [diesem Blogbeitrag](https://ijdykeman.github.io/ml/2016/12/21/cvae.html) von Isaak Dykeman @@ -57,13 +48,13 @@ Variationale Autoencoder verwenden eine komplexe Verlustfunktion, die aus zwei T Ein wichtiger Vorteil von VAEs ist, dass sie es uns ermöglichen, relativ einfach neue Bilder zu generieren, da wir wissen, aus welcher Verteilung wir latente Vektoren entnehmen müssen. Wenn wir beispielsweise einen VAE mit einem 2D-latenten Vektor auf MNIST trainieren, können wir die Komponenten des latenten Vektors variieren, um verschiedene Ziffern zu erhalten: -vaemnist +vaemnist > Bild von [Dmitry Soshnikov](http://soshnikov.com) Beobachten Sie, wie Bilder ineinander übergehen, wenn wir beginnen, latente Vektoren aus verschiedenen Bereichen des latenten Parameterraums zu entnehmen. Wir können diesen Raum auch in 2D visualisieren: -vaemnist cluster +vaemnist cluster > Bild von [Dmitry Soshnikov](http://soshnikov.com) diff --git a/translations/de/lessons/4-ComputerVision/10-GANs/README.md b/translations/de/lessons/4-ComputerVision/10-GANs/README.md index 01c81acf..2547f7cd 100644 --- a/translations/de/lessons/4-ComputerVision/10-GANs/README.md +++ b/translations/de/lessons/4-ComputerVision/10-GANs/README.md @@ -1,12 +1,3 @@ - # Generative Adversarial Networks Im vorherigen Abschnitt haben wir über **generative Modelle** gelernt: Modelle, die neue Bilder erzeugen können, die den Bildern im Trainingsdatensatz ähneln. VAE war ein gutes Beispiel für ein generatives Modell. @@ -17,7 +8,7 @@ Wenn wir jedoch versuchen, etwas wirklich Bedeutungsvolles zu erzeugen, wie ein Die Hauptidee eines GANs besteht darin, zwei neuronale Netzwerke zu haben, die gegeneinander trainiert werden: - + > Bild von [Dmitry Soshnikov](http://soshnikov.com) @@ -41,7 +32,7 @@ Ein Generator ist etwas komplizierter. Man kann ihn als umgekehrten Diskriminato > ✅ Da die Convolution-Schicht als linearer Filter implementiert ist, der das Bild durchläuft, ist Deconvolution im Wesentlichen ähnlich wie Convolution und kann mit derselben Schichtlogik implementiert werden. - + > Bild von [Dmitry Soshnikov](http://soshnikov.com) diff --git a/translations/de/lessons/4-ComputerVision/11-ObjectDetection/README.md b/translations/de/lessons/4-ComputerVision/11-ObjectDetection/README.md index b4d67baa..290049fc 100644 --- a/translations/de/lessons/4-ComputerVision/11-ObjectDetection/README.md +++ b/translations/de/lessons/4-ComputerVision/11-ObjectDetection/README.md @@ -1,12 +1,3 @@ - # Objekterkennung Die Bildklassifizierungsmodelle, die wir bisher behandelt haben, nahmen ein Bild und lieferten ein kategorisches Ergebnis, wie beispielsweise die Klasse "Zahl" in einem MNIST-Problem. In vielen Fällen möchten wir jedoch nicht nur wissen, dass ein Bild Objekte darstellt – wir möchten auch deren genaue Position bestimmen können. Genau darum geht es bei der **Objekterkennung**. diff --git a/translations/de/lessons/4-ComputerVision/11-ObjectDetection/lab/README.md b/translations/de/lessons/4-ComputerVision/11-ObjectDetection/lab/README.md index ff33bbdf..03ce40c2 100644 --- a/translations/de/lessons/4-ComputerVision/11-ObjectDetection/lab/README.md +++ b/translations/de/lessons/4-ComputerVision/11-ObjectDetection/lab/README.md @@ -1,12 +1,3 @@ - # Kopferkennung mit dem Hollywood Heads Dataset Laboraufgabe aus dem [AI for Beginners Curriculum](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/de/lessons/4-ComputerVision/12-Segmentation/README.md b/translations/de/lessons/4-ComputerVision/12-Segmentation/README.md index f03ec586..1a935a32 100644 --- a/translations/de/lessons/4-ComputerVision/12-Segmentation/README.md +++ b/translations/de/lessons/4-ComputerVision/12-Segmentation/README.md @@ -1,12 +1,3 @@ - # Segmentierung Wir haben zuvor über Objekterkennung gelernt, die es uns ermöglicht, Objekte im Bild zu lokalisieren, indem ihre *begrenzenden Boxen* vorhergesagt werden. Für einige Aufgaben benötigen wir jedoch nicht nur begrenzende Boxen, sondern auch eine präzisere Objektlokalisierung. Diese Aufgabe nennt man **Segmentierung**. @@ -20,7 +11,7 @@ Segmentierung kann als **Pixelklassifikation** betrachtet werden, wobei für **j Bei der Instanzsegmentierung sind diese Schafe unterschiedliche Objekte, aber bei der semantischen Segmentierung werden alle Schafe durch eine Klasse repräsentiert. - + > Bild aus [diesem Blogbeitrag](https://nirmalamurali.medium.com/image-classification-vs-semantic-segmentation-vs-instance-segmentation-625c33a08d50) @@ -29,7 +20,7 @@ Es gibt verschiedene neuronale Architekturen für die Segmentierung, aber sie ha * **Encoder** extrahiert Merkmale aus dem Eingabebild. * **Decoder** transformiert diese Merkmale in das **Maskenbild**, mit derselben Größe und einer Anzahl von Kanälen, die der Anzahl der Klassen entspricht. - + > Bild aus [dieser Publikation](https://arxiv.org/pdf/2001.05566.pdf) @@ -43,7 +34,7 @@ In dieser Lektion werden wir die Segmentierung in Aktion sehen, indem wir ein Ne > ✅ Diese Technik ist besonders geeignet für diese Art der medizinischen Bildgebung, aber welche anderen Anwendungen in der realen Welt könnten Sie sich vorstellen? -navi +navi > Bild aus der PH2-Datenbank diff --git a/translations/de/lessons/4-ComputerVision/12-Segmentation/lab/README.md b/translations/de/lessons/4-ComputerVision/12-Segmentation/lab/README.md index 38452536..da9aa6c9 100644 --- a/translations/de/lessons/4-ComputerVision/12-Segmentation/lab/README.md +++ b/translations/de/lessons/4-ComputerVision/12-Segmentation/lab/README.md @@ -1,12 +1,3 @@ - # Segmentierung des menschlichen Körpers Laboraufgabe aus dem [AI for Beginners Curriculum](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/de/lessons/4-ComputerVision/README.md b/translations/de/lessons/4-ComputerVision/README.md index 659c6e2f..cf975b79 100644 --- a/translations/de/lessons/4-ComputerVision/README.md +++ b/translations/de/lessons/4-ComputerVision/README.md @@ -1,12 +1,3 @@ - # Computer Vision ![Zusammenfassung der Inhalte zu Computer Vision in einer Skizze](../../../../lessons/sketchnotes/ai-computervision.png) diff --git a/translations/de/lessons/5-NLP/13-TextRep/README.md b/translations/de/lessons/5-NLP/13-TextRep/README.md index ff8b529c..dfdcb6cc 100644 --- a/translations/de/lessons/5-NLP/13-TextRep/README.md +++ b/translations/de/lessons/5-NLP/13-TextRep/README.md @@ -1,12 +1,3 @@ - # Darstellung von Text als Tensoren ## [Quiz vor der Vorlesung](https://ff-quizzes.netlify.app/en/ai/quiz/25) @@ -25,7 +16,7 @@ Unser Ziel wird es sein, den Nachrichtenartikel basierend auf dem Text einer der Um Aufgaben der natürlichen Sprachverarbeitung (NLP) mit neuronalen Netzwerken zu lösen, benötigen wir eine Methode, um Text als Tensoren darzustellen. Computer repräsentieren Textzeichen bereits als Zahlen, die mit Codierungen wie ASCII oder UTF-8 auf die Schriftarten auf Ihrem Bildschirm abgebildet werden. -Bild zeigt ein Diagramm, das ein Zeichen mit einer ASCII- und Binärdarstellung verknüpft +Bild zeigt ein Diagramm, das ein Zeichen mit einer ASCII- und Binärdarstellung verknüpft > [Bildquelle](https://www.seobility.net/en/wiki/ASCII) @@ -48,7 +39,7 @@ In einigen Fällen können wir auch die Verwendung von Tri-Grammen – Kombinati Bei Aufgaben wie der Textklassifikation müssen wir in der Lage sein, Text durch einen festen Vektor darzustellen, den wir als Eingabe für den abschließenden dichten Klassifikator verwenden. Eine der einfachsten Möglichkeiten, dies zu tun, besteht darin, alle individuellen Wortdarstellungen zu kombinieren, z. B. durch Addition. Wenn wir die One-Hot-Codierungen jedes Wortes addieren, erhalten wir einen Frequenzvektor, der zeigt, wie oft jedes Wort im Text vorkommt. Eine solche Darstellung von Text wird als **Bag-of-Words** (BoW) bezeichnet. - + > Bild vom Autor diff --git a/translations/de/lessons/5-NLP/13-TextRep/assignment.md b/translations/de/lessons/5-NLP/13-TextRep/assignment.md index c0430ca0..2c14bc88 100644 --- a/translations/de/lessons/5-NLP/13-TextRep/assignment.md +++ b/translations/de/lessons/5-NLP/13-TextRep/assignment.md @@ -1,12 +1,3 @@ - # Aufgabe: Notebooks Verwenden Sie die Notebooks, die zu dieser Lektion gehören (entweder die PyTorch- oder die TensorFlow-Version), und führen Sie sie mit Ihrem eigenen Datensatz aus, möglicherweise einem von Kaggle, unter Angabe der Quelle. Überarbeiten Sie das Notebook, um Ihre eigenen Erkenntnisse hervorzuheben. Probieren Sie einige innovative Datensätze aus, die überraschend sein könnten, wie zum Beispiel [diesen über UFO-Sichtungen](https://www.kaggle.com/datasets/NUFORC/ufo-sightings) von NUFORC. diff --git a/translations/de/lessons/5-NLP/14-Embeddings/README.md b/translations/de/lessons/5-NLP/14-Embeddings/README.md index 3647f5ff..62cdb564 100644 --- a/translations/de/lessons/5-NLP/14-Embeddings/README.md +++ b/translations/de/lessons/5-NLP/14-Embeddings/README.md @@ -1,12 +1,3 @@ - # Einbettungen ## [Quiz vor der Vorlesung](https://ff-quizzes.netlify.app/en/ai/quiz/27) diff --git a/translations/de/lessons/5-NLP/14-Embeddings/assignment.md b/translations/de/lessons/5-NLP/14-Embeddings/assignment.md index 7eb319b0..ab511dfe 100644 --- a/translations/de/lessons/5-NLP/14-Embeddings/assignment.md +++ b/translations/de/lessons/5-NLP/14-Embeddings/assignment.md @@ -1,12 +1,3 @@ - # Aufgabe: Notebooks Verwenden Sie die Notebooks, die mit dieser Lektion verbunden sind (entweder die PyTorch- oder die TensorFlow-Version), und führen Sie sie mit Ihrem eigenen Datensatz erneut aus, vielleicht einem von Kaggle, der mit entsprechender Quellenangabe verwendet wird. Überarbeiten Sie das Notebook, um Ihre eigenen Erkenntnisse hervorzuheben. Probieren Sie eine andere Art von Datensatz aus und dokumentieren Sie Ihre Ergebnisse, indem Sie beispielsweise Texte wie [diese Beatles-Songtexte](https://www.kaggle.com/datasets/jenlooper/beatles-lyrics) verwenden. diff --git a/translations/de/lessons/5-NLP/15-LanguageModeling/README.md b/translations/de/lessons/5-NLP/15-LanguageModeling/README.md index 24649cf5..1f85b8d4 100644 --- a/translations/de/lessons/5-NLP/15-LanguageModeling/README.md +++ b/translations/de/lessons/5-NLP/15-LanguageModeling/README.md @@ -1,12 +1,3 @@ - # Sprachmodellierung Semantische Einbettungen wie Word2Vec und GloVe sind tatsächlich ein erster Schritt in Richtung **Sprachmodellierung** – Modelle zu erstellen, die irgendwie die *Natur* der Sprache *verstehen* (oder *repräsentieren*). diff --git a/translations/de/lessons/5-NLP/15-LanguageModeling/lab/README.md b/translations/de/lessons/5-NLP/15-LanguageModeling/lab/README.md index d6deaa49..b2916f76 100644 --- a/translations/de/lessons/5-NLP/15-LanguageModeling/lab/README.md +++ b/translations/de/lessons/5-NLP/15-LanguageModeling/lab/README.md @@ -1,12 +1,3 @@ - # Training eines Skip-Gram-Modells Laboraufgabe aus dem [AI for Beginners Curriculum](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/de/lessons/5-NLP/16-RNN/README.md b/translations/de/lessons/5-NLP/16-RNN/README.md index 91a9e1ac..3cc4b376 100644 --- a/translations/de/lessons/5-NLP/16-RNN/README.md +++ b/translations/de/lessons/5-NLP/16-RNN/README.md @@ -1,12 +1,3 @@ - # Rekurrente Neuronale Netze ## [Quiz vor der Vorlesung](https://ff-quizzes.netlify.app/en/ai/quiz/31) @@ -31,7 +22,7 @@ Schauen wir uns an, wie eine einfache RNN-Zelle organisiert ist. Sie akzeptiert Eine einfache RNN-Zelle hat zwei Gewichtsmatrizen: eine transformiert ein Eingabesymbol (wir nennen sie W), und eine andere transformiert einen Eingabezustand (H). In diesem Fall wird die Ausgabe des Netzwerks als σ(W×Xi+H×Si-1+b) berechnet, wobei σ die Aktivierungsfunktion und b ein zusätzlicher Bias ist. -Anatomie einer RNN-Zelle +Anatomie einer RNN-Zelle > Bild vom Autor diff --git a/translations/de/lessons/5-NLP/16-RNN/assignment.md b/translations/de/lessons/5-NLP/16-RNN/assignment.md index df4fd5b3..142d9c09 100644 --- a/translations/de/lessons/5-NLP/16-RNN/assignment.md +++ b/translations/de/lessons/5-NLP/16-RNN/assignment.md @@ -1,12 +1,3 @@ - # Aufgabe: Notebooks Verwenden Sie die Notebooks, die mit dieser Lektion verbunden sind (entweder die PyTorch- oder die TensorFlow-Version), und führen Sie sie mit Ihrem eigenen Datensatz erneut aus, möglicherweise einem von Kaggle, unter Angabe der Quelle. Überarbeiten Sie das Notebook, um Ihre eigenen Erkenntnisse hervorzuheben. Probieren Sie eine andere Art von Datensatz aus und dokumentieren Sie Ihre Ergebnisse, beispielsweise mit Text wie [diesem Kaggle-Wettbewerbsdatensatz über Wetter-Tweets](https://www.kaggle.com/competitions/crowdflower-weather-twitter/data?select=train.csv). diff --git a/translations/de/lessons/5-NLP/17-GenerativeNetworks/README.md b/translations/de/lessons/5-NLP/17-GenerativeNetworks/README.md index 26170669..597d6ce2 100644 --- a/translations/de/lessons/5-NLP/17-GenerativeNetworks/README.md +++ b/translations/de/lessons/5-NLP/17-GenerativeNetworks/README.md @@ -1,12 +1,3 @@ - # Generative Netzwerke ## [Quiz vor der Vorlesung](https://ff-quizzes.netlify.app/en/ai/quiz/33) @@ -36,7 +27,7 @@ Wir werden dieses RNN trainieren, um Text Schritt für Schritt zu generieren. Be Bei der Textgenerierung (während der Inferenz) beginnen wir mit einem **Prompt**, der durch die RNN-Zellen geleitet wird, um seinen Zwischenzustand zu erzeugen, und dann beginnt die Generierung aus diesem Zustand. Wir generieren ein Zeichen nach dem anderen und übergeben den Zustand und das generierte Zeichen an eine andere RNN-Zelle, um das nächste zu generieren, bis wir genügend Zeichen erzeugt haben. - + > Bild vom Autor diff --git a/translations/de/lessons/5-NLP/17-GenerativeNetworks/lab/README.md b/translations/de/lessons/5-NLP/17-GenerativeNetworks/lab/README.md index 31a4156f..80239898 100644 --- a/translations/de/lessons/5-NLP/17-GenerativeNetworks/lab/README.md +++ b/translations/de/lessons/5-NLP/17-GenerativeNetworks/lab/README.md @@ -1,12 +1,3 @@ - # Wortbasierte Textgenerierung mit RNNs Laboraufgabe aus dem [AI for Beginners Curriculum](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/de/lessons/5-NLP/18-Transformers/README.md b/translations/de/lessons/5-NLP/18-Transformers/README.md index ac09b3ff..28e6316e 100644 --- a/translations/de/lessons/5-NLP/18-Transformers/README.md +++ b/translations/de/lessons/5-NLP/18-Transformers/README.md @@ -1,12 +1,3 @@ - # Aufmerksamkeitsmechanismen und Transformer ## [Quiz vor der Vorlesung](https://ff-quizzes.netlify.app/en/ai/quiz/35) @@ -56,7 +47,7 @@ Die Idee der Positionskodierung ist folgende: * Trainierbare Einbettung, ähnlich wie Token-Einbettung. Dies ist der Ansatz, den wir hier betrachten. Wir wenden Einbettungsschichten sowohl auf Tokens als auch auf ihre Positionen an, was zu Einbettungsvektoren mit denselben Dimensionen führt, die wir dann zusammen addieren. * Feste Positionskodierungsfunktion, wie im ursprünglichen Paper vorgeschlagen. - + > Bild vom Autor diff --git a/translations/de/lessons/5-NLP/18-Transformers/READMEtransformers.md b/translations/de/lessons/5-NLP/18-Transformers/READMEtransformers.md deleted file mode 100644 index 21501ab3..00000000 --- a/translations/de/lessons/5-NLP/18-Transformers/READMEtransformers.md +++ /dev/null @@ -1,112 +0,0 @@ -# Aufmerksamkeitsmechanismen und Transformer - -## [Vorlesungsquiz](https://ff-quizzes.netlify.app/en/ai/quiz/35) - -Eines der wichtigsten Probleme im Bereich der NLP ist die **maschinelle Übersetzung**, eine grundlegende Aufgabe, die Tools wie Google Translate zugrunde liegt. In diesem Abschnitt konzentrieren wir uns auf maschinelle Übersetzung oder, allgemeiner, auf jede *Sequenz-zu-Sequenz*-Aufgabe (die auch als **Satztransduktion** bezeichnet wird). - -Bei RNNs wird Sequenz-zu-Sequenz durch zwei rekursive Netzwerke implementiert, wobei ein Netzwerk, der **Encoder**, eine Eingabesequenz in einen verborgenen Zustand zusammenfasst, während ein anderes Netzwerk, der **Decoder**, diesen verborgenen Zustand in ein übersetztes Ergebnis entfaltet. Es gibt einige Probleme mit diesem Ansatz: - -* Der endgültige Zustand des Encoder-Netzwerks hat Schwierigkeiten, sich an den Anfang eines Satzes zu erinnern, was zu einer schlechten Modellqualität bei langen Sätzen führt. -* Alle Wörter in einer Sequenz haben den gleichen Einfluss auf das Ergebnis. In der Realität haben jedoch bestimmte Wörter in der Eingabesequenz oft einen größeren Einfluss auf die sequenziellen Ausgaben als andere. - -**Aufmerksamkeitsmechanismen** bieten ein Mittel, um den kontextuellen Einfluss jedes Eingangsvektors auf jede Vorhersageausgabe des RNN zu gewichten. Die Implementierung erfolgt durch die Schaffung von Abkürzungen zwischen den Zwischenzuständen des Eingangs-RNN und dem Ausgangs-RNN. Auf diese Weise berücksichtigen wir beim Generieren des Ausgabesymbols yt alle Eingangsversteckzustände hi mit unterschiedlichen Gewichtungskoeffizienten αt,i. - -![Bild, das ein Encoder/Decoder-Modell mit einer additiven Aufmerksamkeitschicht zeigt](../../../../../translated_images/de/encoder-decoder-attention.7a726296894fb567.webp) - -> Das Encoder-Decoder-Modell mit dem additiven Aufmerksamkeitsmechanismus in [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf), zitiert aus [diesem Blogbeitrag](https://lilianweng.github.io/lil-log/2018/06/24/attention-attention.html) - -Die Aufmerksamkeitsmatrix {αi,j} würde den Grad darstellen, in dem bestimmte Eingabewörter an der Generierung eines bestimmten Wortes in der Ausgabesequenz beteiligt sind. Im Folgenden finden Sie ein Beispiel für eine solche Matrix: - -![Bild, das eine Beispielausrichtung zeigt, die von RNNsearch-50 gefunden wurde, entnommen von Bahdanau - arviz.org](../../../../../translated_images/de/bahdanau-fig3.09ba2d37f202a6af.webp) - -> Abbildung aus [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf) (Fig.3) - -Aufmerksamkeitsmechanismen sind für einen Großteil des aktuellen oder nahezu aktuellen Stands der Technik im NLP verantwortlich. Die Hinzufügung von Aufmerksamkeit erhöht jedoch erheblich die Anzahl der Modellparameter, was zu Skalierungsproblemen bei RNNs führte. Eine wichtige Einschränkung bei der Skalierung von RNNs ist, dass die rekursive Natur der Modelle es schwierig macht, das Training zu batchen und zu parallelisieren. In einem RNN muss jedes Element einer Sequenz in sequentieller Reihenfolge verarbeitet werden, was bedeutet, dass es nicht leicht parallelisiert werden kann. - -![Encoder Decoder mit Aufmerksamkeit](../../../../../lessons/5-NLP/18-Transformers/images/EncDecAttention.gif) - -> Abbildung aus [Google's Blog](https://research.googleblog.com/2016/09/a-neural-network-for-machine.html) - -Die Einführung von Aufmerksamkeitsmechanismen in Kombination mit dieser Einschränkung führte zur Schaffung der heutigen State-of-the-Art-Transformer-Modelle, die wir kennen und verwenden, wie BERT und Open-GPT3. - -## Transformermodelle - -Eine der Hauptideen hinter Transformern ist es, die sequentielle Natur von RNNs zu vermeiden und ein Modell zu schaffen, das während des Trainings parallelisierbar ist. Dies wird erreicht, indem zwei Ideen implementiert werden: - -* Positionskodierung -* Verwendung des Selbstaufmerksamkeitsmechanismus zur Erfassung von Mustern anstelle von RNNs (oder CNNs) (deshalb heißt das Papier, das Transformer einführt, *[Attention is all you need](https://arxiv.org/abs/1706.03762)*) - -### Positionskodierung/Einbettung - -Die Idee der Positionskodierung ist die folgende. -1. Bei der Verwendung von RNNs wird die relative Position der Tokens durch die Anzahl der Schritte dargestellt und muss daher nicht explizit dargestellt werden. -2. Sobald wir jedoch zu Aufmerksamkeit wechseln, müssen wir die relativen Positionen der Tokens innerhalb einer Sequenz kennen. -3. Um die Positionskodierung zu erhalten, erweitern wir unsere Sequenz von Tokens um eine Sequenz von Tokenpositionen in der Sequenz (d.h. eine Sequenz von Zahlen 0,1, ...). -4. Wir mischen dann die Tokenposition mit einem Token-Einbettungsvektor. Um die Position (ganzzahlig) in einen Vektor zu transformieren, können wir verschiedene Ansätze verwenden: - -* Trainierbare Einbettung, ähnlich wie bei der Token-Einbettung. Dies ist der Ansatz, den wir hier betrachten. Wir wenden Einbettungsschichten auf sowohl die Tokens als auch ihre Positionen an, was zu Einbettungsvektoren der gleichen Dimensionen führt, die wir dann zusammenaddieren. -* Feste Positionskodierungsfunktion, wie im ursprünglichen Papier vorgeschlagen. - - - -> Bild des Autors - -Das Ergebnis, das wir mit der Positions-Einbettung erhalten, bettet sowohl das ursprüngliche Token als auch seine Position innerhalb einer Sequenz ein. - -### Multi-Head Selbstaufmerksamkeit - -Als Nächstes müssen wir einige Muster innerhalb unserer Sequenz erfassen. Dazu verwenden Transformer einen **Selbstaufmerksamkeits**mechanismus, der im Wesentlichen Aufmerksamkeit auf die gleiche Sequenz anwendet, die als Eingabe und Ausgabe dient. Die Anwendung von Selbstaufmerksamkeit ermöglicht es uns, den **Kontext** innerhalb des Satzes zu berücksichtigen und zu sehen, welche Wörter miteinander in Beziehung stehen. Zum Beispiel ermöglicht es uns zu sehen, welche Wörter durch Kofe referenziert werden, wie *es*, und auch den Kontext zu berücksichtigen: - -![](../../../../../translated_images/de/CoreferenceResolution.861924d6d384a7d6.webp) - -> Bild aus dem [Google Blog](https://research.googleblog.com/2017/08/transformer-novel-neural-network.html) - -In Transformern verwenden wir **Multi-Head Attention**, um dem Netzwerk die Fähigkeit zu geben, mehrere verschiedene Arten von Abhängigkeiten zu erfassen, z.B. langfristige vs. kurzfristige Wortbeziehungen, Kofeferenz vs. etwas anderes usw. - -[TensorFlow Notebook](../../../../../lessons/5-NLP/18-Transformers/TransformersTF.ipynb) enthält weitere Details zur Implementierung von Transformerschichten. - -### Encoder-Decoder-Aufmerksamkeit - -In Transformern wird Aufmerksamkeit an zwei Stellen verwendet: - -* Um Muster innerhalb des Eingabetextes mit Selbstaufmerksamkeit zu erfassen -* Um die Sequenzübersetzung durchzuführen - es ist die Aufmerksamkeitschicht zwischen Encoder und Decoder. - -Die Encoder-Decoder-Aufmerksamkeit ähnelt sehr dem Aufmerksamkeitsmechanismus, der in RNNs verwendet wird, wie zu Beginn dieses Abschnitts beschrieben. Dieses animierte Diagramm erklärt die Rolle der Encoder-Decoder-Aufmerksamkeit. - -![Animiertes GIF, das zeigt, wie die Bewertungen in Transformermodellen durchgeführt werden.](../../../../../lessons/5-NLP/18-Transformers/images/transformer-animated-explanation.gif) - -Da jede Eingabeposition unabhängig auf jede Ausgabeposition abgebildet wird, können Transformer besser parallelisieren als RNNs, was viel größere und ausdrucksstärkere Sprachmodelle ermöglicht. Jeder Aufmerksamkeitskopf kann verwendet werden, um verschiedene Beziehungen zwischen Wörtern zu lernen, die nachgelagerte Aufgaben der natürlichen Sprachverarbeitung verbessern. - -## BERT - -**BERT** (Bidirektionale Encoder-Darstellungen von Transformern) ist ein sehr großes, mehrschichtiges Transformernetzwerk mit 12 Schichten für *BERT-base* und 24 für *BERT-large*. Das Modell wird zunächst auf einem großen Korpus von Textdaten (Wikipedia + Bücher) mit unüberwachtem Training (Vorhersage maskierter Wörter in einem Satz) vortrainiert. Während des Vortrainings absorbiert das Modell erhebliche Niveaus des Sprachverständnisses, die dann mit anderen Datensätzen durch Feinabstimmung genutzt werden können. Dieser Prozess wird als **Transferlernen** bezeichnet. - -![Bild von http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/de/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp) - -> Bild [Quelle](http://jalammar.github.io/illustrated-bert/) - -## ✍️ Übungen: Transformer - -Setzen Sie Ihr Lernen in den folgenden Notebooks fort: - -* [Transformers in PyTorch](../../../../../lessons/5-NLP/18-Transformers/TransformersPyTorch.ipynb) -* [Transformers in TensorFlow](../../../../../lessons/5-NLP/18-Transformers/TransformersTF.ipynb) - -## Fazit - -In dieser Lektion haben Sie über Transformer und Aufmerksamkeitsmechanismen gelernt, alles wesentliche Werkzeuge im NLP-Toolkit. Es gibt viele Variationen von Transformer-Architekturen, darunter BERT, DistilBERT, BigBird, OpenGPT3 und mehr, die feinabgestimmt werden können. Das [HuggingFace-Paket](https://github.com/huggingface/) bietet ein Repository zum Trainieren vieler dieser Architekturen sowohl mit PyTorch als auch mit TensorFlow. - -## 🚀 Herausforderung - -## [Nachlesequiz](https://ff-quizzes.netlify.app/en/ai/quiz/36) - -## Überprüfung & Selbststudium - -* [Blogbeitrag](https://mchromiak.github.io/articles/2017/Sep/12/Transformer-Attention-is-all-you-need/), der das klassische [Attention is all you need](https://arxiv.org/abs/1706.03762) Papier über Transformer erklärt. -* [Eine Reihe von Blogbeiträgen](https://towardsdatascience.com/transformers-explained-visually-part-1-overview-of-functionality-95a6dd460452) über Transformer, die die Architektur im Detail erklären. - -## [Aufgabe](assignment.md) - -**Haftungsausschluss**: -Dieses Dokument wurde mit maschinellen KI-Übersetzungsdiensten übersetzt. Obwohl wir uns um Genauigkeit bemühen, bitten wir zu beachten, dass automatisierte Übersetzungen Fehler oder Ungenauigkeiten enthalten können. Das Originaldokument in seiner ursprünglichen Sprache sollte als die maßgebliche Quelle betrachtet werden. Für wichtige Informationen wird eine professionelle menschliche Übersetzung empfohlen. Wir übernehmen keine Haftung für Missverständnisse oder Fehlinterpretationen, die aus der Verwendung dieser Übersetzung entstehen. \ No newline at end of file diff --git a/translations/de/lessons/5-NLP/18-Transformers/assignment.md b/translations/de/lessons/5-NLP/18-Transformers/assignment.md index d524d5b7..eca2087c 100644 --- a/translations/de/lessons/5-NLP/18-Transformers/assignment.md +++ b/translations/de/lessons/5-NLP/18-Transformers/assignment.md @@ -1,12 +1,3 @@ - # Aufgabe: Transformers Experimentiere mit Transformers auf HuggingFace! Probiere einige der Skripte aus, die sie bereitstellen, um mit den verschiedenen Modellen auf ihrer Website zu arbeiten: https://huggingface.co/docs/transformers/run_scripts. Teste eines ihrer Datensätze, importiere dann einen eigenen aus diesem Lehrplan oder von Kaggle und schau, ob du interessante Texte generieren kannst. Erstelle ein Notebook mit deinen Ergebnissen. diff --git a/translations/de/lessons/5-NLP/19-NER/README.md b/translations/de/lessons/5-NLP/19-NER/README.md index afa4006b..673e63df 100644 --- a/translations/de/lessons/5-NLP/19-NER/README.md +++ b/translations/de/lessons/5-NLP/19-NER/README.md @@ -1,12 +1,3 @@ - # Named Entity Recognition Bis jetzt haben wir uns hauptsächlich auf eine NLP-Aufgabe konzentriert – die Klassifikation. Es gibt jedoch auch andere NLP-Aufgaben, die mit neuronalen Netzwerken gelöst werden können. Eine dieser Aufgaben ist **[Named Entity Recognition](https://wikipedia.org/wiki/Named-entity_recognition)** (NER), die sich mit der Erkennung spezifischer Entitäten im Text befasst, wie Orte, Personennamen, Zeitintervalle, chemische Formeln und vieles mehr. @@ -17,7 +8,7 @@ Bis jetzt haben wir uns hauptsächlich auf eine NLP-Aufgabe konzentriert – die Angenommen, Sie möchten einen natürlichen Sprach-Chatbot entwickeln, ähnlich wie Amazon Alexa oder Google Assistant. Intelligente Chatbots funktionieren, indem sie *verstehen*, was der Benutzer möchte, indem sie eine Textklassifikation auf den Eingabesatz anwenden. Das Ergebnis dieser Klassifikation ist die sogenannte **Intent**, die bestimmt, was der Chatbot tun soll. -Bot NER +Bot NER > Bild vom Autor diff --git a/translations/de/lessons/5-NLP/19-NER/lab/README.md b/translations/de/lessons/5-NLP/19-NER/lab/README.md index 5a986a01..bf9b335d 100644 --- a/translations/de/lessons/5-NLP/19-NER/lab/README.md +++ b/translations/de/lessons/5-NLP/19-NER/lab/README.md @@ -1,12 +1,3 @@ - # NER Laboraufgabe aus dem [AI for Beginners Curriculum](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/de/lessons/5-NLP/20-LangModels/README.md b/translations/de/lessons/5-NLP/20-LangModels/README.md index 086c1012..d6778c15 100644 --- a/translations/de/lessons/5-NLP/20-LangModels/README.md +++ b/translations/de/lessons/5-NLP/20-LangModels/README.md @@ -1,12 +1,3 @@ - # Vorgefertigte große Sprachmodelle In all unseren bisherigen Aufgaben haben wir ein neuronales Netzwerk darauf trainiert, eine bestimmte Aufgabe mithilfe eines beschrifteten Datensatzes auszuführen. Bei großen Transformermodellen wie BERT verwenden wir Sprachmodellierung in selbstüberwachter Weise, um ein Sprachmodell zu erstellen, das anschließend durch weitere domänenspezifische Trainings für spezifische Downstream-Aufgaben spezialisiert wird. Es wurde jedoch gezeigt, dass große Sprachmodelle viele Aufgaben auch ohne jegliches domänenspezifisches Training lösen können. Eine Familie von Modellen, die dazu in der Lage ist, wird als **GPT** bezeichnet: Generative Pre-Trained Transformer. diff --git a/translations/de/lessons/5-NLP/20-LangModels/READMELargeLang.md b/translations/de/lessons/5-NLP/20-LangModels/READMELargeLang.md deleted file mode 100644 index b7bcb1aa..00000000 --- a/translations/de/lessons/5-NLP/20-LangModels/READMELargeLang.md +++ /dev/null @@ -1,56 +0,0 @@ -# Vorgebildete große Sprachmodelle - -In all unseren vorherigen Aufgaben haben wir ein neuronales Netzwerk trainiert, um eine bestimmte Aufgabe mit einem beschrifteten Datensatz auszuführen. Bei großen Transformermodellen wie BERT nutzen wir Sprachmodellierung in selbstüberwachter Weise, um ein Sprachmodell zu erstellen, das dann für spezifische nachgelagerte Aufgaben mit weiterem domänenspezifischem Training spezialisiert wird. Es wurde jedoch gezeigt, dass große Sprachmodelle viele Aufgaben auch ohne jegliches domänenspezifisches Training lösen können. Eine Familie von Modellen, die dazu in der Lage ist, wird als **GPT** bezeichnet: Generative Pre-Trained Transformer. - -## [Vorlesungsquiz](https://ff-quizzes.netlify.app/en/ai/quiz/39) - -## Textgenerierung und Perplexität - -Die Idee, dass ein neuronales Netzwerk in der Lage ist, allgemeine Aufgaben ohne nachgelagertes Training auszuführen, wird im Papier [Language Models are Unsupervised Multitask Learners](https://cdn.openai.com/better-language-models/language_models_are_unsupervised_multitask_learners.pdf) präsentiert. Die Hauptidee ist, dass viele andere Aufgaben mit **Textgenerierung** modelliert werden können, da das Verständnis von Text im Wesentlichen bedeutet, in der Lage zu sein, ihn zu produzieren. Da das Modell auf einer riesigen Menge an Text trainiert wurde, die menschliches Wissen umfasst, wird es auch über eine Vielzahl von Themen informiert. - -> Das Verständnis und die Fähigkeit, Text zu produzieren, bedeutet auch, etwas über die Welt um uns herum zu wissen. Menschen lernen auch in großem Maße durch Lesen, und das GPT-Netzwerk ist in dieser Hinsicht ähnlich. - -Textgenerierungsnetzwerke funktionieren, indem sie die Wahrscheinlichkeit des nächsten Wortes $$P(w_N)$$ vorhersagen. Die bedingungslose Wahrscheinlichkeit des nächsten Wortes entspricht jedoch der Häufigkeit dieses Wortes im Textkorpus. GPT kann uns die **bedingte Wahrscheinlichkeit** des nächsten Wortes geben, basierend auf den vorherigen: $$P(w_N | w_{n-1}, ..., w_0)$$ - -> Sie können mehr über Wahrscheinlichkeiten in unserem [Data Science for Beginners Curriculum](https://github.com/microsoft/Data-Science-For-Beginners/tree/main/1-Introduction/04-stats-and-probability) lesen. - -Die Qualität des sprachgenerierenden Modells kann mit **Perplexität** definiert werden. Es ist eine intrinsische Metrik, die es uns ermöglicht, die Modellqualität ohne einen aufgaben-spezifischen Datensatz zu messen. Sie basiert auf dem Konzept der *Wahrscheinlichkeit eines Satzes* – das Modell weist einem Satz, der wahrscheinlich echt ist (d.h. das Modell ist nicht **verwirrt** von ihm), eine hohe Wahrscheinlichkeit zu, und einer niedrigen Wahrscheinlichkeit für Sätze, die weniger Sinn machen (z.B. *Kann es was tun?*). Wenn wir unserem Modell Sätze aus einem echten Textkorpus geben, erwarten wir, dass sie eine hohe Wahrscheinlichkeit und eine niedrige **Perplexität** haben. Mathematisch wird es als normalisierte inverse Wahrscheinlichkeit des Testdatensatzes definiert: -$$ -\mathrm{Perplexity}(W) = \sqrt[N]{1\over P(W_1,...,W_N)} -$$ - -**Sie können mit der Textgenerierung den [GPT-gestützten Texteditor von Hugging Face](https://transformer.huggingface.co/doc/gpt2-large) ausprobieren.** In diesem Editor beginnen Sie, Ihren Text zu schreiben, und durch Drücken von **[TAB]** erhalten Sie mehrere Vervollständigungsoptionen. Wenn diese zu kurz sind oder Sie nicht zufrieden sind, drücken Sie erneut [TAB], und Sie erhalten weitere Optionen, einschließlich längerer Textabschnitte. - -## GPT ist eine Familie - -GPT ist kein einzelnes Modell, sondern eine Sammlung von Modellen, die von [OpenAI](https://openai.com) entwickelt und trainiert wurden. - -Unter den GPT-Modellen haben wir: - -| [GPT-2](https://huggingface.co/docs/transformers/model_doc/gpt2#openai-gpt2) | [GPT 3](https://openai.com/research/language-models-are-few-shot-learners) | [GPT-4](https://openai.com/gpt-4) | -| -- | -- | -- | -| Sprachmodell mit bis zu 1,5 Milliarden Parametern. | Sprachmodell mit bis zu 175 Milliarden Parametern | 100T Parameter und akzeptiert sowohl Bild- als auch Texteingaben und gibt Text aus. | - - -Die Modelle GPT-3 und GPT-4 sind [als kognitiver Dienst von Microsoft Azure](https://azure.microsoft.com/en-us/services/cognitive-services/openai-service/#overview?WT.mc_id=academic-77998-cacaste) sowie als [OpenAI API](https://openai.com/api/) verfügbar. - -## Prompt-Engineering - -Da GPT auf riesigen Datenmengen trainiert wurde, um Sprache und Code zu verstehen, liefern sie Ausgaben als Antwort auf Eingaben (Prompts). Prompts sind die Eingaben oder Abfragen an GPT, bei denen man den Modellen Anweisungen zu den Aufgaben gibt, die sie als Nächstes ausführen sollen. Um ein gewünschtes Ergebnis zu erzielen, benötigen Sie den effektivsten Prompt, der die Auswahl der richtigen Wörter, Formate, Phrasen oder sogar Symbole umfasst. Dieser Ansatz wird als [Prompt Engineering](https://learn.microsoft.com/en-us/shows/ai-show/the-basics-of-prompt-engineering-with-azure-openai-service?WT.mc_id=academic-77998-bethanycheum) bezeichnet. - -[Diese Dokumentation](https://learn.microsoft.com/en-us/semantic-kernel/prompt-engineering/?WT.mc_id=academic-77998-bethanycheum) bietet Ihnen weitere Informationen zum Prompt-Engineering. - -## ✍️ Beispielnotizbuch: [Spielen mit OpenAI-GPT](../../../../../lessons/5-NLP/20-LangModels/GPT-PyTorch.ipynb) - -Setzen Sie Ihr Lernen in den folgenden Notizbüchern fort: - -* [Textgenerierung mit OpenAI-GPT und Hugging Face Transformers](../../../../../lessons/5-NLP/20-LangModels/GPT-PyTorch.ipynb) - -## Fazit - -Neue allgemeine vortrainierte Sprachmodelle modellieren nicht nur die Sprachstruktur, sondern enthalten auch eine riesige Menge an natürlicher Sprache. Daher können sie effektiv eingesetzt werden, um einige NLP-Aufgaben in Zero-Shot- oder Few-Shot-Einstellungen zu lösen. - -## [Nachlesungsquiz](https://ff-quizzes.netlify.app/en/ai/quiz/40) - -**Haftungsausschluss**: -Dieses Dokument wurde mit maschinellen KI-Übersetzungsdiensten übersetzt. Obwohl wir uns um Genauigkeit bemühen, beachten Sie bitte, dass automatisierte Übersetzungen Fehler oder Ungenauigkeiten enthalten können. Das Originaldokument in seiner ursprünglichen Sprache sollte als die maßgebliche Quelle betrachtet werden. Für kritische Informationen wird eine professionelle menschliche Übersetzung empfohlen. Wir übernehmen keine Haftung für Missverständnisse oder Fehlinterpretationen, die aus der Verwendung dieser Übersetzung entstehen. \ No newline at end of file diff --git a/translations/de/lessons/5-NLP/README.md b/translations/de/lessons/5-NLP/README.md index 5a04f4c6..9beb5e21 100644 --- a/translations/de/lessons/5-NLP/README.md +++ b/translations/de/lessons/5-NLP/README.md @@ -1,12 +1,3 @@ - # Verarbeitung natürlicher Sprache ![Zusammenfassung der NLP-Aufgaben in einer Skizze](../../../../lessons/sketchnotes/ai-nlp.png) diff --git a/translations/de/lessons/6-Other/21-GeneticAlgorithms/README.md b/translations/de/lessons/6-Other/21-GeneticAlgorithms/README.md index 4e2bedf1..fe714bc8 100644 --- a/translations/de/lessons/6-Other/21-GeneticAlgorithms/README.md +++ b/translations/de/lessons/6-Other/21-GeneticAlgorithms/README.md @@ -1,12 +1,3 @@ - # Genetische Algorithmen ## [Quiz vor der Vorlesung](https://ff-quizzes.netlify.app/en/ai/quiz/41) diff --git a/translations/de/lessons/6-Other/22-DeepRL/README.md b/translations/de/lessons/6-Other/22-DeepRL/README.md index a2b88206..aa63ff10 100644 --- a/translations/de/lessons/6-Other/22-DeepRL/README.md +++ b/translations/de/lessons/6-Other/22-DeepRL/README.md @@ -1,12 +1,3 @@ - # Deep Reinforcement Learning Reinforcement Learning (RL) wird als eines der grundlegenden Paradigmen des maschinellen Lernens angesehen, neben überwachten und unüberwachten Lernen. Während wir beim überwachten Lernen auf Datensätze mit bekannten Ergebnissen angewiesen sind, basiert RL auf dem Prinzip des **Lernens durch Handeln**. Zum Beispiel: Wenn wir ein Computerspiel zum ersten Mal sehen, beginnen wir zu spielen, auch ohne die Regeln zu kennen, und verbessern unsere Fähigkeiten allein durch das Spielen und Anpassen unseres Verhaltens. @@ -34,7 +25,7 @@ Ihr habt wahrscheinlich alle moderne Balanciergeräte wie den *Segway* oder *Gyr Eine vereinfachte Version des Balancierens ist als **CartPole**-Problem bekannt. In der CartPole-Welt haben wir einen horizontalen Schlitten, der sich nach links oder rechts bewegen kann, und das Ziel ist es, eine vertikale Stange oben auf dem Schlitten zu balancieren, während er sich bewegt. -a cartpole +a cartpole Um diese Umgebung zu erstellen und zu nutzen, benötigen wir ein paar Zeilen Python-Code: diff --git a/translations/de/lessons/6-Other/22-DeepRL/lab/README.md b/translations/de/lessons/6-Other/22-DeepRL/lab/README.md index 064548de..7c16026b 100644 --- a/translations/de/lessons/6-Other/22-DeepRL/lab/README.md +++ b/translations/de/lessons/6-Other/22-DeepRL/lab/README.md @@ -1,12 +1,3 @@ - ## Die Umgebung Die Mountain Car-Umgebung besteht aus einem Auto, das in einem Tal gefangen ist. Dein Ziel ist es, aus dem Tal herauszuspringen und die Fahne zu erreichen. Die Aktionen, die du ausführen kannst, sind: nach links beschleunigen, nach rechts beschleunigen oder nichts tun. Du kannst die Position des Autos entlang der x-Achse und die Geschwindigkeit beobachten. diff --git a/translations/de/lessons/6-Other/23-MultiagentSystems/README.md b/translations/de/lessons/6-Other/23-MultiagentSystems/README.md index dae57470..07b4cc80 100644 --- a/translations/de/lessons/6-Other/23-MultiagentSystems/README.md +++ b/translations/de/lessons/6-Other/23-MultiagentSystems/README.md @@ -1,12 +1,3 @@ - # Multi-Agenten-Systeme Eine mögliche Methode, Intelligenz zu erreichen, ist der sogenannte **emergente** (oder **synergetische**) Ansatz. Dieser basiert auf der Tatsache, dass das kombinierte Verhalten vieler relativ einfacher Agenten zu einem insgesamt komplexeren (oder intelligenteren) Verhalten des Systems als Ganzes führen kann. Theoretisch basiert dies auf den Prinzipien der [kollektiven Intelligenz](https://en.wikipedia.org/wiki/Collective_intelligence), des [Emergentismus](https://en.wikipedia.org/wiki/Global_brain) und der [evolutionären Kybernetik](https://en.wikipedia.org/wiki/Global_brain), die besagen, dass höherstufige Systeme einen Mehrwert erhalten, wenn sie richtig aus niedrigeren Systemen kombiniert werden (das sogenannte *Prinzip des Metasystemübergangs*). @@ -60,7 +51,7 @@ Sie können [NetLogo herunterladen](https://ccl.northwestern.edu/netlogo/downloa Ein großartiges Merkmal von NetLogo ist, dass es eine Bibliothek mit funktionierenden Modellen enthält, die Sie ausprobieren können. Gehen Sie zu **Datei → Modellbibliothek**, und Sie haben viele Kategorien von Modellen zur Auswahl. -NetLogo Modellbibliothek +NetLogo Modellbibliothek > Ein Screenshot der Modellbibliothek von Dmitry Soshnikov diff --git a/translations/de/lessons/6-Other/23-MultiagentSystems/assignment.md b/translations/de/lessons/6-Other/23-MultiagentSystems/assignment.md index 6e5c7f37..a58d90ed 100644 --- a/translations/de/lessons/6-Other/23-MultiagentSystems/assignment.md +++ b/translations/de/lessons/6-Other/23-MultiagentSystems/assignment.md @@ -1,12 +1,3 @@ - # NetLogo Aufgabe Nehmen Sie eines der Modelle aus der NetLogo-Bibliothek und verwenden Sie es, um eine reale Situation so genau wie möglich zu simulieren. Ein gutes Beispiel wäre, das Virus-Modell im Ordner Alternative Visualisierungen anzupassen, um zu zeigen, wie es verwendet werden kann, um die Ausbreitung von COVID-19 zu modellieren. Können Sie ein Modell erstellen, das die Ausbreitung eines echten Virus nachahmt? diff --git a/translations/de/lessons/7-Ethics/README.md b/translations/de/lessons/7-Ethics/README.md index d19a0bbc..1a01e4f0 100644 --- a/translations/de/lessons/7-Ethics/README.md +++ b/translations/de/lessons/7-Ethics/README.md @@ -1,12 +1,3 @@ - # Ethische und verantwortungsvolle KI Sie haben diesen Kurs fast abgeschlossen, und ich hoffe, dass Sie mittlerweile klar erkennen, dass KI auf einer Reihe formaler mathematischer Methoden basiert, die es uns ermöglichen, Beziehungen in Daten zu finden und Modelle zu trainieren, um bestimmte Aspekte menschlichen Verhaltens nachzubilden. Zu diesem Zeitpunkt in der Geschichte betrachten wir KI als ein sehr mächtiges Werkzeug, um Muster aus Daten zu extrahieren und diese Muster anzuwenden, um neue Probleme zu lösen. diff --git a/translations/de/lessons/README.md b/translations/de/lessons/README.md index b2aae705..2c7c88e4 100644 --- a/translations/de/lessons/README.md +++ b/translations/de/lessons/README.md @@ -1,12 +1,3 @@ - # Überblick ![Überblick in einer Kritzelei](../../../lessons/sketchnotes/ai-overview.png) diff --git a/translations/de/lessons/X-Extras/X1-MultiModal/README.md b/translations/de/lessons/X-Extras/X1-MultiModal/README.md index f0416fe7..07d468e9 100644 --- a/translations/de/lessons/X-Extras/X1-MultiModal/README.md +++ b/translations/de/lessons/X-Extras/X1-MultiModal/README.md @@ -1,12 +1,3 @@ - # Multi-Modale Netzwerke Nach dem Erfolg von Transformer-Modellen bei der Lösung von NLP-Aufgaben wurden ähnliche Architekturen auch auf Aufgaben der Computer Vision angewendet. Es gibt ein wachsendes Interesse daran, Modelle zu entwickeln, die Vision- und Sprachfähigkeiten *kombinieren*. Ein solcher Versuch wurde von OpenAI unternommen und nennt sich CLIP und DALL.E. diff --git a/translations/de/lessons/sketchnotes/LICENSE.md b/translations/de/lessons/sketchnotes/LICENSE.md index 0a737fa5..fc6edb71 100644 --- a/translations/de/lessons/sketchnotes/LICENSE.md +++ b/translations/de/lessons/sketchnotes/LICENSE.md @@ -1,12 +1,3 @@ - Attribution-Weitergabe unter gleichen Bedingungen 4.0 International ======================================================================= diff --git a/translations/de/lessons/sketchnotes/README.md b/translations/de/lessons/sketchnotes/README.md index 615cdfde..89efb81b 100644 --- a/translations/de/lessons/sketchnotes/README.md +++ b/translations/de/lessons/sketchnotes/README.md @@ -1,12 +1,3 @@ - Alle Sketchnotes des Lehrplans können hier heruntergeladen werden. 🎨 Erstellt von: Tomomi Imura (Twitter: [@girlie_mac](https://twitter.com/girlie_mac), GitHub: [girliemac](https://github.com/girliemac)) diff --git a/translations/de/troubleshoot.md b/translations/de/troubleshoot.md index 182221dc..6a8f68b3 100644 --- a/translations/de/troubleshoot.md +++ b/translations/de/troubleshoot.md @@ -1,12 +1,3 @@ - # AI-For-Beginners Fehlerbehebungshandbuch Dieses Handbuch hilft Ihnen, häufige Probleme zu lösen, die beim Verwenden oder Beitragen zum [AI-For-Beginners](https://github.com/microsoft/AI-For-Beginners)-Repository auftreten können. Jedes Problem enthält Hintergrundinformationen, Symptome, Erklärungen und Schritt-für-Schritt-Lösungen. diff --git a/translations/es/.co-op-translator.json b/translations/es/.co-op-translator.json new file mode 100644 index 00000000..592edc5d --- /dev/null +++ b/translations/es/.co-op-translator.json @@ -0,0 +1,398 @@ +{ + "AGENTS.md": { + "original_hash": "6b11a37115944252ab3ed04e358d830d", + "translation_date": "2025-10-03T09:11:21+00:00", + "source_file": "AGENTS.md", + "language_code": "es" + }, + "README.md": { + "original_hash": "1984fc89dd304a8a33ab5584691a99aa", + "translation_date": "2026-01-30T01:01:10+00:00", + "source_file": "README.md", + "language_code": "es" + }, + "SECURITY.md": { + "original_hash": "a583f49d359c7ebba61433e4dfcd05a9", + "translation_date": "2025-08-24T09:11:05+00:00", + "source_file": "SECURITY.md", + "language_code": "es" + }, + "etc/CODE_OF_CONDUCT.md": { + "original_hash": "c06b12caf3c901eb3156e3dd5b0aea56", + "translation_date": "2025-08-24T09:26:49+00:00", + "source_file": "etc/CODE_OF_CONDUCT.md", + "language_code": "es" + }, + "etc/CONTRIBUTING.md": { + "original_hash": "847a587aa1b83f4d00858183ff3ed18a", + "translation_date": "2025-08-24T09:27:05+00:00", + "source_file": "etc/CONTRIBUTING.md", + "language_code": "es" + }, + "etc/Mindmap.md": { + "original_hash": "f2f88dbd2debd38e26149b27b1fd272d", + "translation_date": "2025-08-24T09:27:13+00:00", + "source_file": "etc/Mindmap.md", + "language_code": "es" + }, + "etc/SUPPORT.md": { + "original_hash": "fdfc08baee91e402938a2b1f94fe0949", + "translation_date": "2025-08-24T09:26:44+00:00", + "source_file": "etc/SUPPORT.md", + "language_code": "es" + }, + "etc/TRANSLATIONS.md": { + "original_hash": "62b3e3ad5182edb905eec649a87eeeb4", + "translation_date": "2025-08-24T09:26:55+00:00", + "source_file": "etc/TRANSLATIONS.md", + "language_code": "es" + }, + "etc/quiz-app/README.md": { + "original_hash": "d699cf8509f74baa5b0b838de5cf0662", + "translation_date": "2025-08-24T09:27:33+00:00", + "source_file": "etc/quiz-app/README.md", + "language_code": "es" + }, + "examples/README.md": { + "original_hash": "0d1babfdcbeb46525f2db3fbaaa54cd7", + "translation_date": "2025-10-03T11:26:17+00:00", + "source_file": "examples/README.md", + "language_code": "es" + }, + "lessons/0-course-setup/for-teachers.md": { + "original_hash": "a094ef9927883de1cfcee51dbd143381", + "translation_date": "2025-08-24T09:26:23+00:00", + "source_file": "lessons/0-course-setup/for-teachers.md", + "language_code": "es" + }, + "lessons/0-course-setup/how-to-run.md": { + "original_hash": "a4717bd9103b9f6cd84d534b83534689", + "translation_date": "2026-01-15T10:47:10+00:00", + "source_file": "lessons/0-course-setup/how-to-run.md", + "language_code": "es" + }, + "lessons/0-course-setup/setup.md": { + "original_hash": "7b4e5b8956915870d0a0ed3cc5890042", + "translation_date": "2025-12-12T19:07:08+00:00", + "source_file": "lessons/0-course-setup/setup.md", + "language_code": "es" + }, + "lessons/1-Intro/README.md": { + "original_hash": "f57e8aa46141fd220b16ffed8f11aec7", + "translation_date": "2025-11-18T20:32:29+00:00", + "source_file": "lessons/1-Intro/README.md", + "language_code": "es" + }, + "lessons/1-Intro/assignment.md": { + "original_hash": "a334df77a82aaaf2a29c77065d3e481e", + "translation_date": "2025-11-18T20:33:22+00:00", + "source_file": "lessons/1-Intro/assignment.md", + "language_code": "es" + }, + "lessons/2-Symbolic/README.md": { + "original_hash": "f9f06b266b8b2bfc6b8792ff2bb1bea4", + "translation_date": "2026-01-15T10:47:28+00:00", + "source_file": "lessons/2-Symbolic/README.md", + "language_code": "es" + }, + "lessons/2-Symbolic/assignment.md": { + "original_hash": "a057a8604f3976c3e309884453f1fad0", + "translation_date": "2025-08-24T09:26:03+00:00", + "source_file": "lessons/2-Symbolic/assignment.md", + "language_code": "es" + }, + "lessons/3-NeuralNetworks/03-Perceptron/README.md": { + "original_hash": "c34cbba802058b6fa267e1a294d4e510", + "translation_date": "2025-09-23T12:11:58+00:00", + "source_file": "lessons/3-NeuralNetworks/03-Perceptron/README.md", + "language_code": "es" + }, + "lessons/3-NeuralNetworks/03-Perceptron/lab/README.md": { + "original_hash": "ba5d1eb353d20d3e7181066b3c424b99", + "translation_date": "2025-08-31T15:38:58+00:00", + "source_file": "lessons/3-NeuralNetworks/03-Perceptron/lab/README.md", + "language_code": "es" + }, + "lessons/3-NeuralNetworks/04-OwnFramework/README.md": { + "original_hash": "789d6c3fb6fc7948a470b33078a5983a", + "translation_date": "2025-09-23T12:11:37+00:00", + "source_file": "lessons/3-NeuralNetworks/04-OwnFramework/README.md", + "language_code": "es" + }, + "lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md": { + "original_hash": "48fdd704d483e19bc3d7464074c9fcbe", + "translation_date": "2025-08-24T09:22:24+00:00", + "source_file": "lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md", + "language_code": "es" + }, + "lessons/3-NeuralNetworks/05-Frameworks/README.md": { + "original_hash": "ddd216f558a255260a9374008002c971", + "translation_date": "2025-09-23T12:12:17+00:00", + "source_file": "lessons/3-NeuralNetworks/05-Frameworks/README.md", + "language_code": "es" + }, + "lessons/3-NeuralNetworks/05-Frameworks/lab/README.md": { + "original_hash": "e452d897efb9a89700f41021834cf6e5", + "translation_date": "2025-08-24T09:22:51+00:00", + "source_file": "lessons/3-NeuralNetworks/05-Frameworks/lab/README.md", + "language_code": "es" + }, + "lessons/3-NeuralNetworks/README.md": { + "original_hash": "f862a99d88088163df12270e2f2ad6c3", + "translation_date": "2025-10-03T12:41:21+00:00", + "source_file": "lessons/3-NeuralNetworks/README.md", + "language_code": "es" + }, + "lessons/4-ComputerVision/06-IntroCV/README.md": { + "original_hash": "feeca98225cb420afc89415f24f63d92", + "translation_date": "2025-09-23T12:07:54+00:00", + "source_file": "lessons/4-ComputerVision/06-IntroCV/README.md", + "language_code": "es" + }, + "lessons/4-ComputerVision/06-IntroCV/lab/README.md": { + "original_hash": "3d53d6409f80970f7281a45dee35328a", + "translation_date": "2025-08-24T09:18:55+00:00", + "source_file": "lessons/4-ComputerVision/06-IntroCV/lab/README.md", + "language_code": "es" + }, + "lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md": { + "original_hash": "53faab85adfcebd8c10bcd71dc2fa557", + "translation_date": "2025-09-23T12:07:15+00:00", + "source_file": "lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md", + "language_code": "es" + }, + "lessons/4-ComputerVision/07-ConvNets/README.md": { + "original_hash": "a560d5b845962cf33dc102266e409568", + "translation_date": "2025-09-23T12:06:51+00:00", + "source_file": "lessons/4-ComputerVision/07-ConvNets/README.md", + "language_code": "es" + }, + "lessons/4-ComputerVision/07-ConvNets/lab/README.md": { + "original_hash": "b70fcf7fcee862990f848c679090943f", + "translation_date": "2025-10-03T14:51:55+00:00", + "source_file": "lessons/4-ComputerVision/07-ConvNets/lab/README.md", + "language_code": "es" + }, + "lessons/4-ComputerVision/08-TransferLearning/README.md": { + "original_hash": "178c0b5ee5395733eb18aec51e71a0a9", + "translation_date": "2025-09-23T12:07:31+00:00", + "source_file": "lessons/4-ComputerVision/08-TransferLearning/README.md", + "language_code": "es" + }, + "lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md": { + "original_hash": "ae074cd940fc2f4dc24fc07b66ccbd99", + "translation_date": "2025-08-24T09:19:25+00:00", + "source_file": "lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md", + "language_code": "es" + }, + "lessons/4-ComputerVision/08-TransferLearning/lab/README.md": { + "original_hash": "7765935c35fcee69b9fe2d0cfd6963e2", + "translation_date": "2025-08-24T09:19:55+00:00", + "source_file": "lessons/4-ComputerVision/08-TransferLearning/lab/README.md", + "language_code": "es" + }, + "lessons/4-ComputerVision/09-Autoencoders/README.md": { + "original_hash": "1b8d9e1b3a6f1daa864b1ff3dfc3076d", + "translation_date": "2025-09-23T12:08:58+00:00", + "source_file": "lessons/4-ComputerVision/09-Autoencoders/README.md", + "language_code": "es" + }, + "lessons/4-ComputerVision/10-GANs/README.md": { + "original_hash": "0ff65b4da07b23697235de2beb2a3c25", + "translation_date": "2025-09-23T12:09:39+00:00", + "source_file": "lessons/4-ComputerVision/10-GANs/README.md", + "language_code": "es" + }, + "lessons/4-ComputerVision/11-ObjectDetection/README.md": { + "original_hash": "d76a7eda28de5210c8b1ba50a6216c69", + "translation_date": "2025-09-23T12:08:25+00:00", + "source_file": "lessons/4-ComputerVision/11-ObjectDetection/README.md", + "language_code": "es" + }, + "lessons/4-ComputerVision/11-ObjectDetection/lab/README.md": { + "original_hash": "ad568d55ae65c856fe929fc2b278510a", + "translation_date": "2025-08-24T09:17:32+00:00", + "source_file": "lessons/4-ComputerVision/11-ObjectDetection/lab/README.md", + "language_code": "es" + }, + "lessons/4-ComputerVision/12-Segmentation/README.md": { + "original_hash": "6568aaae7e0e4afed4b5d74b5b223700", + "translation_date": "2025-09-23T12:09:21+00:00", + "source_file": "lessons/4-ComputerVision/12-Segmentation/README.md", + "language_code": "es" + }, + "lessons/4-ComputerVision/12-Segmentation/lab/README.md": { + "original_hash": "365f0decfe0f47b460bbde8227c5009d", + "translation_date": "2025-08-24T09:16:22+00:00", + "source_file": "lessons/4-ComputerVision/12-Segmentation/lab/README.md", + "language_code": "es" + }, + "lessons/4-ComputerVision/README.md": { + "original_hash": "58a52f000089c1d8906a4daa4ab1169b", + "translation_date": "2025-08-24T09:15:28+00:00", + "source_file": "lessons/4-ComputerVision/README.md", + "language_code": "es" + }, + "lessons/5-NLP/13-TextRep/README.md": { + "original_hash": "dbd3f73e4139f030ecb2e20387d70fee", + "translation_date": "2025-09-23T12:14:55+00:00", + "source_file": "lessons/5-NLP/13-TextRep/README.md", + "language_code": "es" + }, + "lessons/5-NLP/13-TextRep/assignment.md": { + "original_hash": "cdc1f2e631f055f3473b36d18e4760b3", + "translation_date": "2025-08-24T09:13:27+00:00", + "source_file": "lessons/5-NLP/13-TextRep/assignment.md", + "language_code": "es" + }, + "lessons/5-NLP/14-Embeddings/README.md": { + "original_hash": "b708c9b85b833864c73c6281f1e6b96e", + "translation_date": "2025-09-23T12:14:36+00:00", + "source_file": "lessons/5-NLP/14-Embeddings/README.md", + "language_code": "es" + }, + "lessons/5-NLP/14-Embeddings/assignment.md": { + "original_hash": "bc690ecf68b38d311cc9e12f3144a28c", + "translation_date": "2025-08-24T09:12:38+00:00", + "source_file": "lessons/5-NLP/14-Embeddings/assignment.md", + "language_code": "es" + }, + "lessons/5-NLP/15-LanguageModeling/README.md": { + "original_hash": "7ba20f54a5bfcd6521018cdfb17c7c57", + "translation_date": "2025-09-23T12:13:04+00:00", + "source_file": "lessons/5-NLP/15-LanguageModeling/README.md", + "language_code": "es" + }, + "lessons/5-NLP/15-LanguageModeling/lab/README.md": { + "original_hash": "5130f01fdc5ebb83032b23d489027aac", + "translation_date": "2025-08-24T09:13:47+00:00", + "source_file": "lessons/5-NLP/15-LanguageModeling/lab/README.md", + "language_code": "es" + }, + "lessons/5-NLP/16-RNN/README.md": { + "original_hash": "e2273cc150380a5e191903cea858f021", + "translation_date": "2025-09-23T12:14:08+00:00", + "source_file": "lessons/5-NLP/16-RNN/README.md", + "language_code": "es" + }, + "lessons/5-NLP/16-RNN/assignment.md": { + "original_hash": "47f7d3c6a5373543e051e4d1140ce898", + "translation_date": "2025-08-24T09:12:15+00:00", + "source_file": "lessons/5-NLP/16-RNN/assignment.md", + "language_code": "es" + }, + "lessons/5-NLP/17-GenerativeNetworks/README.md": { + "original_hash": "51be6057374d01d70e07dd5ec88ebc0d", + "translation_date": "2025-09-23T12:12:41+00:00", + "source_file": "lessons/5-NLP/17-GenerativeNetworks/README.md", + "language_code": "es" + }, + "lessons/5-NLP/17-GenerativeNetworks/lab/README.md": { + "original_hash": "439e12796197a90e7623d4c9c057b9c2", + "translation_date": "2025-08-24T09:13:03+00:00", + "source_file": "lessons/5-NLP/17-GenerativeNetworks/lab/README.md", + "language_code": "es" + }, + "lessons/5-NLP/18-Transformers/README.md": { + "original_hash": "f335dfcb4a993920504c387973a36957", + "translation_date": "2025-09-23T12:13:13+00:00", + "source_file": "lessons/5-NLP/18-Transformers/README.md", + "language_code": "es" + }, + "lessons/5-NLP/18-Transformers/assignment.md": { + "original_hash": "177f3ea3995d725e6f9f5c66af16edcd", + "translation_date": "2025-08-24T09:13:56+00:00", + "source_file": "lessons/5-NLP/18-Transformers/assignment.md", + "language_code": "es" + }, + "lessons/5-NLP/19-NER/README.md": { + "original_hash": "6522312ff835796ca34136a9462fafb2", + "translation_date": "2025-09-23T12:13:52+00:00", + "source_file": "lessons/5-NLP/19-NER/README.md", + "language_code": "es" + }, + "lessons/5-NLP/19-NER/lab/README.md": { + "original_hash": "032bda5068f543d6c1fcb30c34231461", + "translation_date": "2025-08-24T09:14:21+00:00", + "source_file": "lessons/5-NLP/19-NER/lab/README.md", + "language_code": "es" + }, + "lessons/5-NLP/20-LangModels/README.md": { + "original_hash": "97836d30a6bec736f8e3b4411c572bc2", + "translation_date": "2025-09-23T12:13:35+00:00", + "source_file": "lessons/5-NLP/20-LangModels/README.md", + "language_code": "es" + }, + "lessons/5-NLP/README.md": { + "original_hash": "8ef02a9318257ea140ed3ed74442096d", + "translation_date": "2025-08-24T09:11:33+00:00", + "source_file": "lessons/5-NLP/README.md", + "language_code": "es" + }, + "lessons/6-Other/21-GeneticAlgorithms/README.md": { + "original_hash": "6bbd632dfe6c62e5f66bb51fd78c174a", + "translation_date": "2025-09-23T12:05:32+00:00", + "source_file": "lessons/6-Other/21-GeneticAlgorithms/README.md", + "language_code": "es" + }, + "lessons/6-Other/22-DeepRL/README.md": { + "original_hash": "04395657fc01648f8f70484d0e55ab67", + "translation_date": "2025-09-23T12:06:28+00:00", + "source_file": "lessons/6-Other/22-DeepRL/README.md", + "language_code": "es" + }, + "lessons/6-Other/22-DeepRL/lab/README.md": { + "original_hash": "7bd8dc72040e98e35e7225e34058cd4e", + "translation_date": "2025-08-24T09:21:19+00:00", + "source_file": "lessons/6-Other/22-DeepRL/lab/README.md", + "language_code": "es" + }, + "lessons/6-Other/23-MultiagentSystems/README.md": { + "original_hash": "38a1185ae3d54b180378bbd71ae3ef16", + "translation_date": "2025-09-23T12:05:46+00:00", + "source_file": "lessons/6-Other/23-MultiagentSystems/README.md", + "language_code": "es" + }, + "lessons/6-Other/23-MultiagentSystems/assignment.md": { + "original_hash": "cf654ca60c7f86c8dad28596fb42994b", + "translation_date": "2025-08-24T09:20:48+00:00", + "source_file": "lessons/6-Other/23-MultiagentSystems/assignment.md", + "language_code": "es" + }, + "lessons/7-Ethics/README.md": { + "original_hash": "437c988596e751072e41a5aad3fcc5d9", + "translation_date": "2025-08-24T09:11:21+00:00", + "source_file": "lessons/7-Ethics/README.md", + "language_code": "es" + }, + "lessons/README.md": { + "original_hash": "5fef1a0b22498d7188959e2a2cb08af7", + "translation_date": "2025-08-24T09:11:15+00:00", + "source_file": "lessons/README.md", + "language_code": "es" + }, + "lessons/X-Extras/X1-MultiModal/README.md": { + "original_hash": "9c592c26aca16ca085d268c732284187", + "translation_date": "2025-08-24T09:21:24+00:00", + "source_file": "lessons/X-Extras/X1-MultiModal/README.md", + "language_code": "es" + }, + "lessons/sketchnotes/LICENSE.md": { + "original_hash": "45ab63a2cd8f5faef6c9b150618837a4", + "translation_date": "2025-08-24T09:24:20+00:00", + "source_file": "lessons/sketchnotes/LICENSE.md", + "language_code": "es" + }, + "lessons/sketchnotes/README.md": { + "original_hash": "050b8bddebafba55b129414e6ab096ab", + "translation_date": "2025-08-24T09:23:26+00:00", + "source_file": "lessons/sketchnotes/README.md", + "language_code": "es" + }, + "troubleshoot.md": { + "original_hash": "8d9c5a4a7c7798d699672a22cb7fea86", + "translation_date": "2025-10-03T09:34:07+00:00", + "source_file": "troubleshoot.md", + "language_code": "es" + } +} \ No newline at end of file diff --git a/translations/es/AGENTS.md b/translations/es/AGENTS.md index fec8d9e8..df30eb59 100644 --- a/translations/es/AGENTS.md +++ b/translations/es/AGENTS.md @@ -1,12 +1,3 @@ - # AGENTS.md ## Resumen del Proyecto diff --git a/translations/es/README.md b/translations/es/README.md index d36ee0e1..5f0e5e11 100644 --- a/translations/es/README.md +++ b/translations/es/README.md @@ -1,12 +1,3 @@ - [![GitHub license](https://img.shields.io/github/license/microsoft/AI-For-Beginners.svg)](https://github.com/microsoft/AI-For-Beginners/blob/main/LICENSE) [![GitHub contributors](https://img.shields.io/github/contributors/microsoft/AI-For-Beginners.svg)](https://GitHub.com/microsoft/AI-For-Beginners/graphs/contributors/) [![GitHub issues](https://img.shields.io/github/issues/microsoft/AI-For-Beginners.svg)](https://GitHub.com/microsoft/AI-For-Beginners/issues/) @@ -21,212 +12,214 @@ CO_OP_TRANSLATOR_METADATA: [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/nTYy5BXMWG)](https://discord.gg/nTYy5BXMWG) -# Inteligencia Artificial para Principiantes - Un Currículo +# Inteligencia Artificial para Principiantes - Un Currículum -|![Sketchnote by @girlie_mac https://twitter.com/girlie_mac](../../../../translated_images/es/ai-overview.0857791951d19500.webp)| +|![Sketchnote by @girlie_mac https://twitter.com/girlie_mac](../../translated_images/es/ai-overview.0857791951d19500.webp)| |:---:| | AI For Beginners - _Sketchnote por [@girlie_mac](https://twitter.com/girlie_mac)_ | -¡Explora el mundo de la **Inteligencia Artificial** (IA) con nuestro currículo de 12 semanas y 24 lecciones! Incluye lecciones prácticas, cuestionarios y laboratorios. El currículo es accesible para principiantes y cubre herramientas como TensorFlow y PyTorch, así como ética en IA. +¡Explora el mundo de la **Inteligencia Artificial** (IA) con nuestro currículum de 12 semanas y 24 lecciones! Incluye lecciones prácticas, cuestionarios y laboratorios. El currículum es apto para principiantes y cubre herramientas como TensorFlow y PyTorch, así como la ética en la IA. + ### 🌐 Soporte Multilingüe -#### Soportado a través de GitHub Action (Automatizado y Siempre Actualizado) +#### Soportado mediante GitHub Action (Automatizado y Siempre Actualizado) -[Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh/README.md) | [Chinese (Traditional, Hong Kong)](../hk/README.md) | [Chinese (Traditional, Macau)](../mo/README.md) | [Chinese (Traditional, Taiwan)](../tw/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../br/README.md) | [Portuguese (Portugal)](../pt/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](./README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) +[Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](./README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) > **¿Prefieres Clonar Localmente?** -> Este repositorio incluye más de 50 traducciones que aumentan significativamente el tamaño de descarga. Para clonar sin traducciones, usa sparse checkout: +> Este repositorio incluye más de 50 traducciones de idiomas que aumentan significativamente el tamaño de descarga. Para clonar sin traducciones, usa el sparse checkout: > ```bash > git clone --filter=blob:none --sparse https://github.com/microsoft/AI-For-Beginners.git > cd AI-For-Beginners > git sparse-checkout set --no-cone '/*' '!translations' '!translated_images' > ``` -> Esto te da todo lo que necesitas para completar el curso con una descarga mucho más rápida. +> Esto te da todo lo necesario para completar el curso con una descarga mucho más rápida. -**Si deseas que se agreguen más idiomas de traducción, están listados [aquí](https://github.com/Azure/co-op-translator/blob/main/getting_started/supported-languages.md)** +**Si deseas que se soporten idiomas adicionales, están listados [aquí](https://github.com/Azure/co-op-translator/blob/main/getting_started/supported-languages.md)** ## Únete a la Comunidad [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/nTYy5BXMWG)](https://discord.gg/nTYy5BXMWG) -## Lo que aprenderás +## Qué aprenderás -**[Mapa mental del curso](http://soshnikov.com/courses/ai-for-beginners/mindmap.html)** +**[Mapa Mental del Curso](http://soshnikov.com/courses/ai-for-beginners/mindmap.html)** -En este currículo aprenderás: +En este currículum aprenderás: -* Diferentes enfoques de la Inteligencia Artificial, incluyendo el "buen y viejo" enfoque simbólico con **Representación del Conocimiento** y razonamiento ([GOFAI](https://en.wikipedia.org/wiki/Symbolic_artificial_intelligence)). -* **Redes Neuronales** y **Aprendizaje Profundo**, que están en el núcleo de la IA moderna. Ilustraremos los conceptos detrás de estos temas importantes usando código en dos de los frameworks más populares - [TensorFlow](http://Tensorflow.org) y [PyTorch](http://pytorch.org). -* **Arquitecturas Neuronales** para trabajar con imágenes y texto. Cubriremos modelos recientes pero puede que falte un poco del estado del arte. -* Enfoques menos populares de IA, como **Algoritmos Genéticos** y **Sistemas Multiagente**. +* Diferentes enfoques de la Inteligencia Artificial, incluyendo el enfoque simbólico "clásico" con **Representación del Conocimiento** y razonamiento ([GOFAI](https://en.wikipedia.org/wiki/Symbolic_artificial_intelligence)). +* **Redes Neuronales** y **Aprendizaje Profundo**, que están en el núcleo de la IA moderna. Ilustraremos los conceptos detrás de estos temas importantes utilizando código en dos de los frameworks más populares - [TensorFlow](http://Tensorflow.org) y [PyTorch](http://pytorch.org). +* **Arquitecturas Neuronales** para trabajar con imágenes y texto. Cubriremos modelos recientes aunque puede que falte algo del estado del arte. +* Enfoques de IA menos populares, como **Algoritmos Genéticos** y **Sistemas Multi-Agente**. -Lo que no cubriremos en este currículo: +Lo que no cubriremos en este currículum: -> [Encuentra todos los recursos adicionales para este curso en nuestra colección de Microsoft Learn](https://learn.microsoft.com/en-us/collections/7w28iy2xrqzdj0?WT.mc_id=academic-77998-bethanycheum) +> [Encuentra todos los recursos adicionales para este curso en nuestra colección Microsoft Learn](https://learn.microsoft.com/en-us/collections/7w28iy2xrqzdj0?WT.mc_id=academic-77998-bethanycheum) -* Casos de negocio usando **IA en Negocios**. Considera tomar la ruta de aprendizaje [Introducción a la IA para usuarios de negocio](https://docs.microsoft.com/learn/paths/introduction-ai-for-business-users/?WT.mc_id=academic-77998-bethanycheum) en Microsoft Learn, o [AI Business School](https://www.microsoft.com/ai/ai-business-school/?WT.mc_id=academic-77998-bethanycheum), desarrollado en cooperación con [INSEAD](https://www.insead.edu/). -* **Aprendizaje Automático Clásico**, que está bien descrito en nuestro [Currículo de Aprendizaje automático para principiantes](http://github.com/Microsoft/ML-for-Beginners). -* Aplicaciones prácticas de IA construidas usando **[Servicios Cognitivos](https://azure.microsoft.com/services/cognitive-services/?WT.mc_id=academic-77998-bethanycheum)**. Para esto, recomendamos comenzar con los módulos de Microsoft Learn sobre [visión](https://docs.microsoft.com/learn/paths/create-computer-vision-solutions-azure-cognitive-services/?WT.mc_id=academic-77998-bethanycheum), [procesamiento de lenguaje natural](https://docs.microsoft.com/learn/paths/explore-natural-language-processing/?WT.mc_id=academic-77998-bethanycheum), **[IA Generativa con Azure OpenAI Service](https://learn.microsoft.com/en-us/training/paths/develop-ai-solutions-azure-openai/?WT.mc_id=academic-77998-bethanycheum)**, y otros. -* **Frameworks específicos en la nube** para ML, como [Azure Machine Learning](https://azure.microsoft.com/services/machine-learning/?WT.mc_id=academic-77998-bethanycheum), [Microsoft Fabric](https://learn.microsoft.com/en-us/training/paths/get-started-fabric/?WT.mc_id=academic-77998-bethanycheum), o [Azure Databricks](https://docs.microsoft.com/learn/paths/data-engineer-azure-databricks?WT.mc_id=academic-77998-bethanycheum). Considera usar las rutas de aprendizaje [Construye y opera soluciones de aprendizaje automático con Azure Machine Learning](https://docs.microsoft.com/learn/paths/build-ai-solutions-with-azure-ml-service/?WT.mc_id=academic-77998-bethanycheum) y [Construye y opera soluciones de aprendizaje automático con Azure Databricks](https://docs.microsoft.com/learn/paths/build-operate-machine-learning-solutions-azure-databricks/?WT.mc_id=academic-77998-bethanycheum). -* **IA conversacional** y **Chat Bots**. Hay una ruta de aprendizaje separada [Crea soluciones de IA conversacional](https://docs.microsoft.com/learn/paths/create-conversational-ai-solutions/?WT.mc_id=academic-77998-bethanycheum), y también puedes consultar [esta publicación del blog](https://soshnikov.com/azure/hello-bot-conversational-ai-on-microsoft-platform/) para más detalles. -* **Matemáticas profundas** detrás del aprendizaje profundo. Para esto, recomendamos [Deep Learning](https://www.amazon.com/Deep-Learning-Adaptive-Computation-Machine/dp/0262035618) por Ian Goodfellow, Yoshua Bengio y Aaron Courville, que también está disponible en línea en [https://www.deeplearningbook.org/](https://www.deeplearningbook.org/). +* Casos de negocio del uso de la **IA en los Negocios**. Considera tomar la ruta de aprendizaje [Introducción a la IA para usuarios de negocio](https://docs.microsoft.com/learn/paths/introduction-ai-for-business-users/?WT.mc_id=academic-77998-bethanycheum) en Microsoft Learn, o la [AI Business School](https://www.microsoft.com/ai/ai-business-school/?WT.mc_id=academic-77998-bethanycheum), desarrollada en cooperación con [INSEAD](https://www.insead.edu/). +* **Aprendizaje Automático Clásico**, que está bien descrito en nuestro [Currículum para principiantes en Machine Learning](http://github.com/Microsoft/ML-for-Beginners). +* Aplicaciones prácticas de IA construidas usando **[Cognitive Services](https://azure.microsoft.com/services/cognitive-services/?WT.mc_id=academic-77998-bethanycheum)**. Para esto, recomendamos comenzar con módulos de Microsoft Learn para [visión](https://docs.microsoft.com/learn/paths/create-computer-vision-solutions-azure-cognitive-services/?WT.mc_id=academic-77998-bethanycheum), [procesamiento de lenguaje natural](https://docs.microsoft.com/learn/paths/explore-natural-language-processing/?WT.mc_id=academic-77998-bethanycheum), **[IA Generativa con Azure OpenAI Service](https://learn.microsoft.com/en-us/training/paths/develop-ai-solutions-azure-openai/?WT.mc_id=academic-77998-bethanycheum)** y otros. +* **Frameworks en la Nube para ML** específicos, como [Azure Machine Learning](https://azure.microsoft.com/services/machine-learning/?WT.mc_id=academic-77998-bethanycheum), [Microsoft Fabric](https://learn.microsoft.com/en-us/training/paths/get-started-fabric/?WT.mc_id=academic-77998-bethanycheum), o [Azure Databricks](https://docs.microsoft.com/learn/paths/data-engineer-azure-databricks?WT.mc_id=academic-77998-bethanycheum). Considera usar las rutas de aprendizaje [Construye y opera soluciones de machine learning con Azure Machine Learning](https://docs.microsoft.com/learn/paths/build-ai-solutions-with-azure-ml-service/?WT.mc_id=academic-77998-bethanycheum) y [Construye y opera soluciones de Machine Learning con Azure Databricks](https://docs.microsoft.com/learn/paths/build-operate-machine-learning-solutions-azure-databricks/?WT.mc_id=academic-77998-bethanycheum). +* **IA Conversacional** y **Chat Bots**. Existe una ruta de aprendizaje separada [Crea soluciones de IA conversacional](https://docs.microsoft.com/learn/paths/create-conversational-ai-solutions/?WT.mc_id=academic-77998-bethanycheum), y también puedes consultar [esta publicación en el blog](https://soshnikov.com/azure/hello-bot-conversational-ai-on-microsoft-platform/) para más detalles. +* **Matemáticas Profundas** detrás del aprendizaje profundo. Para esto, recomendamos [Deep Learning](https://www.amazon.com/Deep-Learning-Adaptive-Computation-Machine/dp/0262035618) por Ian Goodfellow, Yoshua Bengio y Aaron Courville, que también está disponible en línea en [https://www.deeplearningbook.org/](https://www.deeplearningbook.org/). -Para una introducción suave a temas de _IA en la nube_ puedes considerar tomar la Ruta de Aprendizaje [Comienza con inteligencia artificial en Azure](https://docs.microsoft.com/learn/paths/get-started-with-artificial-intelligence-on-azure/?WT.mc_id=academic-77998-bethanycheum). +Para una introducción suave a temas de _IA en la Nube_ puedes considerar tomar la ruta de aprendizaje [Comienza con inteligencia artificial en Azure](https://docs.microsoft.com/learn/paths/get-started-with-artificial-intelligence-on-azure/?WT.mc_id=academic-77998-bethanycheum). # Contenido -| | Enlace a la Lección | PyTorch/Keras/TensorFlow | Lab | +| | Enlace de la Lección | PyTorch/Keras/TensorFlow | Laboratorio | | :-: | :------------------------------------------------------------------------------------------------------------------------------------------: | :---------------------------------------------------------------------------------------------: | ------------------------------------------------------------------------------ | -| 0 | [Configuración del Curso](./lessons/0-course-setup/setup.md) | [Configura tu Entorno de Desarrollo](./lessons/0-course-setup/how-to-run.md) | | +| 0 | [Configuración del Curso](./lessons/0-course-setup/setup.md) | [Configura tu entorno de desarrollo](./lessons/0-course-setup/how-to-run.md) | | | I | [**Introducción a la IA**](./lessons/1-Intro/README.md) | | | | 01 | [Introducción e Historia de la IA](./lessons/1-Intro/README.md) | - | - | | II | **IA Simbólica** | | 02 | [Representación del Conocimiento y Sistemas Expertos](./lessons/2-Symbolic/README.md) | [Sistemas Expertos](./lessons/2-Symbolic/Animals.ipynb) / [Ontología](./lessons/2-Symbolic/FamilyOntology.ipynb) /[Grafo de Conceptos](./lessons/2-Symbolic/MSConceptGraph.ipynb) | | | III | [**Introducción a las Redes Neuronales**](./lessons/3-NeuralNetworks/README.md) ||| | 03 | [Perceptrón](./lessons/3-NeuralNetworks/03-Perceptron/README.md) | [Cuaderno](./lessons/3-NeuralNetworks/03-Perceptron/Perceptron.ipynb) | [Laboratorio](./lessons/3-NeuralNetworks/03-Perceptron/lab/README.md) | -| 04 | [Perceptrón Multicapa y Creando nuestro propio Framework](./lessons/3-NeuralNetworks/04-OwnFramework/README.md) | [Cuaderno](./lessons/3-NeuralNetworks/04-OwnFramework/OwnFramework.ipynb) | [Laboratorio](./lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md) | +| 04 | [Perceptrón Multicapa y Creación de nuestro propio Framework](./lessons/3-NeuralNetworks/04-OwnFramework/README.md) | [Cuaderno](./lessons/3-NeuralNetworks/04-OwnFramework/OwnFramework.ipynb) | [Laboratorio](./lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md) | | 05 | [Introducción a Frameworks (PyTorch/TensorFlow) y Sobreajuste](./lessons/3-NeuralNetworks/05-Frameworks/README.md) | [PyTorch](./lessons/3-NeuralNetworks/05-Frameworks/IntroPyTorch.ipynb) / [Keras](./lessons/3-NeuralNetworks/05-Frameworks/IntroKeras.ipynb) / [TensorFlow](./lessons/3-NeuralNetworks/05-Frameworks/IntroKerasTF.ipynb) | [Laboratorio](./lessons/3-NeuralNetworks/05-Frameworks/lab/README.md) | | IV | [**Visión por Computadora**](./lessons/4-ComputerVision/README.md) | [PyTorch](https://docs.microsoft.com/learn/modules/intro-computer-vision-pytorch/?WT.mc_id=academic-77998-cacaste) / [TensorFlow](https://docs.microsoft.com/learn/modules/intro-computer-vision-TensorFlow/?WT.mc_id=academic-77998-cacaste)| [Explora Visión por Computadora en Microsoft Azure](https://learn.microsoft.com/en-us/collections/7w28iy2xrqzdj0?WT.mc_id=academic-77998-bethanycheum) | | 06 | [Introducción a Visión por Computadora. OpenCV](./lessons/4-ComputerVision/06-IntroCV/README.md) | [Cuaderno](./lessons/4-ComputerVision/06-IntroCV/OpenCV.ipynb) | [Laboratorio](./lessons/4-ComputerVision/06-IntroCV/lab/README.md) | | 07 | [Redes Neuronales Convolucionales](./lessons/4-ComputerVision/07-ConvNets/README.md) & [Arquitecturas CNN](./lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md) | [PyTorch](./lessons/4-ComputerVision/07-ConvNets/ConvNetsPyTorch.ipynb) /[TensorFlow](./lessons/4-ComputerVision/07-ConvNets/ConvNetsTF.ipynb) | [Laboratorio](./lessons/4-ComputerVision/07-ConvNets/lab/README.md) | -| 08 | [Redes Preentrenadas y Aprendizaje por Transferencia](./lessons/4-ComputerVision/08-TransferLearning/README.md) y [Trucos de Entrenamiento](./lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md) | [PyTorch](./lessons/4-ComputerVision/08-TransferLearning/TransferLearningPyTorch.ipynb) / [TensorFlow](./lessons/3-NeuralNetworks/05-Frameworks/IntroKerasTF.ipynb) | [Laboratorio](./lessons/4-ComputerVision/08-TransferLearning/lab/README.md) | +| 08 | [Redes Pre-entrenadas y Aprendizaje por Transferencia](./lessons/4-ComputerVision/08-TransferLearning/README.md) y [Trucos de Entrenamiento](./lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md) | [PyTorch](./lessons/4-ComputerVision/08-TransferLearning/TransferLearningPyTorch.ipynb) / [TensorFlow](./lessons/3-NeuralNetworks/05-Frameworks/IntroKerasTF.ipynb) | [Laboratorio](./lessons/4-ComputerVision/08-TransferLearning/lab/README.md) | | 09 | [Autoencoders y VAEs](./lessons/4-ComputerVision/09-Autoencoders/README.md) | [PyTorch](./lessons/4-ComputerVision/09-Autoencoders/AutoEncodersPyTorch.ipynb) / [TensorFlow](./lessons/4-ComputerVision/09-Autoencoders/AutoencodersTF.ipynb) | | | 10 | [Redes Generativas Antagónicas y Transferencia de Estilo Artístico](./lessons/4-ComputerVision/10-GANs/README.md) | [PyTorch](./lessons/4-ComputerVision/10-GANs/GANPyTorch.ipynb) / [TensorFlow](./lessons/4-ComputerVision/10-GANs/GANTF.ipynb) | | | 11 | [Detección de Objetos](./lessons/4-ComputerVision/11-ObjectDetection/README.md) | [TensorFlow](./lessons/4-ComputerVision/11-ObjectDetection/ObjectDetection.ipynb) | [Laboratorio](./lessons/4-ComputerVision/11-ObjectDetection/lab/README.md) | | 12 | [Segmentación Semántica. U-Net](./lessons/4-ComputerVision/12-Segmentation/README.md) | [PyTorch](./lessons/4-ComputerVision/12-Segmentation/SemanticSegmentationPytorch.ipynb) / [TensorFlow](./lessons/4-ComputerVision/12-Segmentation/SemanticSegmentationTF.ipynb) | | -| V | [**Procesamiento de Lenguaje Natural**](./lessons/5-NLP/README.md) | [PyTorch](https://docs.microsoft.com/learn/modules/intro-natural-language-processing-pytorch/?WT.mc_id=academic-77998-cacaste) /[TensorFlow](https://docs.microsoft.com/learn/modules/intro-natural-language-processing-TensorFlow/?WT.mc_id=academic-77998-cacaste) | [Explora el Procesamiento de Lenguaje Natural en Microsoft Azure](https://learn.microsoft.com/en-us/collections/7w28iy2xrqzdj0?WT.mc_id=academic-77998-bethanycheum)| +| V | [**Procesamiento de Lenguaje Natural**](./lessons/5-NLP/README.md) | [PyTorch](https://docs.microsoft.com/learn/modules/intro-natural-language-processing-pytorch/?WT.mc_id=academic-77998-cacaste) /[TensorFlow](https://docs.microsoft.com/learn/modules/intro-natural-language-processing-TensorFlow/?WT.mc_id=academic-77998-cacaste) | [Explora Procesamiento de Lenguaje Natural en Microsoft Azure](https://learn.microsoft.com/en-us/collections/7w28iy2xrqzdj0?WT.mc_id=academic-77998-bethanycheum)| | 13 | [Representación de Texto. Bow/TF-IDF](./lessons/5-NLP/13-TextRep/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/13-TextRep/TextRepresentationPyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/13-TextRep/TextRepresentationTF.ipynb) | | | 14 | [Embeddings semánticos de palabras. Word2Vec y GloVe](./lessons/5-NLP/14-Embeddings/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/14-Embeddings/EmbeddingsPyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/14-Embeddings/EmbeddingsTF.ipynb) | | -| 15 | [Modelado de Lenguaje. Entrenando tus propios embeddings](./lessons/5-NLP/15-LanguageModeling/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/15-LanguageModeling/CBoW-PyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/15-LanguageModeling/CBoW-TF.ipynb) | [Laboratorio](./lessons/5-NLP/15-LanguageModeling/lab/README.md) | +| 15 | [Modelado del Lenguaje. Entrenando tus propios embeddings](./lessons/5-NLP/15-LanguageModeling/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/15-LanguageModeling/CBoW-PyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/15-LanguageModeling/CBoW-TF.ipynb) | [Laboratorio](./lessons/5-NLP/15-LanguageModeling/lab/README.md) | | 16 | [Redes Neuronales Recurrentes](./lessons/5-NLP/16-RNN/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/16-RNN/RNNPyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/16-RNN/RNNTF.ipynb) | | | 17 | [Redes Recurrentes Generativas](./lessons/5-NLP/17-GenerativeNetworks/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/17-GenerativeNetworks/GenerativePyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/17-GenerativeNetworks/GenerativeTF.ipynb) | [Laboratorio](./lessons/5-NLP/17-GenerativeNetworks/lab/README.md) | | 18 | [Transformers. BERT.](./lessons/5-NLP/18-Transformers/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/18-Transformers/TransformersPyTorch.ipynb) /[TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/18-Transformers/TransformersTF.ipynb) | | | 19 | [Reconocimiento de Entidades Nombradas](./lessons/5-NLP/19-NER/README.md) | [TensorFlow](https://microsoft.github.io/AI-For-Beginners/lessons/5-NLP/19-NER/NER-TF.ipynb) | [Laboratorio](./lessons/5-NLP/19-NER/lab/README.md) | -| 20 | [Modelos de Lenguaje Grandes, Programación de Prompts y Tareas con Pocas Muestras](./lessons/5-NLP/20-LangModels/README.md) | [PyTorch](https://microsoft.github.io/AI-For-Beginners/lessons/5-NLP/20-LangModels/GPT-PyTorch.ipynb) | | +| 20 | [Modelos de Lenguaje Grandes, Programación de Prompts y Tareas Few-Shot](./lessons/5-NLP/20-LangModels/README.md) | [PyTorch](https://microsoft.github.io/AI-For-Beginners/lessons/5-NLP/20-LangModels/GPT-PyTorch.ipynb) | | | VI | **Otras Técnicas de IA** || | | 21 | [Algoritmos Genéticos](./lessons/6-Other/21-GeneticAlgorithms/README.md) | [Cuaderno](./lessons/6-Other/21-GeneticAlgorithms/Genetic.ipynb) | | -| 22 | [Aprendizaje Profundo por Refuerzo](./lessons/6-Other/22-DeepRL/README.md) | [PyTorch](./lessons/6-Other/22-DeepRL/CartPole-RL-PyTorch.ipynb) /[TensorFlow](./lessons/6-Other/22-DeepRL/CartPole-RL-TF.ipynb) | [Laboratorio](./lessons/6-Other/22-DeepRL/lab/README.md) | -| 23 | [Sistemas Multiagente](./lessons/6-Other/23-MultiagentSystems/README.md) | | | +| 22 | [Aprendizaje por Refuerzo Profundo](./lessons/6-Other/22-DeepRL/README.md) | [PyTorch](./lessons/6-Other/22-DeepRL/CartPole-RL-PyTorch.ipynb) /[TensorFlow](./lessons/6-Other/22-DeepRL/CartPole-RL-TF.ipynb) | [Laboratorio](./lessons/6-Other/22-DeepRL/lab/README.md) | +| 23 | [Sistemas Multi-Agentes](./lessons/6-Other/23-MultiagentSystems/README.md) | | | | VII | **Ética en IA** | | | | 24 | [Ética en IA e IA Responsable](./lessons/7-Ethics/README.md) | [Microsoft Learn: Principios de IA Responsable](https://docs.microsoft.com/learn/paths/responsible-ai-business-principles/?WT.mc_id=academic-77998-cacaste) | | | IX | **Extras** | | | -| 25 | [Redes Multimodales, CLIP y VQGAN](./lessons/X-Extras/X1-MultiModal/README.md) | [Cuaderno](./lessons/X-Extras/X1-MultiModal/Clip.ipynb) | | +| 25 | [Redes Multi-Modal, CLIP y VQGAN](./lessons/X-Extras/X1-MultiModal/README.md) | [Cuaderno](./lessons/X-Extras/X1-MultiModal/Clip.ipynb) | | ## Cada lección contiene -* Material de prelectura -* Cuadernos Jupyter ejecutables, que a menudo son específicos para el framework (**PyTorch** o **TensorFlow**). El cuaderno ejecutable también contiene mucho material teórico, por lo que para entender el tema necesitas revisar al menos una versión del cuaderno (ya sea PyTorch o TensorFlow). -* **Laboratorios** disponibles para algunos temas, que te brindan la oportunidad de probar aplicar el material aprendido a un problema específico. +* Material de lectura previa +* Cuadernos ejecutables de Jupyter, que a menudo son específicos para el framework (**PyTorch** o **TensorFlow**). El cuaderno ejecutable también contiene mucho material teórico, por lo que para entender el tema necesitas revisar al menos una versión del cuaderno (ya sea PyTorch o TensorFlow). +* **Laboratorios** disponibles para algunos temas, que te dan la oportunidad de intentar aplicar el material que has aprendido a un problema específico. * Algunas secciones contienen enlaces a módulos de [**MS Learn**](https://learn.microsoft.com/en-us/collections/7w28iy2xrqzdj0?WT.mc_id=academic-77998-bethanycheum) que cubren temas relacionados. -## Empezando +## Comenzando -### 🎯 ¿Nuevo en IA? ¡Comienza Aquí! +### 🎯 ¿Nuevo en IA? ¡Empieza Aquí! -Si eres completamente nuevo en IA y quieres ejemplos rápidos y prácticos, echa un vistazo a nuestros [**Ejemplos para Principiantes**](./examples/README.md). Estos incluyen: +Si eres completamente nuevo en IA y quieres ejemplos prácticos rápidos, ¡consulta nuestros [**Ejemplos para Principiantes**](./examples/README.md)! Estos incluyen: - 🌟 **Hola Mundo IA** - Tu primer programa de IA (reconocimiento de patrones) - 🧠 **Red Neuronal Simple** - Construye una red neuronal desde cero -- 🖼️ **Clasificador de Imágenes** - Clasifica imágenes con comentarios detallados -- 💬 **Sentimiento de Texto** - Analiza texto positivo/negativo -Estos ejemplos están diseñados para ayudarte a entender conceptos de IA antes de sumergirte en el plan de estudios completo. +- 🖼️ **Clasificador de Imágenes** - Clasifica imágenes con comentarios detallados +- 💬 **Sentimiento del Texto** - Analiza textos positivos/negativos -### 📚 Configuración del Plan de Estudios Completo +Estos ejemplos están diseñados para ayudarte a entender conceptos de IA antes de adentrarte en el currículo completo. -- Hemos creado una [lección de configuración](./lessons/0-course-setup/setup.md) para ayudarte a configurar tu entorno de desarrollo. - Para educadores, también hemos creado una [lección de configuración del plan de estudios](./lessons/0-course-setup/for-teachers.md) para ustedes. -- Cómo [Ejecutar el código en VSCode o en un Codespace](./lessons/0-course-setup/how-to-run.md) +### 📚 Configuración del Currículo Completo + +- Hemos creado una [lección de configuración](./lessons/0-course-setup/setup.md) para ayudarte a configurar tu entorno de desarrollo. - ¡Para educadores, también hemos creado una [lección de configuración del currículo](./lessons/0-course-setup/for-teachers.md)! +- Cómo [Ejecutar el código en VSCode o en un Codespace](./lessons/0-course-setup/how-to-run.md) Sigue estos pasos: -Haz un Fork del Repositorio: Haz clic en el botón "Fork" en la esquina superior derecha de esta página. +Bifurca el Repositorio: Haz clic en el botón "Fork" en la esquina superior derecha de esta página. Clona el Repositorio: `git clone https://github.com/microsoft/AI-For-Beginners.git` -No olvides agregar una estrella (🌟) a este repositorio para encontrarlo más fácilmente después. +No olvides poner una estrella (🌟) a este repositorio para encontrarlo más fácilmente después. -## Conoce a otros Estudiantes +## Conoce a otros Aprendices -Únete a nuestro [servidor oficial de Discord de IA](https://aka.ms/genai-discord?WT.mc_id=academic-105485-bethanycheum) para conocer y conectar con otros estudiantes que están tomando este curso y obtener soporte. +Únete a nuestro [servidor oficial de Discord de IA](https://aka.ms/genai-discord?WT.mc_id=academic-105485-bethanycheum) para conocer y hacer networking con otros alumnos que toman este curso y obtener soporte. -Si tienes comentarios sobre el producto o preguntas mientras construyes visita nuestro [Foro de Desarrolladores Azure AI Foundry](https://aka.ms/foundry/forum) +Si tienes retroalimentación sobre el producto o preguntas mientras construyes, visita nuestro [Foro de Desarrolladores de Azure AI Foundry](https://aka.ms/foundry/forum) -## Cuestionarios +## Cuestionarios -> **Una nota sobre los cuestionarios**: Todos los cuestionarios están contenidos en la carpeta Quiz-app en etc\quiz-app, o [En línea Aquí](https://ff-quizzes.netlify.app/). Están enlazados desde dentro de las lecciones, la app de cuestionario puede ejecutarse localmente o desplegarse en Azure; sigue las instrucciones en la carpeta `quiz-app`. Se están localizando gradualmente. +> **Una nota sobre los cuestionarios**: Todos los cuestionarios están contenidos en la carpeta Quiz-app en etc\quiz-app, o [En línea Aquí](https://ff-quizzes.netlify.app/) Están enlazados dentro de las lecciones. La aplicación de cuestionarios puede ejecutarse localmente o desplegarse en Azure; sigue las instrucciones en la carpeta `quiz-app`. Se están localizando gradualmente. ## Se Busca Ayuda -¿Tienes sugerencias o encontraste errores de ortografía o código? Abre un issue o crea un pull request. +¿Tienes sugerencias o has encontrado errores ortográficos o en el código? Abre un issue o crea un pull request. ## Agradecimientos Especiales -* **✍️ Autor Principal:** [Dmitry Soshnikov](http://soshnikov.com), PhD -* **🔥 Editor:** [Jen Looper](https://twitter.com/jenlooper), PhD -* **🎨 Ilustradora de Sketchnotes:** [Tomomi Imura](https://twitter.com/girlie_mac) -* **✅ Creador de Cuestionarios:** [Lateefah Bello](https://github.com/CinnamonXI), [MLSA](https://studentambassadors.microsoft.com/) -* **🙏 Colaboradores Principales:** [Evgenii Pishchik](https://github.com/Pe4enIks) +* **✍️ Autor Principal:** [Dmitry Soshnikov](http://soshnikov.com), PhD +* **🔥 Editor:** [Jen Looper](https://twitter.com/jenlooper), PhD +* **🎨 Ilustradora de Sketchnotes:** [Tomomi Imura](https://twitter.com/girlie_mac) +* **✅ Creadora de Cuestionarios:** [Lateefah Bello](https://github.com/CinnamonXI), [MLSA](https://studentambassadors.microsoft.com/) +* **🙏 Colaboradores Principales:** [Evgenii Pishchik](https://github.com/Pe4enIks) -## Otros Planes de Estudio +## Otros Currículos -Nuestro equipo produce otros planes de estudio. ¡Mira: +¡Nuestro equipo produce otros currículos! Consulta: -### LangChain -[![LangChain4j para Principiantes](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js para Principiantes](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +### LangChain +[![LangChain4j for Beginners](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js for Beginners](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Agentes -[![AZD para Principiantes](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI para Principiantes](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP para Principiantes](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Agentes de IA para Principiantes](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +### Azure / Edge / MCP / Agentes +[![AZD for Beginners](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI for Beginners](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP for Beginners](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agents for Beginners](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Serie de IA Generativa -[![IA Generativa para Principiantes](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![IA Generativa (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![IA Generativa (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![IA Generativa (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) + +### Serie de IA Generativa +[![Generative AI for Beginners](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - -### Aprendizaje Fundamental -[![ML para Principiantes](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Ciencia de Datos para Principiantes](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![IA para Principiantes](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Ciberseguridad para Principiantes](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Desarrollo Web para Principiantes](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT para Principiantes](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![Desarrollo XR para Principiantes](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) + +### Aprendizaje Básico +[![ML for Beginners](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Data Science for Beginners](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![AI for Beginners](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![Cybersecurity for Beginners](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Web Dev for Beginners](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT for Beginners](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![XR Development for Beginners](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Serie Copilot -[![Copilot para Programación en Pareja con IA](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot para C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Aventura Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) + +### Serie Copiloto +[![Copilot for AI Paired Programming](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot for C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) ## Obtener Ayuda -Si te quedas atascado o tienes alguna pregunta sobre cómo construir aplicaciones de IA, únete a otros estudiantes y desarrolladores experimentados en discusiones sobre MCP. Es una comunidad de apoyo donde las preguntas son bienvenidas y el conocimiento se comparte libremente. +Si te quedas atascado o tienes preguntas sobre cómo construir aplicaciones de IA, únete a otros alumnos y desarrolladores experimentados en discusiones sobre MCP. Es una comunidad de apoyo donde las preguntas son bienvenidas y el conocimiento se comparte libremente. [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/nTYy5BXMWG)](https://discord.gg/nTYy5BXMWG) -Si tienes comentarios sobre el producto o errores mientras construyes, visita: +Si tienes retroalimentación sobre el producto o errores mientras construyes, visita: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Aviso legal**: -Este documento ha sido traducido utilizando el servicio de traducción automática [Co-op Translator](https://github.com/Azure/co-op-translator). Aunque nos esforzamos por la precisión, tenga en cuenta que las traducciones automáticas pueden contener errores o inexactitudes. El documento original en su idioma nativo debe considerarse la fuente autorizada. Para información crítica, se recomienda la traducción profesional humana. No nos hacemos responsables por malentendidos o interpretaciones erróneas derivadas del uso de esta traducción. +**Descargo de responsabilidad**: +Este documento ha sido traducido utilizando el servicio de traducción automática [Co-op Translator](https://github.com/Azure/co-op-translator). Aunque nos esforzamos por la precisión, tenga en cuenta que las traducciones automatizadas pueden contener errores o inexactitudes. El documento original en su idioma nativo debe considerarse la fuente autorizada. Para información crítica, se recomienda una traducción profesional realizada por personas. No nos hacemos responsables de malentendidos o interpretaciones erróneas que resulten del uso de esta traducción. \ No newline at end of file diff --git a/translations/es/SECURITY.md b/translations/es/SECURITY.md index c2fd76d0..091ca539 100644 --- a/translations/es/SECURITY.md +++ b/translations/es/SECURITY.md @@ -1,12 +1,3 @@ - ## Seguridad Microsoft se toma muy en serio la seguridad de nuestros productos y servicios de software, lo que incluye todos los repositorios de código fuente gestionados a través de nuestras organizaciones de GitHub, que incluyen [Microsoft](https://github.com/Microsoft), [Azure](https://github.com/Azure), [DotNet](https://github.com/dotnet), [AspNet](https://github.com/aspnet), [Xamarin](https://github.com/xamarin) y [nuestras organizaciones de GitHub](https://opensource.microsoft.com/). diff --git a/translations/es/etc/CODE_OF_CONDUCT.md b/translations/es/etc/CODE_OF_CONDUCT.md index 600b966b..a5945897 100644 --- a/translations/es/etc/CODE_OF_CONDUCT.md +++ b/translations/es/etc/CODE_OF_CONDUCT.md @@ -1,12 +1,3 @@ - # Código de Conducta de Código Abierto de Microsoft Este proyecto ha adoptado el [Código de Conducta de Código Abierto de Microsoft](https://opensource.microsoft.com/codeofconduct/). diff --git a/translations/es/etc/CONTRIBUTING.md b/translations/es/etc/CONTRIBUTING.md index 1177c41d..7a295d7b 100644 --- a/translations/es/etc/CONTRIBUTING.md +++ b/translations/es/etc/CONTRIBUTING.md @@ -1,12 +1,3 @@ - # Contribuir Este proyecto da la bienvenida a contribuciones y sugerencias. La mayoría de las contribuciones requieren que aceptes un Acuerdo de Licencia de Contribuidor (CLA) declarando que tienes el derecho de, y efectivamente otorgas, los derechos para usar tu contribución. Para más detalles, visita https://cla.microsoft.com. diff --git a/translations/es/etc/Mindmap.md b/translations/es/etc/Mindmap.md index b2f0e8e4..b31a585b 100644 --- a/translations/es/etc/Mindmap.md +++ b/translations/es/etc/Mindmap.md @@ -1,12 +1,3 @@ - # IA ## [Introducción a la IA](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/1-Intro/README.md) diff --git a/translations/es/etc/SUPPORT.md b/translations/es/etc/SUPPORT.md index 930db449..64bf5e95 100644 --- a/translations/es/etc/SUPPORT.md +++ b/translations/es/etc/SUPPORT.md @@ -1,12 +1,3 @@ - # Soporte ## Cómo reportar problemas y obtener ayuda diff --git a/translations/es/etc/TRANSLATIONS.md b/translations/es/etc/TRANSLATIONS.md index af0f3526..751b72cc 100644 --- a/translations/es/etc/TRANSLATIONS.md +++ b/translations/es/etc/TRANSLATIONS.md @@ -1,12 +1,3 @@ - # Contribuye traduciendo lecciones ¡Damos la bienvenida a las traducciones de las lecciones en este plan de estudios! diff --git a/translations/es/etc/quiz-app/README.md b/translations/es/etc/quiz-app/README.md index 48de0502..82f2c6ac 100644 --- a/translations/es/etc/quiz-app/README.md +++ b/translations/es/etc/quiz-app/README.md @@ -1,12 +1,3 @@ - # Cuestionarios Estos cuestionarios son los cuestionarios previos y posteriores a las lecciones del currículo de IA en https://aka.ms/ai-beginners diff --git a/translations/es/examples/README.md b/translations/es/examples/README.md index 9b25a699..213e8dac 100644 --- a/translations/es/examples/README.md +++ b/translations/es/examples/README.md @@ -1,12 +1,3 @@ - # Ejemplos de IA para Principiantes ¡Bienvenido! Este directorio contiene ejemplos simples e independientes para ayudarte a comenzar con la inteligencia artificial y el aprendizaje automático. Cada ejemplo está diseñado para ser accesible para principiantes, con comentarios detallados y explicaciones paso a paso. diff --git a/translations/es/lessons/0-course-setup/for-teachers.md b/translations/es/lessons/0-course-setup/for-teachers.md index 16f969ae..838d40ea 100644 --- a/translations/es/lessons/0-course-setup/for-teachers.md +++ b/translations/es/lessons/0-course-setup/for-teachers.md @@ -1,12 +1,3 @@ - # Para Educadores ¿Te gustaría usar este plan de estudios en tu aula? ¡Siéntete libre de hacerlo! diff --git a/translations/es/lessons/0-course-setup/how-to-run.md b/translations/es/lessons/0-course-setup/how-to-run.md index 892ebfb2..5aa4c0c0 100644 --- a/translations/es/lessons/0-course-setup/how-to-run.md +++ b/translations/es/lessons/0-course-setup/how-to-run.md @@ -1,12 +1,3 @@ - # Cómo Ejecutar el Código Este plan de estudios contiene muchos ejemplos ejecutables y laboratorios que querrás ejecutar. Para hacerlo, necesitas la capacidad de ejecutar código Python en los Jupyter Notebooks proporcionados como parte de este plan de estudios. Tienes varias opciones para ejecutar el código: diff --git a/translations/es/lessons/0-course-setup/setup.md b/translations/es/lessons/0-course-setup/setup.md index d7f14508..1c92501f 100644 --- a/translations/es/lessons/0-course-setup/setup.md +++ b/translations/es/lessons/0-course-setup/setup.md @@ -1,12 +1,3 @@ - # Comenzando con este Currículo ## ¿Eres estudiante? diff --git a/translations/es/lessons/1-Intro/README.md b/translations/es/lessons/1-Intro/README.md index 6b6d82eb..78e65be6 100644 --- a/translations/es/lessons/1-Intro/README.md +++ b/translations/es/lessons/1-Intro/README.md @@ -1,12 +1,3 @@ - # Introducción a la IA ![Resumen del contenido de Introducción a la IA en un dibujo](../../../../translated_images/es/ai-intro.bf28d1ac4235881c.webp) diff --git a/translations/es/lessons/1-Intro/assignment.md b/translations/es/lessons/1-Intro/assignment.md index 45cb5e08..c51c59e1 100644 --- a/translations/es/lessons/1-Intro/assignment.md +++ b/translations/es/lessons/1-Intro/assignment.md @@ -1,12 +1,3 @@ - # Game Jam Los juegos son un área que ha sido fuertemente influenciada por los avances en IA y ML. En esta tarea, escribe un breve ensayo sobre un juego que te guste y que haya sido influenciado por la evolución de la IA. Debe ser un juego lo suficientemente antiguo como para haber sido influenciado por varios tipos de sistemas de procesamiento informático. Un buen ejemplo es el ajedrez o el Go, pero también considera videojuegos como Pong o Pac-Man. Escribe un ensayo que analice el pasado, presente y futuro de la IA en el juego. diff --git a/translations/es/lessons/2-Symbolic/README.md b/translations/es/lessons/2-Symbolic/README.md index 23bc066e..8446c76c 100644 --- a/translations/es/lessons/2-Symbolic/README.md +++ b/translations/es/lessons/2-Symbolic/README.md @@ -1,15 +1,6 @@ - # Representación del Conocimiento y Sistemas Expertos -![Resumen del contenido de IA Simbólica](../../../../../../translated_images/es/ai-symbolic.715a30cb610411a6.webp) +![Resumen del contenido de IA Simbólica](../../../../translated_images/es/ai-symbolic.715a30cb610411a6.webp) > Sketchnote por [Tomomi Imura](https://twitter.com/girlie_mac) @@ -41,7 +32,7 @@ Generalmente, no definimos estrictamente el conocimiento, sino que lo alineamos Así, el problema de la **representación del conocimiento** es hallar alguna forma efectiva de representar el conocimiento dentro de una computadora en forma de datos, para hacerlo automáticamente utilizable. Esto puede verse como un espectro: -![Espectro de representación del conocimiento](../../../../../../translated_images/es/knowledge-spectrum.b60df631852c0217.webp) +![Espectro de representación del conocimiento](../../../../translated_images/es/knowledge-spectrum.b60df631852c0217.webp) > Imagen por [Dmitry Soshnikov](http://soshnikov.com) @@ -94,7 +85,7 @@ Sintaxis de bloques | Indentación | | | Uno de los primeros éxitos de la IA simbólica fueron los llamados **sistemas expertos**: sistemas computacionales diseñados para actuar como un experto en un dominio problemático limitado. Se basaban en una **base de conocimiento** extraída de uno o más expertos humanos, y contenían un **motor de inferencia** que realizaba razonamientos sobre ella. -![Arquitectura humana](../../../../../../translated_images/es/arch-human.5d4d35f1bba3ab1c.webp) | ![Sistema basado en conocimiento](../../../../../../translated_images/es/arch-kbs.3ec5c150b09fa8da.webp) +![Arquitectura humana](../../../../translated_images/es/arch-human.5d4d35f1bba3ab1c.webp) | ![Sistema basado en conocimiento](../../../../translated_images/es/arch-kbs.3ec5c150b09fa8da.webp) ---------------------------------------------|------------------------------------------------ Estructura simplificada del sistema neural humano | Arquitectura de un sistema basado en conocimiento @@ -106,7 +97,7 @@ Los sistemas expertos están construidos como el sistema de razonamiento humano, Como ejemplo, consideremos el siguiente sistema experto para determinar un animal según sus características físicas: -![Árbol AND-OR](../../../../../../translated_images/es/AND-OR-Tree.5592d2c70187f283.webp) +![Árbol AND-OR](../../../../translated_images/es/AND-OR-Tree.5592d2c70187f283.webp) > Imagen por [Dmitry Soshnikov](http://soshnikov.com) diff --git a/translations/es/lessons/2-Symbolic/assignment.md b/translations/es/lessons/2-Symbolic/assignment.md index ef7e9cf7..32a6e877 100644 --- a/translations/es/lessons/2-Symbolic/assignment.md +++ b/translations/es/lessons/2-Symbolic/assignment.md @@ -1,12 +1,3 @@ - # Construir una Ontología Construir una base de conocimiento consiste en categorizar un modelo que representa hechos sobre un tema. Elige un tema, como una persona, un lugar o un objeto, y luego construye un modelo de ese tema. Utiliza algunas de las técnicas y estrategias de construcción de modelos descritas en esta lección. Un ejemplo sería crear una ontología de una sala de estar con muebles, luces, etc. ¿En qué se diferencia la sala de estar de la cocina? ¿Del baño? ¿Cómo sabes que es una sala de estar y no un comedor? Utiliza [Protégé](https://protege.stanford.edu/) para construir tu ontología. diff --git a/translations/es/lessons/3-NeuralNetworks/03-Perceptron/README.md b/translations/es/lessons/3-NeuralNetworks/03-Perceptron/README.md index b0f1fcfe..660a9c4d 100644 --- a/translations/es/lessons/3-NeuralNetworks/03-Perceptron/README.md +++ b/translations/es/lessons/3-NeuralNetworks/03-Perceptron/README.md @@ -1,12 +1,3 @@ - # Introducción a Redes Neuronales: Perceptrón ## [Cuestionario previo a la lección](https://ff-quizzes.netlify.app/en/ai/quiz/5) @@ -15,7 +6,7 @@ Uno de los primeros intentos de implementar algo similar a una red neuronal mode | | | |--------------|-----------| -|Frank Rosenblatt | El Perceptrón Mark 1| +|Frank Rosenblatt | El Perceptrón Mark 1| > Imágenes [de Wikipedia](https://en.wikipedia.org/wiki/Perceptron) @@ -34,7 +25,7 @@ y(x) = f(wTx) donde f es una función de activación escalón. - + ## Entrenamiento del Perceptrón diff --git a/translations/es/lessons/3-NeuralNetworks/03-Perceptron/lab/README.md b/translations/es/lessons/3-NeuralNetworks/03-Perceptron/lab/README.md index d13b95f6..f6182aee 100644 --- a/translations/es/lessons/3-NeuralNetworks/03-Perceptron/lab/README.md +++ b/translations/es/lessons/3-NeuralNetworks/03-Perceptron/lab/README.md @@ -1,12 +1,3 @@ - # Clasificación Multiclase con Perceptrón Asignación de laboratorio del [Currículo de IA para Principiantes](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/es/lessons/3-NeuralNetworks/04-OwnFramework/README.md b/translations/es/lessons/3-NeuralNetworks/04-OwnFramework/README.md index 6df6f4bf..7d750ae3 100644 --- a/translations/es/lessons/3-NeuralNetworks/04-OwnFramework/README.md +++ b/translations/es/lessons/3-NeuralNetworks/04-OwnFramework/README.md @@ -1,12 +1,3 @@ - # Introducción a Redes Neuronales. Perceptrón Multicapa En la sección anterior, aprendiste sobre el modelo de red neuronal más simple: el perceptrón de una sola capa, un modelo lineal de clasificación de dos clases. @@ -65,7 +56,7 @@ El algoritmo de descenso de gradiente permanecería igual, pero sería más dif Nota que la parte más a la izquierda de todas estas expresiones es la misma, y por lo tanto podemos calcular eficazmente las derivadas comenzando desde la función de pérdida y avanzando "hacia atrás" a través del gráfico computacional. Por lo tanto, el método de entrenamiento de un perceptrón multicapa se llama **retropropagación**, o 'backprop'. -compute graph +compute graph > TODO: cita de la imagen diff --git a/translations/es/lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md b/translations/es/lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md index 33c90ebc..445b4d87 100644 --- a/translations/es/lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md +++ b/translations/es/lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md @@ -1,12 +1,3 @@ - # Clasificación MNIST con Nuestro Propio Framework Asignación de laboratorio del [Currículo de AI para Principiantes](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/es/lessons/3-NeuralNetworks/05-Frameworks/README.md b/translations/es/lessons/3-NeuralNetworks/05-Frameworks/README.md index f0b39a6f..241837ce 100644 --- a/translations/es/lessons/3-NeuralNetworks/05-Frameworks/README.md +++ b/translations/es/lessons/3-NeuralNetworks/05-Frameworks/README.md @@ -1,12 +1,3 @@ - # Frameworks de Redes Neuronales Como ya hemos aprendido, para entrenar redes neuronales de manera eficiente necesitamos hacer dos cosas: diff --git a/translations/es/lessons/3-NeuralNetworks/05-Frameworks/lab/README.md b/translations/es/lessons/3-NeuralNetworks/05-Frameworks/lab/README.md index 471463d3..76313052 100644 --- a/translations/es/lessons/3-NeuralNetworks/05-Frameworks/lab/README.md +++ b/translations/es/lessons/3-NeuralNetworks/05-Frameworks/lab/README.md @@ -1,12 +1,3 @@ - # Clasificación con PyTorch/TensorFlow Asignación de laboratorio del [Currículo de AI para Principiantes](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/es/lessons/3-NeuralNetworks/README.md b/translations/es/lessons/3-NeuralNetworks/README.md index 2d46c2de..30be9145 100644 --- a/translations/es/lessons/3-NeuralNetworks/README.md +++ b/translations/es/lessons/3-NeuralNetworks/README.md @@ -1,12 +1,3 @@ - # Introducción a las Redes Neuronales ![Resumen del contenido de Introducción a Redes Neuronales en un dibujo](../../../../translated_images/es/ai-neuralnetworks.1c687ae40bc86e83.webp) diff --git a/translations/es/lessons/4-ComputerVision/06-IntroCV/README.md b/translations/es/lessons/4-ComputerVision/06-IntroCV/README.md index 543d31b2..41dc853a 100644 --- a/translations/es/lessons/4-ComputerVision/06-IntroCV/README.md +++ b/translations/es/lessons/4-ComputerVision/06-IntroCV/README.md @@ -1,12 +1,3 @@ - # Introducción a la Visión por Computadora [Visión por Computadora](https://wikipedia.org/wiki/Computer_vision) es una disciplina cuyo objetivo es permitir que las computadoras comprendan imágenes digitales a un nivel avanzado. Esta es una definición bastante amplia, ya que *comprender* puede significar muchas cosas diferentes, como encontrar un objeto en una imagen (**detección de objetos**), entender qué está ocurriendo (**detección de eventos**), describir una imagen en texto o reconstruir una escena en 3D. También hay tareas específicas relacionadas con imágenes humanas: estimación de edad y emociones, detección e identificación de rostros, y estimación de poses en 3D, entre otras. @@ -115,7 +106,7 @@ Lee más sobre flujo óptico [en este excelente tutorial](https://learnopencv.co En este laboratorio, tomarás un video con gestos simples, y tu objetivo será extraer movimientos hacia arriba/abajo/izquierda/derecha usando flujo óptico. -Fotograma de Movimiento de Palma +Fotograma de Movimiento de Palma --- diff --git a/translations/es/lessons/4-ComputerVision/06-IntroCV/lab/README.md b/translations/es/lessons/4-ComputerVision/06-IntroCV/lab/README.md index b2257d0a..984d02f6 100644 --- a/translations/es/lessons/4-ComputerVision/06-IntroCV/lab/README.md +++ b/translations/es/lessons/4-ComputerVision/06-IntroCV/lab/README.md @@ -1,12 +1,3 @@ - # Detectando Movimientos usando Flujo Óptico Asignación de laboratorio del [Currículo de AI para Principiantes](https://aka.ms/ai-beginners). diff --git a/translations/es/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md b/translations/es/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md index 336e7836..7ff271d0 100644 --- a/translations/es/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md +++ b/translations/es/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md @@ -1,12 +1,3 @@ - # Arquitecturas CNN Reconocidas ### VGG-16 @@ -25,7 +16,7 @@ Como puedes ver, VGG sigue una arquitectura piramidal tradicional, que es una se ResNet es una familia de modelos propuesta por Microsoft Research en 2015. La idea principal de ResNet es usar **bloques residuales**: - + > Imagen de [este artículo](https://arxiv.org/pdf/1512.03385.pdf) @@ -37,7 +28,7 @@ También puedes pensar en esta red como capaz de ajustar su complejidad al conju La arquitectura Google Inception lleva esta idea un paso más allá y construye cada capa de la red como una combinación de varios caminos diferentes: - + > Imagen de [Researchgate](https://www.researchgate.net/figure/Inception-module-with-dimension-reductions-left-and-schema-for-Inception-ResNet-v1_fig2_355547454) diff --git a/translations/es/lessons/4-ComputerVision/07-ConvNets/README.md b/translations/es/lessons/4-ComputerVision/07-ConvNets/README.md index f9a7d9f6..ca198a04 100644 --- a/translations/es/lessons/4-ComputerVision/07-ConvNets/README.md +++ b/translations/es/lessons/4-ComputerVision/07-ConvNets/README.md @@ -1,12 +1,3 @@ - # Redes Neuronales Convolucionales Hemos visto anteriormente que las redes neuronales son bastante buenas para trabajar con imágenes, e incluso un perceptrón de una sola capa es capaz de reconocer dígitos escritos a mano del conjunto de datos MNIST con una precisión razonable. Sin embargo, el conjunto de datos MNIST es muy especial, ya que todos los dígitos están centrados dentro de la imagen, lo que simplifica la tarea. @@ -24,7 +15,7 @@ Para extraer patrones, utilizaremos el concepto de **filtros convolucionales**. Por ejemplo, si aplicamos filtros de borde vertical y horizontal de 3x3 a los dígitos de MNIST, podemos resaltar (por ejemplo, valores altos) donde hay bordes verticales y horizontales en nuestra imagen original. Por lo tanto, esos dos filtros pueden usarse para "buscar" bordes. De manera similar, podemos diseñar diferentes filtros para buscar otros patrones de bajo nivel: - + > Imagen del [Banco de Filtros Leung-Malik](https://www.robots.ox.ac.uk/~vgg/research/texclass/filters.html) diff --git a/translations/es/lessons/4-ComputerVision/07-ConvNets/lab/README.md b/translations/es/lessons/4-ComputerVision/07-ConvNets/lab/README.md index 6be838e7..8de8bd9b 100644 --- a/translations/es/lessons/4-ComputerVision/07-ConvNets/lab/README.md +++ b/translations/es/lessons/4-ComputerVision/07-ConvNets/lab/README.md @@ -1,12 +1,3 @@ - # Clasificación de Caras de Mascotas Asignación de laboratorio del [Currículo de AI para Principiantes](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/es/lessons/4-ComputerVision/08-TransferLearning/README.md b/translations/es/lessons/4-ComputerVision/08-TransferLearning/README.md index bb31012d..8d727b0d 100644 --- a/translations/es/lessons/4-ComputerVision/08-TransferLearning/README.md +++ b/translations/es/lessons/4-ComputerVision/08-TransferLearning/README.md @@ -1,12 +1,3 @@ - # Redes Pre-entrenadas y Aprendizaje por Transferencia Entrenar redes neuronales convolucionales (CNNs) puede llevar mucho tiempo y requiere una gran cantidad de datos. Sin embargo, gran parte del tiempo se invierte en aprender los mejores filtros de bajo nivel que una red puede usar para extraer patrones de las imágenes. Surge una pregunta natural: ¿podemos usar una red neuronal entrenada en un conjunto de datos y adaptarla para clasificar imágenes diferentes sin necesidad de un proceso completo de entrenamiento? diff --git a/translations/es/lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md b/translations/es/lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md index 38516165..705d3138 100644 --- a/translations/es/lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md +++ b/translations/es/lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md @@ -1,12 +1,3 @@ - # Trucos para el Entrenamiento de Redes Neuronales Profundas A medida que las redes neuronales se vuelven más profundas, el proceso de entrenamiento se vuelve cada vez más desafiante. Uno de los principales problemas son los llamados [gradientes que desaparecen](https://en.wikipedia.org/wiki/Vanishing_gradient_problem) o [gradientes que explotan](https://deepai.org/machine-learning-glossary-and-terms/exploding-gradient-problem#:~:text=Exploding%20gradients%20are%20a%20problem,updates%20are%20small%20and%20controlled.). [Este artículo](https://towardsdatascience.com/the-vanishing-exploding-gradient-problem-in-deep-neural-networks-191358470c11) ofrece una buena introducción a estos problemas. diff --git a/translations/es/lessons/4-ComputerVision/08-TransferLearning/lab/README.md b/translations/es/lessons/4-ComputerVision/08-TransferLearning/lab/README.md index 64a7fac3..09066c47 100644 --- a/translations/es/lessons/4-ComputerVision/08-TransferLearning/lab/README.md +++ b/translations/es/lessons/4-ComputerVision/08-TransferLearning/lab/README.md @@ -1,12 +1,3 @@ - # Clasificación de Mascotas de Oxford usando Aprendizaje por Transferencia Asignación de laboratorio del [Currículo de AI para Principiantes](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/es/lessons/4-ComputerVision/09-Autoencoders/README.md b/translations/es/lessons/4-ComputerVision/09-Autoencoders/README.md index 4ad30a8c..7c48c751 100644 --- a/translations/es/lessons/4-ComputerVision/09-Autoencoders/README.md +++ b/translations/es/lessons/4-ComputerVision/09-Autoencoders/README.md @@ -1,12 +1,3 @@ - # Autoencoders Al entrenar CNNs, uno de los problemas es que necesitamos una gran cantidad de datos etiquetados. En el caso de la clasificación de imágenes, debemos separar las imágenes en diferentes clases, lo cual requiere un esfuerzo manual. @@ -46,7 +37,7 @@ En resumen: * Muestreamos un vector `sample` de la distribución N(zmean,exp(zlog\_sigma)). * El decodificador intenta decodificar la imagen original usando `sample` como vector de entrada. - + > Imagen de [este blog](https://ijdykeman.github.io/ml/2016/12/21/cvae.html) por Isaak Dykeman @@ -57,13 +48,13 @@ Los autoencoders variacionales utilizan una función de pérdida compleja que co Una ventaja importante de los VAEs es que nos permiten generar nuevas imágenes de manera relativamente sencilla, porque sabemos de qué distribución muestrear los vectores latentes. Por ejemplo, si entrenamos un VAE con un vector latente 2D en MNIST, luego podemos variar los componentes del vector latente para obtener diferentes dígitos: -vaemnist +vaemnist > Imagen por [Dmitry Soshnikov](http://soshnikov.com) Observa cómo las imágenes se mezclan entre sí, a medida que comenzamos a obtener vectores latentes de diferentes partes del espacio de parámetros latentes. También podemos visualizar este espacio en 2D: -vaemnist cluster +vaemnist cluster > Imagen por [Dmitry Soshnikov](http://soshnikov.com) diff --git a/translations/es/lessons/4-ComputerVision/10-GANs/README.md b/translations/es/lessons/4-ComputerVision/10-GANs/README.md index 40fc67b2..85416589 100644 --- a/translations/es/lessons/4-ComputerVision/10-GANs/README.md +++ b/translations/es/lessons/4-ComputerVision/10-GANs/README.md @@ -1,12 +1,3 @@ - # Redes Generativas Antagónicas En la sección anterior, aprendimos sobre los **modelos generativos**: modelos que pueden generar nuevas imágenes similares a las del conjunto de datos de entrenamiento. VAE fue un buen ejemplo de un modelo generativo. @@ -17,7 +8,7 @@ Sin embargo, si intentamos generar algo realmente significativo, como una pintur La idea principal de una GAN es tener dos redes neuronales que se entrenan una contra la otra: - + > Imagen por [Dmitry Soshnikov](http://soshnikov.com) @@ -41,7 +32,7 @@ Un Generador es un poco más complicado. Puedes considerarlo como un discriminad > ✅ Debido a que la capa de convolución se implementa como un filtro lineal que recorre la imagen, la deconvolución es esencialmente similar a la convolución y puede implementarse utilizando la misma lógica de capa. - + > Imagen por [Dmitry Soshnikov](http://soshnikov.com) diff --git a/translations/es/lessons/4-ComputerVision/11-ObjectDetection/README.md b/translations/es/lessons/4-ComputerVision/11-ObjectDetection/README.md index 6584d9be..594af796 100644 --- a/translations/es/lessons/4-ComputerVision/11-ObjectDetection/README.md +++ b/translations/es/lessons/4-ComputerVision/11-ObjectDetection/README.md @@ -1,12 +1,3 @@ - # Detección de Objetos Los modelos de clasificación de imágenes que hemos tratado hasta ahora toman una imagen y producen un resultado categórico, como la clase 'número' en un problema de MNIST. Sin embargo, en muchos casos no solo queremos saber que una imagen contiene objetos, sino que queremos determinar su ubicación precisa. Este es precisamente el objetivo de la **detección de objetos**. diff --git a/translations/es/lessons/4-ComputerVision/11-ObjectDetection/lab/README.md b/translations/es/lessons/4-ComputerVision/11-ObjectDetection/lab/README.md index 7d388401..77563e79 100644 --- a/translations/es/lessons/4-ComputerVision/11-ObjectDetection/lab/README.md +++ b/translations/es/lessons/4-ComputerVision/11-ObjectDetection/lab/README.md @@ -1,12 +1,3 @@ - # Detección de Cabezas usando el Conjunto de Datos Hollywood Heads Asignación de laboratorio del [Currículo de IA para Principiantes](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/es/lessons/4-ComputerVision/12-Segmentation/README.md b/translations/es/lessons/4-ComputerVision/12-Segmentation/README.md index dda81397..5b4f5228 100644 --- a/translations/es/lessons/4-ComputerVision/12-Segmentation/README.md +++ b/translations/es/lessons/4-ComputerVision/12-Segmentation/README.md @@ -1,12 +1,3 @@ - # Segmentación Anteriormente hemos aprendido sobre la Detección de Objetos, que nos permite localizar objetos en una imagen prediciendo sus *cajas delimitadoras*. Sin embargo, para algunas tareas no solo necesitamos cajas delimitadoras, sino también una localización más precisa de los objetos. Esta tarea se llama **segmentación**. @@ -20,7 +11,7 @@ La segmentación puede considerarse como una **clasificación de píxeles**, en Por ejemplo, en la segmentación por instancia, estas ovejas son objetos diferentes, pero en la segmentación semántica todas las ovejas se representan como una sola clase. - + > Imagen tomada de [este artículo](https://nirmalamurali.medium.com/image-classification-vs-semantic-segmentation-vs-instance-segmentation-625c33a08d50) @@ -29,7 +20,7 @@ Existen diferentes arquitecturas neuronales para la segmentación, pero todas ti * **Codificador** que extrae características de la imagen de entrada. * **Decodificador** que transforma esas características en la **imagen de máscara**, con el mismo tamaño y número de canales correspondientes al número de clases. - + > Imagen tomada de [esta publicación](https://arxiv.org/pdf/2001.05566.pdf) @@ -43,7 +34,7 @@ En esta lección, veremos la segmentación en acción entrenando una red para re > ✅ Esta técnica es particularmente adecuada para este tipo de imágenes médicas, pero ¿qué otras aplicaciones del mundo real podrías imaginar? -navi +navi > Imagen tomada de la Base de Datos PH2 diff --git a/translations/es/lessons/4-ComputerVision/12-Segmentation/lab/README.md b/translations/es/lessons/4-ComputerVision/12-Segmentation/lab/README.md index e619b48e..ef7b7153 100644 --- a/translations/es/lessons/4-ComputerVision/12-Segmentation/lab/README.md +++ b/translations/es/lessons/4-ComputerVision/12-Segmentation/lab/README.md @@ -1,12 +1,3 @@ - # Segmentación del Cuerpo Humano Asignación de laboratorio del [Currículo de IA para Principiantes](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/es/lessons/4-ComputerVision/README.md b/translations/es/lessons/4-ComputerVision/README.md index 330e8c10..7092d158 100644 --- a/translations/es/lessons/4-ComputerVision/README.md +++ b/translations/es/lessons/4-ComputerVision/README.md @@ -1,12 +1,3 @@ - # Visión por Computadora ![Resumen del contenido de Visión por Computadora en un dibujo](../../../../lessons/sketchnotes/ai-computervision.png) diff --git a/translations/es/lessons/5-NLP/13-TextRep/README.md b/translations/es/lessons/5-NLP/13-TextRep/README.md index ab119003..9d740ab7 100644 --- a/translations/es/lessons/5-NLP/13-TextRep/README.md +++ b/translations/es/lessons/5-NLP/13-TextRep/README.md @@ -1,12 +1,3 @@ - # Representar texto como tensores ## [Cuestionario previo a la clase](https://ff-quizzes.netlify.app/en/ai/quiz/25) @@ -25,7 +16,7 @@ Nuestro objetivo será clasificar el artículo de noticias en una de las categor Si queremos resolver tareas de Procesamiento de Lenguaje Natural (NLP) con redes neuronales, necesitamos una forma de representar el texto como tensores. Las computadoras ya representan caracteres textuales como números que se asignan a fuentes en tu pantalla utilizando codificaciones como ASCII o UTF-8. -Imagen que muestra un diagrama que mapea un carácter a una representación ASCII y binaria +Imagen que muestra un diagrama que mapea un carácter a una representación ASCII y binaria > [Fuente de la imagen](https://www.seobility.net/en/wiki/ASCII) @@ -48,7 +39,7 @@ En algunos casos, podemos considerar usar tri-gramas, combinaciones de tres pala Cuando resolvemos tareas como la clasificación de texto, necesitamos poder representar el texto mediante un vector de tamaño fijo, que utilizaremos como entrada para el clasificador denso final. Una de las formas más simples de hacerlo es combinar todas las representaciones individuales de palabras, por ejemplo, sumándolas. Si sumamos las codificaciones one-hot de cada palabra, terminaremos con un vector de frecuencias que muestra cuántas veces aparece cada palabra dentro del texto. Tal representación del texto se llama **bag of words** (BoW). - + > Imagen del autor diff --git a/translations/es/lessons/5-NLP/13-TextRep/assignment.md b/translations/es/lessons/5-NLP/13-TextRep/assignment.md index 4683516a..6a975246 100644 --- a/translations/es/lessons/5-NLP/13-TextRep/assignment.md +++ b/translations/es/lessons/5-NLP/13-TextRep/assignment.md @@ -1,12 +1,3 @@ - # Tarea: Cuadernos Usando los cuadernos asociados a esta lección (ya sea la versión de PyTorch o TensorFlow), ejecútalos nuevamente utilizando tu propio conjunto de datos, tal vez uno de Kaggle, usado con la debida atribución. Reescribe el cuaderno para resaltar tus propios hallazgos. Prueba con algunos conjuntos de datos innovadores que puedan resultar sorprendentes, como [este sobre avistamientos de ovnis](https://www.kaggle.com/datasets/NUFORC/ufo-sightings) de NUFORC. diff --git a/translations/es/lessons/5-NLP/14-Embeddings/README.md b/translations/es/lessons/5-NLP/14-Embeddings/README.md index f56ca266..a80e615f 100644 --- a/translations/es/lessons/5-NLP/14-Embeddings/README.md +++ b/translations/es/lessons/5-NLP/14-Embeddings/README.md @@ -1,12 +1,3 @@ - # Embeddings ## [Pre-lecture quiz](https://ff-quizzes.netlify.app/en/ai/quiz/27) diff --git a/translations/es/lessons/5-NLP/14-Embeddings/assignment.md b/translations/es/lessons/5-NLP/14-Embeddings/assignment.md index 70170acc..3f6fb7e7 100644 --- a/translations/es/lessons/5-NLP/14-Embeddings/assignment.md +++ b/translations/es/lessons/5-NLP/14-Embeddings/assignment.md @@ -1,12 +1,3 @@ - # Asignación: Cuadernos Usando los cuadernos asociados a esta lección (ya sea la versión de PyTorch o TensorFlow), vuelve a ejecutarlos utilizando tu propio conjunto de datos, tal vez uno de Kaggle, usado con atribución. Reescribe el cuaderno para destacar tus propios hallazgos. Prueba con un tipo diferente de conjunto de datos y documenta tus descubrimientos, utilizando texto como [estas letras de los Beatles](https://www.kaggle.com/datasets/jenlooper/beatles-lyrics). diff --git a/translations/es/lessons/5-NLP/15-LanguageModeling/README.md b/translations/es/lessons/5-NLP/15-LanguageModeling/README.md index 76c3d30d..dd8ea2fa 100644 --- a/translations/es/lessons/5-NLP/15-LanguageModeling/README.md +++ b/translations/es/lessons/5-NLP/15-LanguageModeling/README.md @@ -1,12 +1,3 @@ - # Modelado de Lenguaje Las incrustaciones semánticas, como Word2Vec y GloVe, son de hecho un primer paso hacia el **modelado de lenguaje**: crear modelos que de alguna manera *entiendan* (o *representen*) la naturaleza del lenguaje. diff --git a/translations/es/lessons/5-NLP/15-LanguageModeling/lab/README.md b/translations/es/lessons/5-NLP/15-LanguageModeling/lab/README.md index 0d345a36..a151d796 100644 --- a/translations/es/lessons/5-NLP/15-LanguageModeling/lab/README.md +++ b/translations/es/lessons/5-NLP/15-LanguageModeling/lab/README.md @@ -1,12 +1,3 @@ - # Entrenando el Modelo Skip-Gram Asignación de laboratorio del [Currículo de AI para Principiantes](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/es/lessons/5-NLP/16-RNN/README.md b/translations/es/lessons/5-NLP/16-RNN/README.md index 4f064773..fc4dc959 100644 --- a/translations/es/lessons/5-NLP/16-RNN/README.md +++ b/translations/es/lessons/5-NLP/16-RNN/README.md @@ -1,12 +1,3 @@ - # Redes Neuronales Recurrentes ## [Cuestionario previo a la clase](https://ff-quizzes.netlify.app/en/ai/quiz/31) @@ -31,7 +22,7 @@ Veamos cómo está organizada una célula RNN simple. Acepta el estado anterior Una célula RNN simple tiene dos matrices de pesos internas: una transforma un símbolo de entrada (llamémosla W) y otra transforma un estado de entrada (H). En este caso, la salida de la red se calcula como σ(W×Xi+H×Si-1+b), donde σ es la función de activación y b es un sesgo adicional. -Anatomía de una célula RNN +Anatomía de una célula RNN > Imagen del autor diff --git a/translations/es/lessons/5-NLP/16-RNN/assignment.md b/translations/es/lessons/5-NLP/16-RNN/assignment.md index d41073db..6136d1db 100644 --- a/translations/es/lessons/5-NLP/16-RNN/assignment.md +++ b/translations/es/lessons/5-NLP/16-RNN/assignment.md @@ -1,12 +1,3 @@ - # Tarea: Cuadernos Usando los cuadernos asociados a esta lección (ya sea la versión de PyTorch o la de TensorFlow), ejecútalos nuevamente utilizando tu propio conjunto de datos, tal vez uno de Kaggle, usado con la debida atribución. Reescribe el cuaderno para resaltar tus propios hallazgos. Prueba con un tipo diferente de conjunto de datos y documenta tus descubrimientos, utilizando texto como [este conjunto de datos de una competencia de Kaggle sobre tweets relacionados con el clima](https://www.kaggle.com/competitions/crowdflower-weather-twitter/data?select=train.csv). diff --git a/translations/es/lessons/5-NLP/17-GenerativeNetworks/README.md b/translations/es/lessons/5-NLP/17-GenerativeNetworks/README.md index 3760d170..c9092558 100644 --- a/translations/es/lessons/5-NLP/17-GenerativeNetworks/README.md +++ b/translations/es/lessons/5-NLP/17-GenerativeNetworks/README.md @@ -1,12 +1,3 @@ - # Redes generativas ## [Cuestionario previo a la clase](https://ff-quizzes.netlify.app/en/ai/quiz/33) @@ -36,7 +27,7 @@ Entrenaremos esta RNN para generar texto paso a paso. En cada paso, tomaremos un Cuando generamos texto (durante la inferencia), comenzamos con un **prompt**, que se pasa por las celdas RNN para generar su estado intermedio, y luego comienza la generación desde este estado. Generamos un carácter a la vez y pasamos el estado y el carácter generado a otra celda RNN para generar el siguiente, hasta que generemos suficientes caracteres. - + > Imagen del autor diff --git a/translations/es/lessons/5-NLP/17-GenerativeNetworks/lab/README.md b/translations/es/lessons/5-NLP/17-GenerativeNetworks/lab/README.md index 915b0b1b..062aae6f 100644 --- a/translations/es/lessons/5-NLP/17-GenerativeNetworks/lab/README.md +++ b/translations/es/lessons/5-NLP/17-GenerativeNetworks/lab/README.md @@ -1,12 +1,3 @@ - # Generación de texto a nivel de palabra usando RNNs Asignación de laboratorio del [Currículo de IA para Principiantes](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/es/lessons/5-NLP/18-Transformers/README.md b/translations/es/lessons/5-NLP/18-Transformers/README.md index 0435dc8c..62549576 100644 --- a/translations/es/lessons/5-NLP/18-Transformers/README.md +++ b/translations/es/lessons/5-NLP/18-Transformers/README.md @@ -1,12 +1,3 @@ - # Mecanismos de Atención y Transformadores ## [Cuestionario previo a la clase](https://ff-quizzes.netlify.app/en/ai/quiz/35) @@ -56,7 +47,7 @@ La idea de la codificación posicional es la siguiente. * Embebido entrenable, similar al embebido de tokens. Este es el enfoque que consideramos aquí. Aplicamos capas de embebido tanto a los tokens como a sus posiciones, obteniendo vectores de embebido de las mismas dimensiones, que luego sumamos. * Función de codificación posicional fija, como se propone en el artículo original. - + > Imagen del autor diff --git a/translations/es/lessons/5-NLP/18-Transformers/READMEtransformers.md b/translations/es/lessons/5-NLP/18-Transformers/READMEtransformers.md deleted file mode 100644 index 578fe89d..00000000 --- a/translations/es/lessons/5-NLP/18-Transformers/READMEtransformers.md +++ /dev/null @@ -1,112 +0,0 @@ -# Mecanismos de Atención y Transformadores - -## [Cuestionario previo a la clase](https://ff-quizzes.netlify.app/en/ai/quiz/35) - -Uno de los problemas más importantes en el dominio del PLN es la **traducción automática**, una tarea esencial que subyace a herramientas como Google Translate. En esta sección, nos enfocaremos en la traducción automática, o, de manera más general, en cualquier tarea de *secuencia a secuencia* (que también se llama **transducción de oraciones**). - -Con las RNN, la secuencia a secuencia se implementa mediante dos redes recurrentes, donde una red, el **codificador**, colapsa una secuencia de entrada en un estado oculto, mientras que otra red, el **decodificador**, despliega este estado oculto en un resultado traducido. Hay un par de problemas con este enfoque: - -* El estado final de la red del codificador tiene dificultades para recordar el comienzo de una oración, lo que provoca una mala calidad del modelo para oraciones largas. -* Todas las palabras en una secuencia tienen el mismo impacto en el resultado. Sin embargo, en la realidad, ciertas palabras en la secuencia de entrada a menudo tienen más impacto en las salidas secuenciales que otras. - -**Los Mecanismos de Atención** proporcionan un medio para ponderar el impacto contextual de cada vector de entrada en cada predicción de salida de la RNN. La forma en que se implementa es creando atajos entre los estados intermedios de la RNN de entrada y la RNN de salida. De esta manera, al generar el símbolo de salida yt, tomaremos en cuenta todos los estados ocultos de entrada hi, con diferentes coeficientes de peso αt,i. - -![Imagen que muestra un modelo de codificador/decodificador con una capa de atención aditiva](../../../../../translated_images/es/encoder-decoder-attention.7a726296894fb567.webp) - -> El modelo codificador-decodificador con mecanismo de atención aditiva en [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf), citado de [esta publicación de blog](https://lilianweng.github.io/lil-log/2018/06/24/attention-attention.html) - -La matriz de atención {αi,j} representaría el grado en que ciertas palabras de entrada juegan un papel en la generación de una palabra dada en la secuencia de salida. A continuación se muestra un ejemplo de tal matriz: - -![Imagen que muestra una alineación de muestra encontrada por RNNsearch-50, tomada de Bahdanau - arviz.org](../../../../../translated_images/es/bahdanau-fig3.09ba2d37f202a6af.webp) - -> Figura de [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf) (Fig.3) - -Los mecanismos de atención son responsables de gran parte del estado actual o casi actual de la técnica en PLN. Sin embargo, añadir atención aumenta considerablemente el número de parámetros del modelo, lo que llevó a problemas de escalabilidad con las RNN. Una restricción clave de la escalabilidad de las RNN es que la naturaleza recurrente de los modelos hace que sea un desafío agrupar y paralelizar el entrenamiento. En una RNN, cada elemento de una secuencia necesita ser procesado en orden secuencial, lo que significa que no se puede paralelizar fácilmente. - -![Codificador Decodificador con Atención](../../../../../lessons/5-NLP/18-Transformers/images/EncDecAttention.gif) - -> Figura de [Google's Blog](https://research.googleblog.com/2016/09/a-neural-network-for-machine.html) - -La adopción de mecanismos de atención combinados con esta restricción llevó a la creación de los ahora modelos transformadores de última generación que conocemos y utilizamos hoy, como BERT y Open-GPT3. - -## Modelos Transformadores - -Una de las ideas principales detrás de los transformadores es evitar la naturaleza secuencial de las RNN y crear un modelo que sea paralelizable durante el entrenamiento. Esto se logra implementando dos ideas: - -* codificación posicional -* uso de un mecanismo de autoatención para capturar patrones en lugar de RNNs (o CNNs) (por eso el artículo que introduce los transformadores se llama *[Attention is all you need](https://arxiv.org/abs/1706.03762)*) - -### Codificación/Embebido Posicional - -La idea de la codificación posicional es la siguiente. -1. Al usar RNNs, la posición relativa de los tokens está representada por el número de pasos, y por lo tanto no necesita ser representada explícitamente. -2. Sin embargo, una vez que cambiamos a atención, necesitamos conocer las posiciones relativas de los tokens dentro de una secuencia. -3. Para obtener la codificación posicional, aumentamos nuestra secuencia de tokens con una secuencia de posiciones de tokens en la secuencia (es decir, una secuencia de números 0,1, ...). -4. Luego mezclamos la posición del token con un vector de embebido del token. Para transformar la posición (entero) en un vector, podemos usar diferentes enfoques: - -* Embebido entrenable, similar al embebido de tokens. Este es el enfoque que consideramos aquí. Aplicamos capas de embebido sobre ambos, los tokens y sus posiciones, resultando en vectores de embebido de las mismas dimensiones, que luego sumamos. -* Función de codificación de posición fija, como se propuso en el artículo original. - - - -> Imagen del autor - -El resultado que obtenemos con el embebido posicional integra tanto el token original como su posición dentro de una secuencia. - -### Autoatención Multi-Cabeza - -A continuación, necesitamos capturar algunos patrones dentro de nuestra secuencia. Para hacer esto, los transformadores utilizan un mecanismo de **autoatención**, que es esencialmente atención aplicada a la misma secuencia como entrada y salida. Aplicar autoatención nos permite tener en cuenta el **contexto** dentro de la oración y ver qué palabras están interrelacionadas. Por ejemplo, nos permite ver qué palabras son referidas por co-referencias, como *ello*, y también tener en cuenta el contexto: - -![](../../../../../translated_images/es/CoreferenceResolution.861924d6d384a7d6.webp) - -> Imagen del [Blog de Google](https://research.googleblog.com/2017/08/transformer-novel-neural-network.html) - -En los transformadores, utilizamos **Atención Multi-Cabeza** para darle al modelo la capacidad de capturar varios tipos diferentes de dependencias, por ejemplo, relaciones de palabras a largo plazo frente a corto plazo, co-referencia frente a algo más, etc. - -[Notebook de TensorFlow](../../../../../lessons/5-NLP/18-Transformers/TransformersTF.ipynb) contiene más detalles sobre la implementación de las capas transformadoras. - -### Atención Codificador-Decodificador - -En los transformadores, la atención se utiliza en dos lugares: - -* Para capturar patrones dentro del texto de entrada utilizando autoatención -* Para realizar traducción de secuencias - es la capa de atención entre el codificador y el decodificador. - -La atención codificador-decodificador es muy similar al mecanismo de atención utilizado en las RNN, como se describió al principio de esta sección. Este diagrama animado explica el papel de la atención codificador-decodificador. - -![GIF animado que muestra cómo se realizan las evaluaciones en modelos transformadores.](../../../../../lessons/5-NLP/18-Transformers/images/transformer-animated-explanation.gif) - -Dado que cada posición de entrada se mapea independientemente a cada posición de salida, los transformadores pueden paralelizarse mejor que las RNN, lo que permite modelos de lenguaje mucho más grandes y expresivos. Cada cabeza de atención puede ser utilizada para aprender diferentes relaciones entre palabras que mejoran las tareas de Procesamiento de Lenguaje Natural posteriores. - -## BERT - -**BERT** (Representaciones de Codificador Bidireccional de Transformadores) es una red transformadora de múltiples capas muy grande con 12 capas para *BERT-base*, y 24 para *BERT-large*. El modelo se preentrena primero en un gran corpus de datos textuales (WikiPedia + libros) utilizando entrenamiento no supervisado (prediciendo palabras enmascaradas en una oración). Durante el preentrenamiento, el modelo absorbe niveles significativos de comprensión del lenguaje que luego pueden ser aprovechados con otros conjuntos de datos utilizando ajuste fino. Este proceso se llama **aprendizaje por transferencia**. - -![imagen de http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/es/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp) - -> Imagen [fuente](http://jalammar.github.io/illustrated-bert/) - -## ✍️ Ejercicios: Transformadores - -Continúa tu aprendizaje en los siguientes cuadernos: - -* [Transformadores en PyTorch](../../../../../lessons/5-NLP/18-Transformers/TransformersPyTorch.ipynb) -* [Transformadores en TensorFlow](../../../../../lessons/5-NLP/18-Transformers/TransformersTF.ipynb) - -## Conclusión - -En esta lección aprendiste sobre Transformadores y Mecanismos de Atención, herramientas esenciales en la caja de herramientas del PLN. Hay muchas variaciones de arquitecturas de Transformadores, incluyendo BERT, DistilBERT, BigBird, OpenGPT3 y más que pueden ser ajustadas. El [paquete HuggingFace](https://github.com/huggingface/) proporciona un repositorio para entrenar muchas de estas arquitecturas tanto con PyTorch como con TensorFlow. - -## 🚀 Desafío - -## [Cuestionario posterior a la clase](https://ff-quizzes.netlify.app/en/ai/quiz/36) - -## Revisión y Autoestudio - -* [Publicación de blog](https://mchromiak.github.io/articles/2017/Sep/12/Transformer-Attention-is-all-you-need/), que explica el clásico artículo [Attention is all you need](https://arxiv.org/abs/1706.03762) sobre transformadores. -* [Una serie de publicaciones de blog](https://towardsdatascience.com/transformers-explained-visually-part-1-overview-of-functionality-95a6dd460452) sobre transformadores, que explican la arquitectura en detalle. - -## [Asignación](assignment.md) - -**Descargo de responsabilidad**: -Este documento ha sido traducido utilizando servicios de traducción automática basados en inteligencia artificial. Si bien nos esforzamos por lograr precisión, tenga en cuenta que las traducciones automatizadas pueden contener errores o inexactitudes. El documento original en su idioma nativo debe considerarse la fuente autorizada. Para información crítica, se recomienda una traducción profesional realizada por humanos. No nos hacemos responsables de malentendidos o malas interpretaciones que surjan del uso de esta traducción. \ No newline at end of file diff --git a/translations/es/lessons/5-NLP/18-Transformers/assignment.md b/translations/es/lessons/5-NLP/18-Transformers/assignment.md index c8c8b752..f2820155 100644 --- a/translations/es/lessons/5-NLP/18-Transformers/assignment.md +++ b/translations/es/lessons/5-NLP/18-Transformers/assignment.md @@ -1,12 +1,3 @@ - # Asignación: Transformers ¡Experimenta con Transformers en HuggingFace! Prueba algunos de los scripts que ofrecen para trabajar con los diversos modelos disponibles en su sitio: https://huggingface.co/docs/transformers/run_scripts. Prueba uno de sus conjuntos de datos, luego importa uno propio de este currículo o de Kaggle y mira si puedes generar textos interesantes. Produce un cuaderno con tus hallazgos. diff --git a/translations/es/lessons/5-NLP/19-NER/README.md b/translations/es/lessons/5-NLP/19-NER/README.md index c92511f6..aa1b5d67 100644 --- a/translations/es/lessons/5-NLP/19-NER/README.md +++ b/translations/es/lessons/5-NLP/19-NER/README.md @@ -1,12 +1,3 @@ - # Reconocimiento de Entidades Nombradas Hasta ahora, nos hemos concentrado principalmente en una tarea de PLN: la clasificación. Sin embargo, existen otras tareas de PLN que pueden realizarse con redes neuronales. Una de esas tareas es **[Reconocimiento de Entidades Nombradas](https://wikipedia.org/wiki/Named-entity_recognition)** (NER), que se ocupa de identificar entidades específicas dentro del texto, como lugares, nombres de personas, intervalos de tiempo, fórmulas químicas, entre otros. @@ -17,7 +8,7 @@ Hasta ahora, nos hemos concentrado principalmente en una tarea de PLN: la clasif Supongamos que quieres desarrollar un chatbot de lenguaje natural, similar a Amazon Alexa o Google Assistant. La forma en que funcionan los chatbots inteligentes es *entendiendo* lo que el usuario quiere mediante la clasificación de texto en la frase de entrada. El resultado de esta clasificación es el llamado **intento**, que determina lo que el chatbot debe hacer. -Bot NER +Bot NER > Imagen del autor diff --git a/translations/es/lessons/5-NLP/19-NER/lab/README.md b/translations/es/lessons/5-NLP/19-NER/lab/README.md index fdbcd787..d5e3b059 100644 --- a/translations/es/lessons/5-NLP/19-NER/lab/README.md +++ b/translations/es/lessons/5-NLP/19-NER/lab/README.md @@ -1,12 +1,3 @@ - # NER Asignación de laboratorio del [Currículo de AI para Principiantes](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/es/lessons/5-NLP/20-LangModels/README.md b/translations/es/lessons/5-NLP/20-LangModels/README.md index 105b950e..83a5c5aa 100644 --- a/translations/es/lessons/5-NLP/20-LangModels/README.md +++ b/translations/es/lessons/5-NLP/20-LangModels/README.md @@ -1,12 +1,3 @@ - # Modelos de Lenguaje Grandes Preentrenados En todas nuestras tareas anteriores, entrenábamos una red neuronal para realizar una tarea específica utilizando un conjunto de datos etiquetado. Con modelos transformadores grandes, como BERT, utilizamos el modelado de lenguaje de manera auto-supervisada para construir un modelo de lenguaje, que luego se especializa en tareas específicas con entrenamiento adicional en dominios concretos. Sin embargo, se ha demostrado que los modelos de lenguaje grandes también pueden resolver muchas tareas sin necesidad de entrenamiento específico en un dominio. Una familia de modelos capaz de hacer esto se llama **GPT**: Transformador Generativo Preentrenado. diff --git a/translations/es/lessons/5-NLP/20-LangModels/READMELargeLang.md b/translations/es/lessons/5-NLP/20-LangModels/READMELargeLang.md deleted file mode 100644 index bbc00375..00000000 --- a/translations/es/lessons/5-NLP/20-LangModels/READMELargeLang.md +++ /dev/null @@ -1,55 +0,0 @@ -# Modelos de Lenguaje de Gran Tamaño Preentrenados - -En todas nuestras tareas anteriores, entrenábamos una red neuronal para realizar una tarea específica utilizando un conjunto de datos etiquetados. Con modelos de transformador grandes, como BERT, utilizamos el modelado del lenguaje de manera auto-supervisada para construir un modelo de lenguaje, que luego se especializa para tareas específicas de acuerdo con un entrenamiento adicional en dominios específicos. Sin embargo, se ha demostrado que los grandes modelos de lenguaje también pueden resolver muchas tareas sin NINGÚN entrenamiento específico de dominio. Una familia de modelos capaces de hacer esto se llama **GPT**: Transformador Generativo Preentrenado. - -## [Cuestionario previo a la conferencia](https://ff-quizzes.netlify.app/en/ai/quiz/39) - -## Generación de Texto y Perplejidad - -La idea de que una red neuronal pueda realizar tareas generales sin entrenamiento posterior se presenta en el artículo [Language Models are Unsupervised Multitask Learners](https://cdn.openai.com/better-language-models/language_models_are_unsupervised_multitask_learners.pdf). La idea principal es que muchas otras tareas pueden ser modeladas utilizando **generación de texto**, porque entender texto significa, esencialmente, ser capaz de producirlo. Dado que el modelo se entrena con una gran cantidad de texto que abarca el conocimiento humano, también se vuelve conocedor de una amplia variedad de temas. - -> Entender y ser capaz de producir texto también implica saber algo sobre el mundo que nos rodea. Las personas también aprenden leyendo en gran medida, y la red GPT es similar en este aspecto. - -Las redes de generación de texto funcionan prediciendo la probabilidad de la siguiente palabra $$P(w_N)$$. Sin embargo, la probabilidad incondicional de la siguiente palabra es igual a la frecuencia de esta palabra en el corpus de texto. GPT es capaz de darnos la **probabilidad condicional** de la siguiente palabra, dado las anteriores: $$P(w_N | w_{n-1}, ..., w_0)$$ - -> Puedes leer más sobre probabilidades en nuestro [Currículo de Ciencia de Datos para Principiantes](https://github.com/microsoft/Data-Science-For-Beginners/tree/main/1-Introduction/04-stats-and-probability) - -La calidad del modelo de generación de lenguaje se puede definir utilizando **perplejidad**. Es una métrica intrínseca que nos permite medir la calidad del modelo sin ningún conjunto de datos específico para la tarea. Se basa en la noción de *probabilidad de una oración*: el modelo asigna alta probabilidad a una oración que es probable que sea real (es decir, el modelo no está **perplejo** por ella), y baja probabilidad a oraciones que tienen menos sentido (por ejemplo, *¿Puede hacer qué?*). Cuando le damos a nuestro modelo oraciones de un corpus de texto real, esperaríamos que tuvieran alta probabilidad y baja **perplejidad**. Matemáticamente, se define como la probabilidad inversa normalizada del conjunto de prueba: -$$ -\mathrm{Perplexity}(W) = \sqrt[N]{1\over P(W_1,...,W_N)} -$$ - -**Puedes experimentar con la generación de texto utilizando [el editor de texto impulsado por GPT de Hugging Face](https://transformer.huggingface.co/doc/gpt2-large)**. En este editor, comienzas a escribir tu texto, y al presionar **[TAB]** se te ofrecerán varias opciones de finalización. Si son demasiado cortas, o no estás satisfecho con ellas, presiona [TAB] nuevamente y tendrás más opciones, incluyendo fragmentos de texto más largos. - -## GPT es una Familia - -GPT no es un solo modelo, sino una colección de modelos desarrollados y entrenados por [OpenAI](https://openai.com). - -Bajo los modelos GPT, tenemos: - -| [GPT-2](https://huggingface.co/docs/transformers/model_doc/gpt2#openai-gpt2) | [GPT 3](https://openai.com/research/language-models-are-few-shot-learners) | [GPT-4](https://openai.com/gpt-4) | -| -- | -- | -- | -|Modelo de lenguaje con hasta 1.5 mil millones de parámetros. | Modelo de lenguaje con hasta 175 mil millones de parámetros | 100T parámetros y acepta tanto entradas de imagen como de texto y produce texto. | - -Los modelos GPT-3 y GPT-4 están disponibles [como un servicio cognitivo de Microsoft Azure](https://azure.microsoft.com/en-us/services/cognitive-services/openai-service/#overview?WT.mc_id=academic-77998-cacaste), y como [API de OpenAI](https://openai.com/api/). - -## Ingeniería de Prompts - -Debido a que GPT ha sido entrenado con grandes volúmenes de datos para entender el lenguaje y el código, proporciona salidas en respuesta a entradas (prompts). Los prompts son las entradas o consultas de GPT mediante las cuales se dan instrucciones a los modelos sobre las tareas que deben completar a continuación. Para obtener un resultado deseado, necesitas el prompt más efectivo, lo que implica seleccionar las palabras, formatos, frases o incluso símbolos correctos. Este enfoque se llama [Ingeniería de Prompts](https://learn.microsoft.com/en-us/shows/ai-show/the-basics-of-prompt-engineering-with-azure-openai-service?WT.mc_id=academic-77998-bethanycheum) - -[Esta documentación](https://learn.microsoft.com/en-us/semantic-kernel/prompt-engineering/?WT.mc_id=academic-77998-bethanycheum) te proporciona más información sobre la ingeniería de prompts. - -## ✍️ Cuaderno de Ejemplo: [Jugando con OpenAI-GPT](../../../../../lessons/5-NLP/20-LangModels/GPT-PyTorch.ipynb) - -Continúa tu aprendizaje en los siguientes cuadernos: - -* [Generando texto con OpenAI-GPT y Transformers de Hugging Face](../../../../../lessons/5-NLP/20-LangModels/GPT-PyTorch.ipynb) - -## Conclusión - -Los nuevos modelos de lenguaje preentrenados generales no solo modelan la estructura del lenguaje, sino que también contienen una gran cantidad de lenguaje natural. Por lo tanto, pueden ser utilizados de manera efectiva para resolver algunas tareas de PLN en configuraciones de cero disparos o pocos disparos. - -## [Cuestionario posterior a la conferencia](https://ff-quizzes.netlify.app/en/ai/quiz/40) - -**Descargo de responsabilidad**: -Este documento ha sido traducido utilizando servicios de traducción automática basados en IA. Si bien nos esforzamos por la precisión, tenga en cuenta que las traducciones automáticas pueden contener errores o inexactitudes. El documento original en su idioma nativo debe considerarse la fuente autorizada. Para información crítica, se recomienda una traducción profesional realizada por humanos. No nos hacemos responsables de malentendidos o malas interpretaciones que surjan del uso de esta traducción. \ No newline at end of file diff --git a/translations/es/lessons/5-NLP/README.md b/translations/es/lessons/5-NLP/README.md index dc7c5fd3..0df5ef9f 100644 --- a/translations/es/lessons/5-NLP/README.md +++ b/translations/es/lessons/5-NLP/README.md @@ -1,12 +1,3 @@ - # Procesamiento de Lenguaje Natural ![Resumen de tareas de PLN en un dibujo](../../../../lessons/sketchnotes/ai-nlp.png) diff --git a/translations/es/lessons/6-Other/21-GeneticAlgorithms/README.md b/translations/es/lessons/6-Other/21-GeneticAlgorithms/README.md index a2d0a2f0..0c8066fc 100644 --- a/translations/es/lessons/6-Other/21-GeneticAlgorithms/README.md +++ b/translations/es/lessons/6-Other/21-GeneticAlgorithms/README.md @@ -1,12 +1,3 @@ - # Algoritmos Genéticos ## [Cuestionario previo a la clase](https://ff-quizzes.netlify.app/en/ai/quiz/41) diff --git a/translations/es/lessons/6-Other/22-DeepRL/README.md b/translations/es/lessons/6-Other/22-DeepRL/README.md index 43d1960a..46ae8ca8 100644 --- a/translations/es/lessons/6-Other/22-DeepRL/README.md +++ b/translations/es/lessons/6-Other/22-DeepRL/README.md @@ -1,12 +1,3 @@ - # Aprendizaje por Refuerzo Profundo El aprendizaje por refuerzo (RL, por sus siglas en inglés) se considera uno de los paradigmas básicos del aprendizaje automático, junto con el aprendizaje supervisado y el aprendizaje no supervisado. Mientras que en el aprendizaje supervisado dependemos de un conjunto de datos con resultados conocidos, el RL se basa en **aprender haciendo**. Por ejemplo, cuando vemos un videojuego por primera vez, comenzamos a jugar sin conocer las reglas, y pronto somos capaces de mejorar nuestras habilidades simplemente jugando y ajustando nuestro comportamiento. @@ -34,7 +25,7 @@ Probablemente todos han visto dispositivos modernos de equilibrio como el *Segwa Una versión simplificada del equilibrio se conoce como el problema de **CartPole**. En el mundo de CartPole, tenemos un deslizador horizontal que puede moverse hacia la izquierda o hacia la derecha, y el objetivo es equilibrar un poste vertical sobre el deslizador mientras se mueve. -un cartpole +un cartpole Para crear y usar este entorno, necesitamos unas pocas líneas de código en Python: diff --git a/translations/es/lessons/6-Other/22-DeepRL/lab/README.md b/translations/es/lessons/6-Other/22-DeepRL/lab/README.md index 4d415935..c0d1fba4 100644 --- a/translations/es/lessons/6-Other/22-DeepRL/lab/README.md +++ b/translations/es/lessons/6-Other/22-DeepRL/lab/README.md @@ -1,12 +1,3 @@ - ## El Entorno El entorno Mountain Car consiste en un coche atrapado dentro de un valle. Tu objetivo es salir del valle y alcanzar la bandera. Las acciones que puedes realizar son acelerar hacia la izquierda, hacia la derecha o no hacer nada. Puedes observar la posición del coche a lo largo del eje x y su velocidad. diff --git a/translations/es/lessons/6-Other/23-MultiagentSystems/README.md b/translations/es/lessons/6-Other/23-MultiagentSystems/README.md index f14924ea..7781c62e 100644 --- a/translations/es/lessons/6-Other/23-MultiagentSystems/README.md +++ b/translations/es/lessons/6-Other/23-MultiagentSystems/README.md @@ -1,12 +1,3 @@ - # Sistemas Multi-Agente Una de las posibles formas de lograr inteligencia es el enfoque llamado **emergente** (o **sinérgico**), que se basa en el hecho de que el comportamiento combinado de muchos agentes relativamente simples puede resultar en un comportamiento general más complejo (o inteligente) del sistema en su conjunto. Teóricamente, esto se basa en los principios de [Inteligencia Colectiva](https://en.wikipedia.org/wiki/Collective_intelligence), [Emergentismo](https://en.wikipedia.org/wiki/Global_brain) y [Cibernética Evolutiva](https://en.wikipedia.org/wiki/Global_brain), que afirman que los sistemas de nivel superior obtienen algún tipo de valor añadido cuando se combinan adecuadamente a partir de sistemas de nivel inferior (el llamado *principio de transición de metasistema*). @@ -60,7 +51,7 @@ Puedes [descargar](https://ccl.northwestern.edu/netlogo/download.shtml) e instal Una gran ventaja de NetLogo es que contiene una biblioteca de modelos funcionales que puedes probar. Ve a **File → Models Library**, y tendrás muchas categorías de modelos para elegir. -Biblioteca de Modelos de NetLogo +Biblioteca de Modelos de NetLogo > Una captura de pantalla de la biblioteca de modelos por Dmitry Soshnikov diff --git a/translations/es/lessons/6-Other/23-MultiagentSystems/assignment.md b/translations/es/lessons/6-Other/23-MultiagentSystems/assignment.md index e9b205c8..7d10d3b2 100644 --- a/translations/es/lessons/6-Other/23-MultiagentSystems/assignment.md +++ b/translations/es/lessons/6-Other/23-MultiagentSystems/assignment.md @@ -1,12 +1,3 @@ - # Tarea de NetLogo Toma uno de los modelos de la biblioteca de NetLogo y úsalo para simular una situación de la vida real lo más fielmente posible. Un buen ejemplo sería ajustar el modelo Virus en la carpeta de Visualizaciones Alternativas para mostrar cómo puede usarse para modelar la propagación del COVID-19. ¿Puedes construir un modelo que imite la propagación viral en la vida real? diff --git a/translations/es/lessons/7-Ethics/README.md b/translations/es/lessons/7-Ethics/README.md index 2ba93fd2..7971e44f 100644 --- a/translations/es/lessons/7-Ethics/README.md +++ b/translations/es/lessons/7-Ethics/README.md @@ -1,12 +1,3 @@ - # IA Ética y Responsable Has casi terminado este curso, y espero que para ahora veas claramente que la IA se basa en una serie de métodos matemáticos formales que nos permiten encontrar relaciones en los datos y entrenar modelos para replicar algunos aspectos del comportamiento humano. En este momento de la historia, consideramos que la IA es una herramienta muy poderosa para extraer patrones de los datos y aplicar esos patrones para resolver nuevos problemas. diff --git a/translations/es/lessons/README.md b/translations/es/lessons/README.md index d53f8e33..9da466bf 100644 --- a/translations/es/lessons/README.md +++ b/translations/es/lessons/README.md @@ -1,12 +1,3 @@ - # Descripción general ![Descripción general en un dibujo](../../../lessons/sketchnotes/ai-overview.png) diff --git a/translations/es/lessons/X-Extras/X1-MultiModal/README.md b/translations/es/lessons/X-Extras/X1-MultiModal/README.md index 098e9f48..564a9b22 100644 --- a/translations/es/lessons/X-Extras/X1-MultiModal/README.md +++ b/translations/es/lessons/X-Extras/X1-MultiModal/README.md @@ -1,12 +1,3 @@ - # Redes Multi-Modales Tras el éxito de los modelos transformadores para resolver tareas de procesamiento de lenguaje natural (NLP), se han aplicado arquitecturas similares a tareas de visión por computadora. Existe un creciente interés en construir modelos que *combinen* capacidades de visión y lenguaje natural. Uno de estos intentos fue realizado por OpenAI, y se llama CLIP y DALL.E. diff --git a/translations/es/lessons/sketchnotes/LICENSE.md b/translations/es/lessons/sketchnotes/LICENSE.md index 7240dd90..643baf8d 100644 --- a/translations/es/lessons/sketchnotes/LICENSE.md +++ b/translations/es/lessons/sketchnotes/LICENSE.md @@ -1,12 +1,3 @@ - Attribution-ShareAlike 4.0 Internacional ======================================================================= diff --git a/translations/es/lessons/sketchnotes/README.md b/translations/es/lessons/sketchnotes/README.md index 583b79e6..42da6496 100644 --- a/translations/es/lessons/sketchnotes/README.md +++ b/translations/es/lessons/sketchnotes/README.md @@ -1,12 +1,3 @@ - Todas las notas visuales del plan de estudios se pueden descargar aquí. 🎨 Creado por: Tomomi Imura (Twitter: [@girlie_mac](https://twitter.com/girlie_mac), GitHub: [girliemac](https://github.com/girliemac)) diff --git a/translations/es/troubleshoot.md b/translations/es/troubleshoot.md index 6b4504ad..3e9ed5d2 100644 --- a/translations/es/troubleshoot.md +++ b/translations/es/troubleshoot.md @@ -1,12 +1,3 @@ - # Guía de Solución de Problemas para AI-For-Beginners Esta guía te ayuda a resolver problemas comunes al usar o contribuir al repositorio [AI-For-Beginners](https://github.com/microsoft/AI-For-Beginners). Cada problema incluye antecedentes, síntomas, explicaciones y soluciones paso a paso. diff --git a/translations/fr/.co-op-translator.json b/translations/fr/.co-op-translator.json new file mode 100644 index 00000000..b6a83894 --- /dev/null +++ b/translations/fr/.co-op-translator.json @@ -0,0 +1,398 @@ +{ + "AGENTS.md": { + "original_hash": "6b11a37115944252ab3ed04e358d830d", + "translation_date": "2025-10-03T09:10:55+00:00", + "source_file": "AGENTS.md", + "language_code": "fr" + }, + "README.md": { + "original_hash": "1984fc89dd304a8a33ab5584691a99aa", + "translation_date": "2026-01-30T00:59:31+00:00", + "source_file": "README.md", + "language_code": "fr" + }, + "SECURITY.md": { + "original_hash": "a583f49d359c7ebba61433e4dfcd05a9", + "translation_date": "2025-08-24T20:44:45+00:00", + "source_file": "SECURITY.md", + "language_code": "fr" + }, + "etc/CODE_OF_CONDUCT.md": { + "original_hash": "c06b12caf3c901eb3156e3dd5b0aea56", + "translation_date": "2025-08-24T21:00:13+00:00", + "source_file": "etc/CODE_OF_CONDUCT.md", + "language_code": "fr" + }, + "etc/CONTRIBUTING.md": { + "original_hash": "847a587aa1b83f4d00858183ff3ed18a", + "translation_date": "2025-08-24T21:00:29+00:00", + "source_file": "etc/CONTRIBUTING.md", + "language_code": "fr" + }, + "etc/Mindmap.md": { + "original_hash": "f2f88dbd2debd38e26149b27b1fd272d", + "translation_date": "2025-08-24T21:00:36+00:00", + "source_file": "etc/Mindmap.md", + "language_code": "fr" + }, + "etc/SUPPORT.md": { + "original_hash": "fdfc08baee91e402938a2b1f94fe0949", + "translation_date": "2025-08-24T21:00:07+00:00", + "source_file": "etc/SUPPORT.md", + "language_code": "fr" + }, + "etc/TRANSLATIONS.md": { + "original_hash": "62b3e3ad5182edb905eec649a87eeeb4", + "translation_date": "2025-08-24T21:00:20+00:00", + "source_file": "etc/TRANSLATIONS.md", + "language_code": "fr" + }, + "etc/quiz-app/README.md": { + "original_hash": "d699cf8509f74baa5b0b838de5cf0662", + "translation_date": "2025-08-24T21:01:05+00:00", + "source_file": "etc/quiz-app/README.md", + "language_code": "fr" + }, + "examples/README.md": { + "original_hash": "0d1babfdcbeb46525f2db3fbaaa54cd7", + "translation_date": "2025-10-03T11:26:04+00:00", + "source_file": "examples/README.md", + "language_code": "fr" + }, + "lessons/0-course-setup/for-teachers.md": { + "original_hash": "a094ef9927883de1cfcee51dbd143381", + "translation_date": "2025-08-24T20:59:47+00:00", + "source_file": "lessons/0-course-setup/for-teachers.md", + "language_code": "fr" + }, + "lessons/0-course-setup/how-to-run.md": { + "original_hash": "a4717bd9103b9f6cd84d534b83534689", + "translation_date": "2026-01-15T10:42:15+00:00", + "source_file": "lessons/0-course-setup/how-to-run.md", + "language_code": "fr" + }, + "lessons/0-course-setup/setup.md": { + "original_hash": "7b4e5b8956915870d0a0ed3cc5890042", + "translation_date": "2025-12-12T19:05:38+00:00", + "source_file": "lessons/0-course-setup/setup.md", + "language_code": "fr" + }, + "lessons/1-Intro/README.md": { + "original_hash": "f57e8aa46141fd220b16ffed8f11aec7", + "translation_date": "2025-11-18T20:30:45+00:00", + "source_file": "lessons/1-Intro/README.md", + "language_code": "fr" + }, + "lessons/1-Intro/assignment.md": { + "original_hash": "a334df77a82aaaf2a29c77065d3e481e", + "translation_date": "2025-11-18T20:31:41+00:00", + "source_file": "lessons/1-Intro/assignment.md", + "language_code": "fr" + }, + "lessons/2-Symbolic/README.md": { + "original_hash": "f9f06b266b8b2bfc6b8792ff2bb1bea4", + "translation_date": "2026-01-15T10:43:40+00:00", + "source_file": "lessons/2-Symbolic/README.md", + "language_code": "fr" + }, + "lessons/2-Symbolic/assignment.md": { + "original_hash": "a057a8604f3976c3e309884453f1fad0", + "translation_date": "2025-08-24T20:59:26+00:00", + "source_file": "lessons/2-Symbolic/assignment.md", + "language_code": "fr" + }, + "lessons/3-NeuralNetworks/03-Perceptron/README.md": { + "original_hash": "c34cbba802058b6fa267e1a294d4e510", + "translation_date": "2025-09-23T12:02:02+00:00", + "source_file": "lessons/3-NeuralNetworks/03-Perceptron/README.md", + "language_code": "fr" + }, + "lessons/3-NeuralNetworks/03-Perceptron/lab/README.md": { + "original_hash": "ba5d1eb353d20d3e7181066b3c424b99", + "translation_date": "2025-08-31T14:20:13+00:00", + "source_file": "lessons/3-NeuralNetworks/03-Perceptron/lab/README.md", + "language_code": "fr" + }, + "lessons/3-NeuralNetworks/04-OwnFramework/README.md": { + "original_hash": "789d6c3fb6fc7948a470b33078a5983a", + "translation_date": "2025-09-23T12:01:37+00:00", + "source_file": "lessons/3-NeuralNetworks/04-OwnFramework/README.md", + "language_code": "fr" + }, + "lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md": { + "original_hash": "48fdd704d483e19bc3d7464074c9fcbe", + "translation_date": "2025-08-24T20:55:59+00:00", + "source_file": "lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md", + "language_code": "fr" + }, + "lessons/3-NeuralNetworks/05-Frameworks/README.md": { + "original_hash": "ddd216f558a255260a9374008002c971", + "translation_date": "2025-09-23T12:02:17+00:00", + "source_file": "lessons/3-NeuralNetworks/05-Frameworks/README.md", + "language_code": "fr" + }, + "lessons/3-NeuralNetworks/05-Frameworks/lab/README.md": { + "original_hash": "e452d897efb9a89700f41021834cf6e5", + "translation_date": "2025-08-24T20:56:34+00:00", + "source_file": "lessons/3-NeuralNetworks/05-Frameworks/lab/README.md", + "language_code": "fr" + }, + "lessons/3-NeuralNetworks/README.md": { + "original_hash": "f862a99d88088163df12270e2f2ad6c3", + "translation_date": "2025-10-03T12:41:07+00:00", + "source_file": "lessons/3-NeuralNetworks/README.md", + "language_code": "fr" + }, + "lessons/4-ComputerVision/06-IntroCV/README.md": { + "original_hash": "feeca98225cb420afc89415f24f63d92", + "translation_date": "2025-09-23T11:56:49+00:00", + "source_file": "lessons/4-ComputerVision/06-IntroCV/README.md", + "language_code": "fr" + }, + "lessons/4-ComputerVision/06-IntroCV/lab/README.md": { + "original_hash": "3d53d6409f80970f7281a45dee35328a", + "translation_date": "2025-08-24T20:52:35+00:00", + "source_file": "lessons/4-ComputerVision/06-IntroCV/lab/README.md", + "language_code": "fr" + }, + "lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md": { + "original_hash": "53faab85adfcebd8c10bcd71dc2fa557", + "translation_date": "2025-09-23T11:56:12+00:00", + "source_file": "lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md", + "language_code": "fr" + }, + "lessons/4-ComputerVision/07-ConvNets/README.md": { + "original_hash": "a560d5b845962cf33dc102266e409568", + "translation_date": "2025-09-23T11:55:55+00:00", + "source_file": "lessons/4-ComputerVision/07-ConvNets/README.md", + "language_code": "fr" + }, + "lessons/4-ComputerVision/07-ConvNets/lab/README.md": { + "original_hash": "b70fcf7fcee862990f848c679090943f", + "translation_date": "2025-10-03T14:51:48+00:00", + "source_file": "lessons/4-ComputerVision/07-ConvNets/lab/README.md", + "language_code": "fr" + }, + "lessons/4-ComputerVision/08-TransferLearning/README.md": { + "original_hash": "178c0b5ee5395733eb18aec51e71a0a9", + "translation_date": "2025-09-23T11:56:29+00:00", + "source_file": "lessons/4-ComputerVision/08-TransferLearning/README.md", + "language_code": "fr" + }, + "lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md": { + "original_hash": "ae074cd940fc2f4dc24fc07b66ccbd99", + "translation_date": "2025-08-24T20:52:58+00:00", + "source_file": "lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md", + "language_code": "fr" + }, + "lessons/4-ComputerVision/08-TransferLearning/lab/README.md": { + "original_hash": "7765935c35fcee69b9fe2d0cfd6963e2", + "translation_date": "2025-08-24T20:53:21+00:00", + "source_file": "lessons/4-ComputerVision/08-TransferLearning/lab/README.md", + "language_code": "fr" + }, + "lessons/4-ComputerVision/09-Autoencoders/README.md": { + "original_hash": "1b8d9e1b3a6f1daa864b1ff3dfc3076d", + "translation_date": "2025-09-23T11:57:45+00:00", + "source_file": "lessons/4-ComputerVision/09-Autoencoders/README.md", + "language_code": "fr" + }, + "lessons/4-ComputerVision/10-GANs/README.md": { + "original_hash": "0ff65b4da07b23697235de2beb2a3c25", + "translation_date": "2025-09-23T11:59:37+00:00", + "source_file": "lessons/4-ComputerVision/10-GANs/README.md", + "language_code": "fr" + }, + "lessons/4-ComputerVision/11-ObjectDetection/README.md": { + "original_hash": "d76a7eda28de5210c8b1ba50a6216c69", + "translation_date": "2025-09-23T11:57:14+00:00", + "source_file": "lessons/4-ComputerVision/11-ObjectDetection/README.md", + "language_code": "fr" + }, + "lessons/4-ComputerVision/11-ObjectDetection/lab/README.md": { + "original_hash": "ad568d55ae65c856fe929fc2b278510a", + "translation_date": "2025-08-24T20:51:20+00:00", + "source_file": "lessons/4-ComputerVision/11-ObjectDetection/lab/README.md", + "language_code": "fr" + }, + "lessons/4-ComputerVision/12-Segmentation/README.md": { + "original_hash": "6568aaae7e0e4afed4b5d74b5b223700", + "translation_date": "2025-09-23T11:59:24+00:00", + "source_file": "lessons/4-ComputerVision/12-Segmentation/README.md", + "language_code": "fr" + }, + "lessons/4-ComputerVision/12-Segmentation/lab/README.md": { + "original_hash": "365f0decfe0f47b460bbde8227c5009d", + "translation_date": "2025-08-24T20:50:12+00:00", + "source_file": "lessons/4-ComputerVision/12-Segmentation/lab/README.md", + "language_code": "fr" + }, + "lessons/4-ComputerVision/README.md": { + "original_hash": "58a52f000089c1d8906a4daa4ab1169b", + "translation_date": "2025-08-24T20:49:24+00:00", + "source_file": "lessons/4-ComputerVision/README.md", + "language_code": "fr" + }, + "lessons/5-NLP/13-TextRep/README.md": { + "original_hash": "dbd3f73e4139f030ecb2e20387d70fee", + "translation_date": "2025-09-23T12:05:13+00:00", + "source_file": "lessons/5-NLP/13-TextRep/README.md", + "language_code": "fr" + }, + "lessons/5-NLP/13-TextRep/assignment.md": { + "original_hash": "cdc1f2e631f055f3473b36d18e4760b3", + "translation_date": "2025-08-24T20:47:34+00:00", + "source_file": "lessons/5-NLP/13-TextRep/assignment.md", + "language_code": "fr" + }, + "lessons/5-NLP/14-Embeddings/README.md": { + "original_hash": "b708c9b85b833864c73c6281f1e6b96e", + "translation_date": "2025-09-23T12:04:57+00:00", + "source_file": "lessons/5-NLP/14-Embeddings/README.md", + "language_code": "fr" + }, + "lessons/5-NLP/14-Embeddings/assignment.md": { + "original_hash": "bc690ecf68b38d311cc9e12f3144a28c", + "translation_date": "2025-08-24T20:46:35+00:00", + "source_file": "lessons/5-NLP/14-Embeddings/assignment.md", + "language_code": "fr" + }, + "lessons/5-NLP/15-LanguageModeling/README.md": { + "original_hash": "7ba20f54a5bfcd6521018cdfb17c7c57", + "translation_date": "2025-09-23T12:03:15+00:00", + "source_file": "lessons/5-NLP/15-LanguageModeling/README.md", + "language_code": "fr" + }, + "lessons/5-NLP/15-LanguageModeling/lab/README.md": { + "original_hash": "5130f01fdc5ebb83032b23d489027aac", + "translation_date": "2025-08-24T20:47:51+00:00", + "source_file": "lessons/5-NLP/15-LanguageModeling/lab/README.md", + "language_code": "fr" + }, + "lessons/5-NLP/16-RNN/README.md": { + "original_hash": "e2273cc150380a5e191903cea858f021", + "translation_date": "2025-09-23T12:04:29+00:00", + "source_file": "lessons/5-NLP/16-RNN/README.md", + "language_code": "fr" + }, + "lessons/5-NLP/16-RNN/assignment.md": { + "original_hash": "47f7d3c6a5373543e051e4d1140ce898", + "translation_date": "2025-08-24T20:46:10+00:00", + "source_file": "lessons/5-NLP/16-RNN/assignment.md", + "language_code": "fr" + }, + "lessons/5-NLP/17-GenerativeNetworks/README.md": { + "original_hash": "51be6057374d01d70e07dd5ec88ebc0d", + "translation_date": "2025-09-23T12:02:46+00:00", + "source_file": "lessons/5-NLP/17-GenerativeNetworks/README.md", + "language_code": "fr" + }, + "lessons/5-NLP/17-GenerativeNetworks/lab/README.md": { + "original_hash": "439e12796197a90e7623d4c9c057b9c2", + "translation_date": "2025-08-24T20:47:04+00:00", + "source_file": "lessons/5-NLP/17-GenerativeNetworks/lab/README.md", + "language_code": "fr" + }, + "lessons/5-NLP/18-Transformers/README.md": { + "original_hash": "f335dfcb4a993920504c387973a36957", + "translation_date": "2025-09-23T12:03:26+00:00", + "source_file": "lessons/5-NLP/18-Transformers/README.md", + "language_code": "fr" + }, + "lessons/5-NLP/18-Transformers/assignment.md": { + "original_hash": "177f3ea3995d725e6f9f5c66af16edcd", + "translation_date": "2025-08-24T20:47:59+00:00", + "source_file": "lessons/5-NLP/18-Transformers/assignment.md", + "language_code": "fr" + }, + "lessons/5-NLP/19-NER/README.md": { + "original_hash": "6522312ff835796ca34136a9462fafb2", + "translation_date": "2025-09-23T12:04:16+00:00", + "source_file": "lessons/5-NLP/19-NER/README.md", + "language_code": "fr" + }, + "lessons/5-NLP/19-NER/lab/README.md": { + "original_hash": "032bda5068f543d6c1fcb30c34231461", + "translation_date": "2025-08-24T20:48:23+00:00", + "source_file": "lessons/5-NLP/19-NER/lab/README.md", + "language_code": "fr" + }, + "lessons/5-NLP/20-LangModels/README.md": { + "original_hash": "97836d30a6bec736f8e3b4411c572bc2", + "translation_date": "2025-09-23T12:03:56+00:00", + "source_file": "lessons/5-NLP/20-LangModels/README.md", + "language_code": "fr" + }, + "lessons/5-NLP/README.md": { + "original_hash": "8ef02a9318257ea140ed3ed74442096d", + "translation_date": "2025-08-24T20:45:19+00:00", + "source_file": "lessons/5-NLP/README.md", + "language_code": "fr" + }, + "lessons/6-Other/21-GeneticAlgorithms/README.md": { + "original_hash": "6bbd632dfe6c62e5f66bb51fd78c174a", + "translation_date": "2025-09-23T11:54:29+00:00", + "source_file": "lessons/6-Other/21-GeneticAlgorithms/README.md", + "language_code": "fr" + }, + "lessons/6-Other/22-DeepRL/README.md": { + "original_hash": "04395657fc01648f8f70484d0e55ab67", + "translation_date": "2025-09-23T11:55:29+00:00", + "source_file": "lessons/6-Other/22-DeepRL/README.md", + "language_code": "fr" + }, + "lessons/6-Other/22-DeepRL/lab/README.md": { + "original_hash": "7bd8dc72040e98e35e7225e34058cd4e", + "translation_date": "2025-08-24T20:54:55+00:00", + "source_file": "lessons/6-Other/22-DeepRL/lab/README.md", + "language_code": "fr" + }, + "lessons/6-Other/23-MultiagentSystems/README.md": { + "original_hash": "38a1185ae3d54b180378bbd71ae3ef16", + "translation_date": "2025-09-23T11:54:54+00:00", + "source_file": "lessons/6-Other/23-MultiagentSystems/README.md", + "language_code": "fr" + }, + "lessons/6-Other/23-MultiagentSystems/assignment.md": { + "original_hash": "cf654ca60c7f86c8dad28596fb42994b", + "translation_date": "2025-08-24T20:54:29+00:00", + "source_file": "lessons/6-Other/23-MultiagentSystems/assignment.md", + "language_code": "fr" + }, + "lessons/7-Ethics/README.md": { + "original_hash": "437c988596e751072e41a5aad3fcc5d9", + "translation_date": "2025-08-24T20:45:04+00:00", + "source_file": "lessons/7-Ethics/README.md", + "language_code": "fr" + }, + "lessons/README.md": { + "original_hash": "5fef1a0b22498d7188959e2a2cb08af7", + "translation_date": "2025-08-24T20:44:59+00:00", + "source_file": "lessons/README.md", + "language_code": "fr" + }, + "lessons/X-Extras/X1-MultiModal/README.md": { + "original_hash": "9c592c26aca16ca085d268c732284187", + "translation_date": "2025-08-24T20:55:01+00:00", + "source_file": "lessons/X-Extras/X1-MultiModal/README.md", + "language_code": "fr" + }, + "lessons/sketchnotes/LICENSE.md": { + "original_hash": "45ab63a2cd8f5faef6c9b150618837a4", + "translation_date": "2025-08-24T20:57:48+00:00", + "source_file": "lessons/sketchnotes/LICENSE.md", + "language_code": "fr" + }, + "lessons/sketchnotes/README.md": { + "original_hash": "050b8bddebafba55b129414e6ab096ab", + "translation_date": "2025-08-24T20:57:07+00:00", + "source_file": "lessons/sketchnotes/README.md", + "language_code": "fr" + }, + "troubleshoot.md": { + "original_hash": "8d9c5a4a7c7798d699672a22cb7fea86", + "translation_date": "2025-10-03T09:33:51+00:00", + "source_file": "troubleshoot.md", + "language_code": "fr" + } +} \ No newline at end of file diff --git a/translations/fr/AGENTS.md b/translations/fr/AGENTS.md index d237a0d8..0700f5f6 100644 --- a/translations/fr/AGENTS.md +++ b/translations/fr/AGENTS.md @@ -1,12 +1,3 @@ - # AGENTS.md ## Aperçu du projet diff --git a/translations/fr/README.md b/translations/fr/README.md index d0ddbd29..fdf5a14c 100644 --- a/translations/fr/README.md +++ b/translations/fr/README.md @@ -1,12 +1,3 @@ - [![GitHub license](https://img.shields.io/github/license/microsoft/AI-For-Beginners.svg)](https://github.com/microsoft/AI-For-Beginners/blob/main/LICENSE) [![GitHub contributors](https://img.shields.io/github/contributors/microsoft/AI-For-Beginners.svg)](https://GitHub.com/microsoft/AI-For-Beginners/graphs/contributors/) [![GitHub issues](https://img.shields.io/github/issues/microsoft/AI-For-Beginners.svg)](https://GitHub.com/microsoft/AI-For-Beginners/issues/) @@ -23,33 +14,34 @@ CO_OP_TRANSLATOR_METADATA: # Intelligence Artificielle pour Débutants - Un Curriculum -|![Sketchnote by @girlie_mac https://twitter.com/girlie_mac](../../../../translated_images/fr/ai-overview.0857791951d19500.webp)| +|![Sketchnote par @girlie_mac https://twitter.com/girlie_mac](../../translated_images/fr/ai-overview.0857791951d19500.webp)| |:---:| -| Intelligence Artificielle pour Débutants - _Sketchnote par [@girlie_mac](https://twitter.com/girlie_mac)_ | +| AI For Beginners - _Sketchnote par [@girlie_mac](https://twitter.com/girlie_mac)_ | -Explorez le monde de **l'Intelligence Artificielle** (IA) avec notre curriculum de 12 semaines et 24 leçons ! Il comprend des leçons pratiques, des quiz et des laboratoires. Le curriculum est adapté aux débutants et couvre des outils comme TensorFlow et PyTorch, ainsi que l'éthique en IA. +Explorez le monde de **l’Intelligence Artificielle** (IA) avec notre curriculum de 12 semaines et 24 leçons ! Il inclut des leçons pratiques, des quiz et des laboratoires. Le curriculum est adapté aux débutants et couvre des outils comme TensorFlow et PyTorch, ainsi que l’éthique en IA. -### 🌐 Support Multilingue + +### 🌐 Support multilingue #### Pris en charge via GitHub Action (Automatisé & Toujours à jour) -[Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh/README.md) | [Chinese (Traditional, Hong Kong)](../hk/README.md) | [Chinese (Traditional, Macau)](../mo/README.md) | [Chinese (Traditional, Taiwan)](../tw/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](./README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../br/README.md) | [Portuguese (Portugal)](../pt/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) +[Arabe](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgare](../bg/README.md) | [Birman (Myanmar)](../my/README.md) | [Chinois (simplifié)](../zh-CN/README.md) | [Chinois (traditionnel, Hong Kong)](../zh-HK/README.md) | [Chinois (traditionnel, Macau)](../zh-MO/README.md) | [Chinois (traditionnel, Taïwan)](../zh-TW/README.md) | [Croate](../hr/README.md) | [Tchèque](../cs/README.md) | [Danois](../da/README.md) | [Néerlandais](../nl/README.md) | [Estonien](../et/README.md) | [Finnois](../fi/README.md) | [Français](./README.md) | [Allemand](../de/README.md) | [Grec](../el/README.md) | [Hébreu](../he/README.md) | [Hindi](../hi/README.md) | [Hongrois](../hu/README.md) | [Indonésien](../id/README.md) | [Italien](../it/README.md) | [Japonais](../ja/README.md) | [Kannada](../kn/README.md) | [Coréen](../ko/README.md) | [Lituanien](../lt/README.md) | [Malais](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Népalais](../ne/README.md) | [Pidgin nigérian](../pcm/README.md) | [Norvégien](../no/README.md) | [Persan (Farsi)](../fa/README.md) | [Polonais](../pl/README.md) | [Portugais (Brésil)](../pt-BR/README.md) | [Portugais (Portugal)](../pt-PT/README.md) | [Pendjabi (Gurmukhi)](../pa/README.md) | [Roumain](../ro/README.md) | [Russe](../ru/README.md) | [Serbe (cyrillique)](../sr/README.md) | [Slovaque](../sk/README.md) | [Slovène](../sl/README.md) | [Espagnol](../es/README.md) | [Swahili](../sw/README.md) | [Suédois](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamoul](../ta/README.md) | [Télougou](../te/README.md) | [Thaï](../th/README.md) | [Turc](../tr/README.md) | [Ukrainien](../uk/README.md) | [Ourdou](../ur/README.md) | [Vietnamien](../vi/README.md) -> **Préférez-vous cloner localement ?** +> **Vous préférez cloner localement ?** -> Ce référentiel inclut plus de 50 traductions linguistiques ce qui augmente significativement la taille du téléchargement. Pour cloner sans les traductions, utilisez le sparse checkout: +> Ce dépôt inclut plus de 50 traductions de langue ce qui augmente significativement la taille de téléchargement. Pour cloner sans traductions, utilisez le sparse checkout : > ```bash > git clone --filter=blob:none --sparse https://github.com/microsoft/AI-For-Beginners.git > cd AI-For-Beginners > git sparse-checkout set --no-cone '/*' '!translations' '!translated_images' > ``` -> Cela vous donne tout ce dont vous avez besoin pour compléter le cours avec un téléchargement beaucoup plus rapide. +> Cela vous donne tout ce dont vous avez besoin pour suivre le cours avec un téléchargement beaucoup plus rapide. -**Si vous souhaitez que d'autres langues soient prises en charge, elles sont listées [ici](https://github.com/Azure/co-op-translator/blob/main/getting_started/supported-languages.md)** +**Si vous souhaitez que d’autres langues soient prises en charge, elles sont listées [ici](https://github.com/Azure/co-op-translator/blob/main/getting_started/supported-languages.md)** -## Rejoignez la Communauté +## Rejoignez la communauté [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/nTYy5BXMWG)](https://discord.gg/nTYy5BXMWG) ## Ce que vous apprendrez @@ -58,21 +50,21 @@ Explorez le monde de **l'Intelligence Artificielle** (IA) avec notre curriculum Dans ce curriculum, vous apprendrez : -* Différentes approches de l'Intelligence Artificielle, incluant l'approche "classique" symbolique avec la **Représentation des Connaissances** et le raisonnement ([GOFAI](https://en.wikipedia.org/wiki/Symbolic_artificial_intelligence)). -* **Réseaux de Neurones** et **Apprentissage Profond**, qui sont au cœur de l'IA moderne. Nous illustrerons les concepts derrière ces sujets importants en utilisant du code dans deux des frameworks les plus populaires - [TensorFlow](http://Tensorflow.org) et [PyTorch](http://pytorch.org). -* **Architectures Neuronales** pour travailler avec des images et du texte. Nous couvrirons des modèles récents mais qui peuvent être un peu en retard sur l'état de l'art. -* Des approches moins populaires de l’IA, telles que les **Algorithmes Génétiques** et les **Systèmes Multi-Agents**. +* Différentes approches de l’Intelligence Artificielle, y compris la « bonne vieille » approche symbolique avec la **Représentation des connaissances** et le raisonnement ([GOFAI](https://en.wikipedia.org/wiki/Symbolic_artificial_intelligence)). +* **Réseaux neuronaux** et **Apprentissage profond**, qui sont au cœur de l’IA moderne. Nous illustrerons les concepts derrière ces sujets importants en utilisant du code dans deux des frameworks les plus populaires - [TensorFlow](http://Tensorflow.org) et [PyTorch](http://pytorch.org). +* **Architectures neuronales** pour travailler avec des images et du texte. Nous couvrirons des modèles récents bien que cela puisse être un peu en deçà de l’état de l’art. +* Des approches d’IA moins populaires, telles que les **Algorithmes génétiques** et les **Systèmes multi-agents**. Ce que nous ne couvrirons pas dans ce curriculum : > [Trouvez toutes les ressources supplémentaires pour ce cours dans notre collection Microsoft Learn](https://learn.microsoft.com/en-us/collections/7w28iy2xrqzdj0?WT.mc_id=academic-77998-bethanycheum) -* Les cas d'utilisation de l'**IA en entreprise**. Pensez à suivre le parcours d’apprentissage [Introduction à l’IA pour les utilisateurs métier](https://docs.microsoft.com/learn/paths/introduction-ai-for-business-users/?WT.mc_id=academic-77998-bethanycheum) sur Microsoft Learn, ou [AI Business School](https://www.microsoft.com/ai/ai-business-school/?WT.mc_id=academic-77998-bethanycheum), développé en coopération avec [INSEAD](https://www.insead.edu/). -* L'**Apprentissage Machine Classique**, qui est bien décrit dans notre [Curriculum Apprentissage Machine pour Débutants](http://github.com/Microsoft/ML-for-Beginners). -* Applications pratiques de l’IA construites en utilisant **[les Services Cognitifs](https://azure.microsoft.com/services/cognitive-services/?WT.mc_id=academic-77998-bethanycheum)**. Pour cela, nous recommandons de commencer par les modules Microsoft Learn pour la [vision](https://docs.microsoft.com/learn/paths/create-computer-vision-solutions-azure-cognitive-services/?WT.mc_id=academic-77998-bethanycheum), le [traitement du langage naturel](https://docs.microsoft.com/learn/paths/explore-natural-language-processing/?WT.mc_id=academic-77998-bethanycheum), **[IA Générative avec le service Azure OpenAI](https://learn.microsoft.com/en-us/training/paths/develop-ai-solutions-azure-openai/?WT.mc_id=academic-77998-bethanycheum)** et d’autres. -* **Frameworks Cloud spécifiques pour le ML**, tels que [Azure Machine Learning](https://azure.microsoft.com/services/machine-learning/?WT.mc_id=academic-77998-bethanycheum), [Microsoft Fabric](https://learn.microsoft.com/en-us/training/paths/get-started-fabric/?WT.mc_id=academic-77998-bethanycheum) ou [Azure Databricks](https://docs.microsoft.com/learn/paths/data-engineer-azure-databricks?WT.mc_id=academic-77998-bethanycheum). Pensez à consulter les parcours d’apprentissage [Construire et exploiter des solutions de machine learning avec Azure Machine Learning](https://docs.microsoft.com/learn/paths/build-ai-solutions-with-azure-ml-service/?WT.mc_id=academic-77998-bethanycheum) et [Construire et exploiter des solutions de Machine Learning avec Azure Databricks](https://docs.microsoft.com/learn/paths/build-operate-machine-learning-solutions-azure-databricks/?WT.mc_id=academic-77998-bethanycheum). -* **IA Conversationnelle** et **Chat Bots**. Il existe un parcours d’apprentissage distinct [Créer des solutions d’IA conversationnelle](https://docs.microsoft.com/learn/paths/create-conversational-ai-solutions/?WT.mc_id=academic-77998-bethanycheum) et vous pouvez aussi consulter [ce billet de blog](https://soshnikov.com/azure/hello-bot-conversational-ai-on-microsoft-platform/) pour plus de détails. -* **Mathématiques approfondies** derrière l’apprentissage profond. Pour cela, nous recommandons [Deep Learning](https://www.amazon.com/Deep-Learning-Adaptive-Computation-Machine/dp/0262035618) par Ian Goodfellow, Yoshua Bengio et Aaron Courville, également disponible en ligne à [https://www.deeplearningbook.org/](https://www.deeplearningbook.org/). +* Les cas d’usage de l’**IA en entreprise**. Envisagez de suivre le parcours d’apprentissage [Introduction à l’IA pour les utilisateurs métier](https://docs.microsoft.com/learn/paths/introduction-ai-for-business-users/?WT.mc_id=academic-77998-bethanycheum) sur Microsoft Learn, ou [AI Business School](https://www.microsoft.com/ai/ai-business-school/?WT.mc_id=academic-77998-bethanycheum), développé en coopération avec [INSEAD](https://www.insead.edu/). +* **Apprentissage automatique classique**, qui est bien décrit dans notre [curriculum Machine Learning pour débutants](http://github.com/Microsoft/ML-for-Beginners). +* Applications pratiques d’IA construites à l’aide des **[Services cognitifs](https://azure.microsoft.com/services/cognitive-services/?WT.mc_id=academic-77998-bethanycheum)**. Pour cela, nous recommandons de commencer par les modules Microsoft Learn pour la [vision](https://docs.microsoft.com/learn/paths/create-computer-vision-solutions-azure-cognitive-services/?WT.mc_id=academic-77998-bethanycheum), le [traitement du langage naturel](https://docs.microsoft.com/learn/paths/explore-natural-language-processing/?WT.mc_id=academic-77998-bethanycheum), **[IA générative avec Azure OpenAI Service](https://learn.microsoft.com/en-us/training/paths/develop-ai-solutions-azure-openai/?WT.mc_id=academic-77998-bethanycheum)** et autres. +* **Frameworks Cloud spécifiques au ML**, tels que [Azure Machine Learning](https://azure.microsoft.com/services/machine-learning/?WT.mc_id=academic-77998-bethanycheum), [Microsoft Fabric](https://learn.microsoft.com/en-us/training/paths/get-started-fabric/?WT.mc_id=academic-77998-bethanycheum), ou [Azure Databricks](https://docs.microsoft.com/learn/paths/data-engineer-azure-databricks?WT.mc_id=academic-77998-bethanycheum). Envisagez d’utiliser les parcours d’apprentissage [Construire et exploiter des solutions de machine learning avec Azure Machine Learning](https://docs.microsoft.com/learn/paths/build-ai-solutions-with-azure-ml-service/?WT.mc_id=academic-77998-bethanycheum) et [Construire et exploiter des solutions de machine learning avec Azure Databricks](https://docs.microsoft.com/learn/paths/build-operate-machine-learning-solutions-azure-databricks/?WT.mc_id=academic-77998-bethanycheum). +* **IA conversationnelle** et **chatbots**. Il existe un parcours d’apprentissage séparé [Créer des solutions d’IA conversationnelle](https://docs.microsoft.com/learn/paths/create-conversational-ai-solutions/?WT.mc_id=academic-77998-bethanycheum), et vous pouvez également vous référer à [cet article de blog](https://soshnikov.com/azure/hello-bot-conversational-ai-on-microsoft-platform/) pour plus de détails. +* **Mathématiques approfondies** derrière l’apprentissage profond. Pour cela, nous recommanderions [Deep Learning](https://www.amazon.com/Deep-Learning-Adaptive-Computation-Machine/dp/0262035618) par Ian Goodfellow, Yoshua Bengio et Aaron Courville, également disponible en ligne sur [https://www.deeplearningbook.org/](https://www.deeplearningbook.org/). Pour une introduction douce aux sujets _IA dans le Cloud_, vous pouvez envisager de suivre le parcours d’apprentissage [Commencer avec l’intelligence artificielle sur Azure](https://docs.microsoft.com/learn/paths/get-started-with-artificial-intelligence-on-azure/?WT.mc_id=academic-77998-bethanycheum). @@ -80,147 +72,148 @@ Pour une introduction douce aux sujets _IA dans le Cloud_, vous pouvez envisager | | Lien de la leçon | PyTorch/Keras/TensorFlow | Laboratoire | | :-: | :------------------------------------------------------------------------------------------------------------------------------------------: | :---------------------------------------------------------------------------------------------: | ------------------------------------------------------------------------------ | -| 0 | [Configuration du cours](./lessons/0-course-setup/setup.md) | [Préparez votre environnement de développement](./lessons/0-course-setup/how-to-run.md) | | -| I | [**Introduction à l'IA**](./lessons/1-Intro/README.md) | | | -| 01 | [Introduction et histoire de l'IA](./lessons/1-Intro/README.md) | - | - | -| II | **IA Symbolique** | -| 02 | [Représentation des connaissances et systèmes experts](./lessons/2-Symbolic/README.md) | [Systèmes Experts](./lessons/2-Symbolic/Animals.ipynb) / [Ontologie](./lessons/2-Symbolic/FamilyOntology.ipynb) /[Graphe de Concepts](./lessons/2-Symbolic/MSConceptGraph.ipynb) | | -| III | [**Introduction aux Réseaux de Neurones**](./lessons/3-NeuralNetworks/README.md) ||| -| 03 | [Perceptron](./lessons/3-NeuralNetworks/03-Perceptron/README.md) | [Notebook](./lessons/3-NeuralNetworks/03-Perceptron/Perceptron.ipynb) | [Lab](./lessons/3-NeuralNetworks/03-Perceptron/lab/README.md) | -| 04 | [Perceptron Multi-couches et Création de notre propre Framework](./lessons/3-NeuralNetworks/04-OwnFramework/README.md) | [Notebook](./lessons/3-NeuralNetworks/04-OwnFramework/OwnFramework.ipynb) | [Lab](./lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md) | -| 05 | [Introduction aux Frameworks (PyTorch/TensorFlow) et Surapprentissage](./lessons/3-NeuralNetworks/05-Frameworks/README.md) | [PyTorch](./lessons/3-NeuralNetworks/05-Frameworks/IntroPyTorch.ipynb) / [Keras](./lessons/3-NeuralNetworks/05-Frameworks/IntroKeras.ipynb) / [TensorFlow](./lessons/3-NeuralNetworks/05-Frameworks/IntroKerasTF.ipynb) | [Lab](./lessons/3-NeuralNetworks/05-Frameworks/lab/README.md) | -| IV | [**Vision par ordinateur**](./lessons/4-ComputerVision/README.md) | [PyTorch](https://docs.microsoft.com/learn/modules/intro-computer-vision-pytorch/?WT.mc_id=academic-77998-cacaste) / [TensorFlow](https://docs.microsoft.com/learn/modules/intro-computer-vision-TensorFlow/?WT.mc_id=academic-77998-cacaste)| [Explore Computer Vision on Microsoft Azure](https://learn.microsoft.com/en-us/collections/7w28iy2xrqzdj0?WT.mc_id=academic-77998-bethanycheum) | -| 06 | [Introduction à la Vision par Ordinateur. OpenCV](./lessons/4-ComputerVision/06-IntroCV/README.md) | [Notebook](./lessons/4-ComputerVision/06-IntroCV/OpenCV.ipynb) | [Lab](./lessons/4-ComputerVision/06-IntroCV/lab/README.md) | -| 07 | [Réseaux de Neurones Convolutifs](./lessons/4-ComputerVision/07-ConvNets/README.md) & [Architectures CNN](./lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md) | [PyTorch](./lessons/4-ComputerVision/07-ConvNets/ConvNetsPyTorch.ipynb) /[TensorFlow](./lessons/4-ComputerVision/07-ConvNets/ConvNetsTF.ipynb) | [Lab](./lessons/4-ComputerVision/07-ConvNets/lab/README.md) | -| 08 | [Réseaux pré-entraînés et Apprentissage par Transfert](./lessons/4-ComputerVision/08-TransferLearning/README.md) et [Astuces d'entraînement](./lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md) | [PyTorch](./lessons/4-ComputerVision/08-TransferLearning/TransferLearningPyTorch.ipynb) / [TensorFlow](./lessons/3-NeuralNetworks/05-Frameworks/IntroKerasTF.ipynb) | [Lab](./lessons/4-ComputerVision/08-TransferLearning/lab/README.md) | +| 0 | [Installation du cours](./lessons/0-course-setup/setup.md) | [Configurer votre environnement de développement](./lessons/0-course-setup/how-to-run.md) | | +| I | [**Introduction à l’IA**](./lessons/1-Intro/README.md) | | | +| 01 | [Introduction et histoire de l’IA](./lessons/1-Intro/README.md) | - | - | +| II | **IA symbolique** | +| 02 | [Représentation des connaissances et systèmes experts](./lessons/2-Symbolic/README.md) | [Systèmes experts](./lessons/2-Symbolic/Animals.ipynb) / [Ontologie](./lessons/2-Symbolic/FamilyOntology.ipynb) /[Graphe de concepts](./lessons/2-Symbolic/MSConceptGraph.ipynb) | | +| III | [**Introduction aux réseaux neuronaux**](./lessons/3-NeuralNetworks/README.md) ||| +| 03 | [Perceptron](./lessons/3-NeuralNetworks/03-Perceptron/README.md) | [Carnet](./lessons/3-NeuralNetworks/03-Perceptron/Perceptron.ipynb) | [Laboratoire](./lessons/3-NeuralNetworks/03-Perceptron/lab/README.md) | +| 04 | [Perceptron multicouche et création de notre propre framework](./lessons/3-NeuralNetworks/04-OwnFramework/README.md) | [Carnet](./lessons/3-NeuralNetworks/04-OwnFramework/OwnFramework.ipynb) | [Laboratoire](./lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md) | +| 05 | [Introduction aux frameworks (PyTorch/TensorFlow) et sur-apprentissage](./lessons/3-NeuralNetworks/05-Frameworks/README.md) | [PyTorch](./lessons/3-NeuralNetworks/05-Frameworks/IntroPyTorch.ipynb) / [Keras](./lessons/3-NeuralNetworks/05-Frameworks/IntroKeras.ipynb) / [TensorFlow](./lessons/3-NeuralNetworks/05-Frameworks/IntroKerasTF.ipynb) | [Laboratoire](./lessons/3-NeuralNetworks/05-Frameworks/lab/README.md) | +| IV | [**Vision par ordinateur**](./lessons/4-ComputerVision/README.md) | [PyTorch](https://docs.microsoft.com/learn/modules/intro-computer-vision-pytorch/?WT.mc_id=academic-77998-cacaste) / [TensorFlow](https://docs.microsoft.com/learn/modules/intro-computer-vision-TensorFlow/?WT.mc_id=academic-77998-cacaste)| [Explorez la vision par ordinateur sur Microsoft Azure](https://learn.microsoft.com/en-us/collections/7w28iy2xrqzdj0?WT.mc_id=academic-77998-bethanycheum) | +| 06 | [Introduction à la vision par ordinateur. OpenCV](./lessons/4-ComputerVision/06-IntroCV/README.md) | [Carnet](./lessons/4-ComputerVision/06-IntroCV/OpenCV.ipynb) | [Laboratoire](./lessons/4-ComputerVision/06-IntroCV/lab/README.md) | +| 07 | [Réseaux neuronaux convolutifs](./lessons/4-ComputerVision/07-ConvNets/README.md) & [Architectures CNN](./lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md) | [PyTorch](./lessons/4-ComputerVision/07-ConvNets/ConvNetsPyTorch.ipynb) /[TensorFlow](./lessons/4-ComputerVision/07-ConvNets/ConvNetsTF.ipynb) | [Laboratoire](./lessons/4-ComputerVision/07-ConvNets/lab/README.md) | +| 08 | [Réseaux pré-entraînés et transfert d’apprentissage](./lessons/4-ComputerVision/08-TransferLearning/README.md) et [Astuces d’entraînement](./lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md) | [PyTorch](./lessons/4-ComputerVision/08-TransferLearning/TransferLearningPyTorch.ipynb) / [TensorFlow](./lessons/3-NeuralNetworks/05-Frameworks/IntroKerasTF.ipynb) | [Laboratoire](./lessons/4-ComputerVision/08-TransferLearning/lab/README.md) | | 09 | [Autoencodeurs et VAEs](./lessons/4-ComputerVision/09-Autoencoders/README.md) | [PyTorch](./lessons/4-ComputerVision/09-Autoencoders/AutoEncodersPyTorch.ipynb) / [TensorFlow](./lessons/4-ComputerVision/09-Autoencoders/AutoencodersTF.ipynb) | | -| 10 | [Réseaux Adverses Génératifs & Transfert de Style Artistique](./lessons/4-ComputerVision/10-GANs/README.md) | [PyTorch](./lessons/4-ComputerVision/10-GANs/GANPyTorch.ipynb) / [TensorFlow](./lessons/4-ComputerVision/10-GANs/GANTF.ipynb) | | -| 11 | [Détection d'Objets](./lessons/4-ComputerVision/11-ObjectDetection/README.md) | [TensorFlow](./lessons/4-ComputerVision/11-ObjectDetection/ObjectDetection.ipynb) | [Lab](./lessons/4-ComputerVision/11-ObjectDetection/lab/README.md) | -| 12 | [Segmentation Sémantique. U-Net](./lessons/4-ComputerVision/12-Segmentation/README.md) | [PyTorch](./lessons/4-ComputerVision/12-Segmentation/SemanticSegmentationPytorch.ipynb) / [TensorFlow](./lessons/4-ComputerVision/12-Segmentation/SemanticSegmentationTF.ipynb) | | -| V | [**Traitement du langage naturel**](./lessons/5-NLP/README.md) | [PyTorch](https://docs.microsoft.com/learn/modules/intro-natural-language-processing-pytorch/?WT.mc_id=academic-77998-cacaste) /[TensorFlow](https://docs.microsoft.com/learn/modules/intro-natural-language-processing-TensorFlow/?WT.mc_id=academic-77998-cacaste) | [Explore Natural Language Processing on Microsoft Azure](https://learn.microsoft.com/en-us/collections/7w28iy2xrqzdj0?WT.mc_id=academic-77998-bethanycheum)| -| 13 | [Représentation du texte. Sac de mots/TF-IDF](./lessons/5-NLP/13-TextRep/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/13-TextRep/TextRepresentationPyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/13-TextRep/TextRepresentationTF.ipynb) | | -| 14 | [Représentations sémantiques des mots. Word2Vec et GloVe](./lessons/5-NLP/14-Embeddings/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/14-Embeddings/EmbeddingsPyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/14-Embeddings/EmbeddingsTF.ipynb) | | -| 15 | [Modélisation du langage. Entraînement de vos propres embeddings](./lessons/5-NLP/15-LanguageModeling/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/15-LanguageModeling/CBoW-PyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/15-LanguageModeling/CBoW-TF.ipynb) | [Lab](./lessons/5-NLP/15-LanguageModeling/lab/README.md) | -| 16 | [Réseaux de Neurones Récurrents](./lessons/5-NLP/16-RNN/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/16-RNN/RNNPyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/16-RNN/RNNTF.ipynb) | | -| 17 | [Réseaux Récurrents Génératifs](./lessons/5-NLP/17-GenerativeNetworks/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/17-GenerativeNetworks/GenerativePyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/17-GenerativeNetworks/GenerativeTF.ipynb) | [Lab](./lessons/5-NLP/17-GenerativeNetworks/lab/README.md) | +| 10 | [Réseaux antagonistes génératifs & transfert de style artistique](./lessons/4-ComputerVision/10-GANs/README.md) | [PyTorch](./lessons/4-ComputerVision/10-GANs/GANPyTorch.ipynb) / [TensorFlow](./lessons/4-ComputerVision/10-GANs/GANTF.ipynb) | | +| 11 | [Détection d’objets](./lessons/4-ComputerVision/11-ObjectDetection/README.md) | [TensorFlow](./lessons/4-ComputerVision/11-ObjectDetection/ObjectDetection.ipynb) | [Laboratoire](./lessons/4-ComputerVision/11-ObjectDetection/lab/README.md) | +| 12 | [Segmentation sémantique. U-Net](./lessons/4-ComputerVision/12-Segmentation/README.md) | [PyTorch](./lessons/4-ComputerVision/12-Segmentation/SemanticSegmentationPytorch.ipynb) / [TensorFlow](./lessons/4-ComputerVision/12-Segmentation/SemanticSegmentationTF.ipynb) | | +| V | [**Traitement du langage naturel**](./lessons/5-NLP/README.md) | [PyTorch](https://docs.microsoft.com/learn/modules/intro-natural-language-processing-pytorch/?WT.mc_id=academic-77998-cacaste) /[TensorFlow](https://docs.microsoft.com/learn/modules/intro-natural-language-processing-TensorFlow/?WT.mc_id=academic-77998-cacaste) | [Explorez le traitement du langage naturel sur Microsoft Azure](https://learn.microsoft.com/en-us/collections/7w28iy2xrqzdj0?WT.mc_id=academic-77998-bethanycheum)| +| 13 | [Représentation du texte. Bow/TF-IDF](./lessons/5-NLP/13-TextRep/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/13-TextRep/TextRepresentationPyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/13-TextRep/TextRepresentationTF.ipynb) | | +| 14 | [Incorporations sémantiques de mots. Word2Vec et GloVe](./lessons/5-NLP/14-Embeddings/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/14-Embeddings/EmbeddingsPyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/14-Embeddings/EmbeddingsTF.ipynb) | | +| 15 | [Modélisation du langage. Entraînement de vos propres incorporations](./lessons/5-NLP/15-LanguageModeling/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/15-LanguageModeling/CBoW-PyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/15-LanguageModeling/CBoW-TF.ipynb) | [Laboratoire](./lessons/5-NLP/15-LanguageModeling/lab/README.md) | +| 16 | [Réseaux neuronaux récurrents](./lessons/5-NLP/16-RNN/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/16-RNN/RNNPyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/16-RNN/RNNTF.ipynb) | | +| 17 | [Réseaux récurrents génératifs](./lessons/5-NLP/17-GenerativeNetworks/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/17-GenerativeNetworks/GenerativePyTorch.ipynb) / [TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/17-GenerativeNetworks/GenerativeTF.ipynb) | [Laboratoire](./lessons/5-NLP/17-GenerativeNetworks/lab/README.md) | | 18 | [Transformers. BERT.](./lessons/5-NLP/18-Transformers/README.md) | [PyTorch](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/18-Transformers/TransformersPyTorch.ipynb) /[TensorFlow](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/5-NLP/18-Transformers/TransformersTF.ipynb) | | -| 19 | [Reconnaissance d'entités nommées](./lessons/5-NLP/19-NER/README.md) | [TensorFlow](https://microsoft.github.io/AI-For-Beginners/lessons/5-NLP/19-NER/NER-TF.ipynb) | [Lab](./lessons/5-NLP/19-NER/lab/README.md) | -| 20 | [Grands modèles de langage, programmation par prompt et tâches en few-shot](./lessons/5-NLP/20-LangModels/README.md) | [PyTorch](https://microsoft.github.io/AI-For-Beginners/lessons/5-NLP/20-LangModels/GPT-PyTorch.ipynb) | | -| VI | **Autres techniques d'IA** || | -| 21 | [Algorithmes génétiques](./lessons/6-Other/21-GeneticAlgorithms/README.md) | [Notebook](./lessons/6-Other/21-GeneticAlgorithms/Genetic.ipynb) | | -| 22 | [Apprentissage par renforcement profond](./lessons/6-Other/22-DeepRL/README.md) | [PyTorch](./lessons/6-Other/22-DeepRL/CartPole-RL-PyTorch.ipynb) /[TensorFlow](./lessons/6-Other/22-DeepRL/CartPole-RL-TF.ipynb) | [Lab](./lessons/6-Other/22-DeepRL/lab/README.md) | +| 19 | [Reconnaissance d’entités nommées](./lessons/5-NLP/19-NER/README.md) | [TensorFlow](https://microsoft.github.io/AI-For-Beginners/lessons/5-NLP/19-NER/NER-TF.ipynb) | [Laboratoire](./lessons/5-NLP/19-NER/lab/README.md) | +| 20 | [Modèles de langage volumineux, programmation de prompt et tâches few-shot](./lessons/5-NLP/20-LangModels/README.md) | [PyTorch](https://microsoft.github.io/AI-For-Beginners/lessons/5-NLP/20-LangModels/GPT-PyTorch.ipynb) | | +| VI | **Autres techniques d’IA** || | +| 21 | [Algorithmes génétiques](./lessons/6-Other/21-GeneticAlgorithms/README.md) | [Carnet](./lessons/6-Other/21-GeneticAlgorithms/Genetic.ipynb) | | +| 22 | [Apprentissage par renforcement profond](./lessons/6-Other/22-DeepRL/README.md) | [PyTorch](./lessons/6-Other/22-DeepRL/CartPole-RL-PyTorch.ipynb) /[TensorFlow](./lessons/6-Other/22-DeepRL/CartPole-RL-TF.ipynb) | [Laboratoire](./lessons/6-Other/22-DeepRL/lab/README.md) | | 23 | [Systèmes multi-agents](./lessons/6-Other/23-MultiagentSystems/README.md) | | | -| VII | **Éthique de l'IA** | | | -| 24 | [Éthique de l'IA et IA responsable](./lessons/7-Ethics/README.md) | [Microsoft Learn : Principes de l'IA responsable](https://docs.microsoft.com/learn/paths/responsible-ai-business-principles/?WT.mc_id=academic-77998-cacaste) | | -| IX | **Suppléments** | | | -| 25 | [Réseaux multi-modaux, CLIP et VQGAN](./lessons/X-Extras/X1-MultiModal/README.md) | [Notebook](./lessons/X-Extras/X1-MultiModal/Clip.ipynb) | | +| VII | **Éthique de l’IA** | | | +| 24 | [Éthique de l’IA et IA responsable](./lessons/7-Ethics/README.md) | [Microsoft Learn : Principes de l’IA responsable](https://docs.microsoft.com/learn/paths/responsible-ai-business-principles/?WT.mc_id=academic-77998-cacaste) | | +| IX | **Extras** | | | +| 25 | [Réseaux multimodaux, CLIP et VQGAN](./lessons/X-Extras/X1-MultiModal/README.md) | [Carnet](./lessons/X-Extras/X1-MultiModal/Clip.ipynb) | | ## Chaque leçon contient * Du matériel de pré-lecture -* Des notebooks Jupyter exécutables, souvent spécifiques au framework (**PyTorch** ou **TensorFlow**). Le notebook exécutable contient également beaucoup de théorie, donc pour comprendre le sujet il est nécessaire de parcourir au moins une version du notebook (soit PyTorch soit TensorFlow). -* Des **labs** disponibles pour certains sujets, qui vous donnent l’opportunité d’appliquer le matériel appris à un problème spécifique. +* Des carnets Jupyter exécutables, souvent spécifiques à un framework (**PyTorch** ou **TensorFlow**). Le carnet exécutable contient également beaucoup de matériel théorique, donc pour comprendre le sujet, il faut parcourir au moins une version du carnet (soit PyTorch, soit TensorFlow). +* Des **laboratoires** disponibles pour certains sujets, qui vous donnent l’opportunité d’essayer d’appliquer le matériel appris à un problème spécifique. * Certaines sections contiennent des liens vers des modules [**MS Learn**](https://learn.microsoft.com/en-us/collections/7w28iy2xrqzdj0?WT.mc_id=academic-77998-bethanycheum) qui couvrent des sujets connexes. ## Pour commencer -### 🎯 Nouveau en IA ? Commencez ici ! +### 🎯 Débutant en IA ? Commencez ici ! -Si vous êtes complètement novice en IA et souhaitez des exemples rapides et pratiques, consultez nos [**Exemples accessibles aux débutants**](./examples/README.md) ! Ils comprennent : +Si vous êtes complètement débutant en IA et souhaitez des exemples rapides et pratiques, consultez nos [**Exemples pour débutants**](./examples/README.md) ! Ceux-ci incluent : -- 🌟 **Hello AI World** - Votre premier programme IA (reconnaissance de motifs) +- 🌟 **Hello AI World** - Votre premier programme IA (reconnaissance de motifs) - 🧠 **Réseau neuronal simple** - Construisez un réseau neuronal à partir de zéro -- 🖼️ **Classificateur d'images** - Classifier des images avec des commentaires détaillés -- 💬 **Sentiment du texte** - Analyse du texte positif/négatif +- 🖼️ **Classificateur d’images** - Classifiez des images avec des commentaires détaillés +- 💬 **Sentiment de texte** - Analysez le texte positif/négatif -Ces exemples sont conçus pour vous aider à comprendre les concepts d'IA avant de plonger dans le programme complet. +Ces exemples sont conçus pour vous aider à comprendre les concepts d’IA avant de plonger dans le curriculum complet. -### 📚 Configuration du programme complet +### 📚 Configuration complète du curriculum -- Nous avons créé une [leçon d'installation](./lessons/0-course-setup/setup.md) pour vous aider à configurer votre environnement de développement. - Pour les éducateurs, nous avons également créé une [leçon de configuration du programme](./lessons/0-course-setup/for-teachers.md) ! -- Comment [exécuter le code dans VSCode ou un Codespace](./lessons/0-course-setup/how-to-run.md) +- Nous avons créé une [leçon d’installation](./lessons/0-course-setup/setup.md) pour vous aider à configurer votre environnement de développement. +- Pour les éducateurs, nous avons aussi créé une [leçon de configuration du curriculum](./lessons/0-course-setup/for-teachers.md) pour vous ! +- Comment [exécuter le code dans VSCode ou un Codespace](./lessons/0-course-setup/how-to-run.md) -Suivez ces étapes : +Suivez ces étapes : -Forkez le dépôt : Cliquez sur le bouton "Fork" en haut à droite de cette page. +Forkez le dépôt : Cliquez sur le bouton « Fork » en haut à droite de cette page. -Clonez le dépôt : `git clone https://github.com/microsoft/AI-For-Beginners.git` +Clonez le dépôt : `git clone https://github.com/microsoft/AI-For-Beginners.git` -N’oubliez pas de mettre une étoile (🌟) sur ce dépôt pour le retrouver plus facilement plus tard. +N’oubliez pas d’étoiler (🌟) ce dépôt pour le retrouver plus facilement plus tard. -## Rencontrez d'autres apprenants +## Rencontrez d’autres apprenants -Rejoignez notre [serveur Discord officiel de l'IA](https://aka.ms/genai-discord?WT.mc_id=academic-105485-bethanycheum) pour rencontrer et réseauter avec d'autres apprenants suivant ce cours et obtenir du support. +Rejoignez notre [serveur Discord officiel AI](https://aka.ms/genai-discord?WT.mc_id=academic-105485-bethanycheum) pour rencontrer et réseauter avec d’autres apprenants suivant ce cours et obtenir du support. -Si vous avez des retours sur le produit ou des questions pendant la construction, visitez notre [Forum développeur Azure AI Foundry](https://aka.ms/foundry/forum). +Si vous avez des retours produits ou des questions pendant la création, rendez-vous sur notre [Forum des développeurs Azure AI Foundry](https://aka.ms/foundry/forum) -## Quiz +## Quiz -> **Une note à propos des quiz** : Tous les quiz se trouvent dans le dossier Quiz-app dans etc\quiz-app, ou [en ligne ici](https://ff-quizzes.netlify.app/) Ils sont liés depuis les leçons, l'application quiz peut être exécutée localement ou déployée sur Azure ; suivez les instructions dans le dossier `quiz-app`. Ils sont progressivement localisés. +> **Une note au sujet des quiz** : Tous les quiz sont contenus dans le dossier Quiz-app dans etc\quiz-app, ou [en ligne ici](https://ff-quizzes.netlify.app/) Ils sont reliés depuis les leçons, l’application quiz peut être exécutée localement ou déployée sur Azure ; suivez les instructions dans le dossier `quiz-app`. Ils sont progressivement localisés. -## Besoin d'aide +## Recherche d’aide -Avez-vous des suggestions ou avez-vous trouvé des fautes d’orthographe ou d’erreurs de code ? Ouvrez une issue ou créez une pull request. +Avez-vous des suggestions ou trouvé des fautes d’orthographe ou d’erreurs de code ? Ouvrez un ticket ou créez une demande de tirage. -## Remerciements spéciaux +## Remerciements spéciaux -* **✍️ Auteur principal :** [Dmitry Soshnikov](http://soshnikov.com), PhD -* **🔥 Éditeur :** [Jen Looper](https://twitter.com/jenlooper), PhD -* **🎨 Illustratrice Sketchnote :** [Tomomi Imura](https://twitter.com/girlie_mac) -* **✅ Créatrice de quiz :** [Lateefah Bello](https://github.com/CinnamonXI), [MLSA](https://studentambassadors.microsoft.com/) -* **🙏 Contributeurs principaux :** [Evgenii Pishchik](https://github.com/Pe4enIks) +* **✍️ Auteur principal :** [Dmitry Soshnikov](http://soshnikov.com), PhD +* **🔥 Éditeur :** [Jen Looper](https://twitter.com/jenlooper), PhD +* **🎨 Illustrateur Sketchnote :** [Tomomi Imura](https://twitter.com/girlie_mac) +* **✅ Créatrice de quiz :** [Lateefah Bello](https://github.com/CinnamonXI), [MLSA](https://studentambassadors.microsoft.com/) +* **🙏 Contributeurs principaux :** [Evgenii Pishchik](https://github.com/Pe4enIks) -## Autres programmes +## Autres Curricula -Notre équipe produit d’autres programmes ! Découvrez : +Notre équipe produit d’autres curricula ! Découvrez : - -### LangChain -[![LangChain4j for Beginners](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js for Beginners](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) + +### LangChain +[![LangChain4j for Beginners](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js for Beginners](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) ---- +--- -### Azure / Edge / MCP / Agents -[![AZD for Beginners](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI for Beginners](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP for Beginners](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agents for Beginners](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +### Azure / Edge / MCP / Agents +[![AZD for Beginners](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI for Beginners](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP for Beginners](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agents for Beginners](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) ---- - -### Série IA générative -[![Generative AI for Beginners](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) +--- + +### Série IA Générative +[![Generative AI for Beginners](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) ---- - -### Apprentissage fondamental -[![ML for Beginners](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Data Science for Beginners](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![AI for Beginners](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Cybersecurity for Beginners](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Web Dev for Beginners](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT for Beginners](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR Development for Beginners](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) +--- + +### Apprentissage Fondamental +[![ML for Beginners](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Data Science for Beginners](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![AI for Beginners](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![Cybersecurity for Beginners](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Web Dev for Beginners](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT for Beginners](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![XR Development for Beginners](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) ---- - -### Série Copilot -[![Copilot for AI Paired Programming](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot for C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) - +--- + +### Série Copilot +[![Copilot for AI Paired Programming](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot for C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) + -## Obtenir de l’aide +## Obtenir de l’aide -Si vous êtes bloqué ou si vous avez des questions sur la création d’applications IA. Rejoignez d’autres apprenants et développeurs expérimentés dans des discussions sur MCP. C’est une communauté de soutien où les questions sont les bienvenues et où le savoir est partagé librement. +Si vous êtes bloqué ou avez des questions sur la création d’applications IA. Rejoignez d’autres apprenants et développeurs expérimentés dans les discussions sur MCP. C’est une communauté bienveillante où les questions sont les bienvenues et le savoir est partagé librement. -[![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/nTYy5BXMWG)](https://discord.gg/nTYy5BXMWG) +[![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/nTYy5BXMWG)](https://discord.gg/nTYy5BXMWG) -Si vous avez des retours produits ou des erreurs lors de la création, visitez : +Si vous avez des retours produits ou des erreurs pendant la création, visitez : [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) @@ -228,5 +221,5 @@ Si vous avez des retours produits ou des erreurs lors de la création, visitez : **Avertissement** : -Ce document a été traduit à l’aide du service de traduction par IA [Co-op Translator](https://github.com/Azure/co-op-translator). Bien que nous nous efforçons d’assurer l’exactitude, veuillez noter que les traductions automatisées peuvent comporter des erreurs ou des imprécisions. Le document original dans sa langue d’origine doit être considéré comme la source fiable. Pour toute information critique, une traduction professionnelle réalisée par un humain est recommandée. Nous déclinons toute responsabilité en cas de malentendus ou d’interprétations erronées résultant de l’utilisation de cette traduction. +Ce document a été traduit à l’aide du service de traduction automatique [Co-op Translator](https://github.com/Azure/co-op-translator). Bien que nous nous efforçons d’assurer l’exactitude, veuillez noter que les traductions automatiques peuvent contenir des erreurs ou des inexactitudes. Le document original dans sa langue d’origine doit être considéré comme la source officielle. Pour les informations critiques, il est recommandé de recourir à une traduction professionnelle humaine. Nous ne saurions être tenus responsables des malentendus ou interprétations erronées résultant de l’utilisation de cette traduction. \ No newline at end of file diff --git a/translations/fr/SECURITY.md b/translations/fr/SECURITY.md index 3252ad6f..7e45f4bd 100644 --- a/translations/fr/SECURITY.md +++ b/translations/fr/SECURITY.md @@ -1,12 +1,3 @@ - ## Sécurité Microsoft prend très au sérieux la sécurité de ses produits logiciels et services, y compris tous les dépôts de code source gérés via nos organisations GitHub, qui incluent [Microsoft](https://github.com/Microsoft), [Azure](https://github.com/Azure), [DotNet](https://github.com/dotnet), [AspNet](https://github.com/aspnet), [Xamarin](https://github.com/xamarin) et [nos organisations GitHub](https://opensource.microsoft.com/). diff --git a/translations/fr/etc/CODE_OF_CONDUCT.md b/translations/fr/etc/CODE_OF_CONDUCT.md index 6423ff85..282a6474 100644 --- a/translations/fr/etc/CODE_OF_CONDUCT.md +++ b/translations/fr/etc/CODE_OF_CONDUCT.md @@ -1,12 +1,3 @@ - # Code de conduite Open Source de Microsoft Ce projet a adopté le [Code de conduite Open Source de Microsoft](https://opensource.microsoft.com/codeofconduct/). diff --git a/translations/fr/etc/CONTRIBUTING.md b/translations/fr/etc/CONTRIBUTING.md index 31fc1b37..3ba220b7 100644 --- a/translations/fr/etc/CONTRIBUTING.md +++ b/translations/fr/etc/CONTRIBUTING.md @@ -1,12 +1,3 @@ - # Contribuer Ce projet accueille volontiers les contributions et suggestions. La plupart des contributions nécessitent que vous acceptiez un Accord de Licence de Contributeur (CLA) déclarant que vous avez le droit, et que vous accordez effectivement, les droits nécessaires pour utiliser votre contribution. Pour plus de détails, visitez https://cla.microsoft.com. diff --git a/translations/fr/etc/Mindmap.md b/translations/fr/etc/Mindmap.md index b0fa252b..6653324b 100644 --- a/translations/fr/etc/Mindmap.md +++ b/translations/fr/etc/Mindmap.md @@ -1,12 +1,3 @@ - # IA ## [Introduction à l'IA](https://github.com/microsoft/AI-For-Beginners/blob/main/lessons/1-Intro/README.md) diff --git a/translations/fr/etc/SUPPORT.md b/translations/fr/etc/SUPPORT.md index f2adb9c2..75e010af 100644 --- a/translations/fr/etc/SUPPORT.md +++ b/translations/fr/etc/SUPPORT.md @@ -1,12 +1,3 @@ - # Support ## Comment signaler des problèmes et obtenir de l'aide diff --git a/translations/fr/etc/TRANSLATIONS.md b/translations/fr/etc/TRANSLATIONS.md index 32ddbc69..e73a9b05 100644 --- a/translations/fr/etc/TRANSLATIONS.md +++ b/translations/fr/etc/TRANSLATIONS.md @@ -1,12 +1,3 @@ - # Contribuer en traduisant les leçons Nous accueillons avec plaisir les traductions des leçons de ce programme ! diff --git a/translations/fr/etc/quiz-app/README.md b/translations/fr/etc/quiz-app/README.md index 0264b0b6..f137581e 100644 --- a/translations/fr/etc/quiz-app/README.md +++ b/translations/fr/etc/quiz-app/README.md @@ -1,12 +1,3 @@ - # Quiz Ces quiz sont les quiz pré- et post-cours pour le programme d'IA disponible sur https://aka.ms/ai-beginners diff --git a/translations/fr/examples/README.md b/translations/fr/examples/README.md index 0a517654..b748d1f3 100644 --- a/translations/fr/examples/README.md +++ b/translations/fr/examples/README.md @@ -1,12 +1,3 @@ - # Exemples d'IA pour Débutants Bienvenue ! Ce répertoire contient des exemples simples et autonomes pour vous aider à démarrer avec l'IA et l'apprentissage automatique. Chaque exemple est conçu pour être accessible aux débutants, avec des commentaires détaillés et des explications étape par étape. diff --git a/translations/fr/lessons/0-course-setup/for-teachers.md b/translations/fr/lessons/0-course-setup/for-teachers.md index d367dff4..69c62d2d 100644 --- a/translations/fr/lessons/0-course-setup/for-teachers.md +++ b/translations/fr/lessons/0-course-setup/for-teachers.md @@ -1,12 +1,3 @@ - # Pour les enseignants Souhaitez-vous utiliser ce programme dans votre classe ? N'hésitez pas ! diff --git a/translations/fr/lessons/0-course-setup/how-to-run.md b/translations/fr/lessons/0-course-setup/how-to-run.md index 62771996..95f32fee 100644 --- a/translations/fr/lessons/0-course-setup/how-to-run.md +++ b/translations/fr/lessons/0-course-setup/how-to-run.md @@ -1,12 +1,3 @@ - # Comment exécuter le code Ce programme contient beaucoup d’exemples exécutables et de travaux pratiques que vous voudrez exécuter. Pour ce faire, vous avez besoin de pouvoir exécuter du code Python dans les notebooks Jupyter fournis dans le cadre de ce programme. Vous disposez de plusieurs options pour exécuter le code : diff --git a/translations/fr/lessons/0-course-setup/setup.md b/translations/fr/lessons/0-course-setup/setup.md index 7f3aa8dc..22e02c13 100644 --- a/translations/fr/lessons/0-course-setup/setup.md +++ b/translations/fr/lessons/0-course-setup/setup.md @@ -1,12 +1,3 @@ - # Commencer avec ce programme ## Êtes-vous étudiant ? diff --git a/translations/fr/lessons/1-Intro/README.md b/translations/fr/lessons/1-Intro/README.md index ad8c3c74..e9499396 100644 --- a/translations/fr/lessons/1-Intro/README.md +++ b/translations/fr/lessons/1-Intro/README.md @@ -1,12 +1,3 @@ - # Introduction à l'IA ![Résumé du contenu de l'introduction à l'IA sous forme de dessin](../../../../translated_images/fr/ai-intro.bf28d1ac4235881c.webp) diff --git a/translations/fr/lessons/1-Intro/assignment.md b/translations/fr/lessons/1-Intro/assignment.md index 87491e05..f435f409 100644 --- a/translations/fr/lessons/1-Intro/assignment.md +++ b/translations/fr/lessons/1-Intro/assignment.md @@ -1,12 +1,3 @@ - # Game Jam Les jeux sont un domaine qui a été fortement influencé par les avancées en IA et en apprentissage automatique. Dans cet exercice, rédigez un court article sur un jeu que vous appréciez et qui a été influencé par l'évolution de l'IA. Il devrait s'agir d'un jeu suffisamment ancien pour avoir été impacté par plusieurs types de systèmes de traitement informatique. Un bon exemple est le jeu d'échecs ou le jeu de Go, mais vous pouvez également examiner des jeux vidéo comme Pong ou Pac-Man. Rédigez un essai qui discute du passé, du présent et de l'avenir de l'IA dans ce jeu. diff --git a/translations/fr/lessons/2-Symbolic/README.md b/translations/fr/lessons/2-Symbolic/README.md index ce33de1f..702b4ce3 100644 --- a/translations/fr/lessons/2-Symbolic/README.md +++ b/translations/fr/lessons/2-Symbolic/README.md @@ -1,15 +1,6 @@ - # Représentation des Connaissances et Systèmes Experts -![Résumé du contenu de l'IA symbolique](../../../../../../translated_images/fr/ai-symbolic.715a30cb610411a6.webp) +![Résumé du contenu de l'IA symbolique](../../../../translated_images/fr/ai-symbolic.715a30cb610411a6.webp) > Sketchnote par [Tomomi Imura](https://twitter.com/girlie_mac) @@ -41,7 +32,7 @@ Le plus souvent, nous ne définissons pas strictement la connaissance, mais nous Ainsi, le problème de la **représentation des connaissances** est de trouver un moyen efficace de représenter la connaissance à l'intérieur d'un ordinateur sous forme de données, afin de la rendre automatiquement utilisable. Ceci peut être vu comme un spectre : -![Spectre de la représentation des connaissances](../../../../../../translated_images/fr/knowledge-spectrum.b60df631852c0217.webp) +![Spectre de la représentation des connaissances](../../../../translated_images/fr/knowledge-spectrum.b60df631852c0217.webp) > Image par [Dmitry Soshnikov](http://soshnikov.com) @@ -94,7 +85,7 @@ Syntaxe de bloc | Indentation | | | Un des premiers succès de l’IA symbolique fut les systèmes dits **experts** – des systèmes informatiques conçus pour agir en tant qu’expert dans un domaine de problème limité. Ils étaient basés sur une **base de connaissances** extraite de un ou plusieurs experts humains, et contenaient un **moteur d’inférence** qui effectuait un raisonnement sur cette base. -![Architecture Humaine](../../../../../../translated_images/fr/arch-human.5d4d35f1bba3ab1c.webp) | ![Système à base de connaissances](../../../../../../translated_images/fr/arch-kbs.3ec5c150b09fa8da.webp) +![Architecture Humaine](../../../../translated_images/fr/arch-human.5d4d35f1bba3ab1c.webp) | ![Système à base de connaissances](../../../../translated_images/fr/arch-kbs.3ec5c150b09fa8da.webp) ---------------------------------------------|------------------------------------------------ Structure simplifiée du système neural humain | Architecture d’un système à base de connaissances @@ -106,7 +97,7 @@ Les systèmes experts sont construits comme le système de raisonnement humain, Par exemple, considérons le système expert suivant permettant de déterminer un animal en fonction de ses caractéristiques physiques : -![Arbre ET-OU](../../../../../../translated_images/fr/AND-OR-Tree.5592d2c70187f283.webp) +![Arbre ET-OU](../../../../translated_images/fr/AND-OR-Tree.5592d2c70187f283.webp) > Image par [Dmitry Soshnikov](http://soshnikov.com) diff --git a/translations/fr/lessons/2-Symbolic/assignment.md b/translations/fr/lessons/2-Symbolic/assignment.md index d83ce15e..1de93c7b 100644 --- a/translations/fr/lessons/2-Symbolic/assignment.md +++ b/translations/fr/lessons/2-Symbolic/assignment.md @@ -1,12 +1,3 @@ - # Construire une Ontologie Construire une base de connaissances consiste à catégoriser un modèle représentant des faits sur un sujet. Choisissez un sujet - comme une personne, un lieu ou un objet - puis construisez un modèle de ce sujet. Utilisez certaines des techniques et stratégies de construction de modèles décrites dans cette leçon. Un exemple serait de créer une ontologie d'un salon avec des meubles, des lumières, etc. En quoi le salon diffère-t-il de la cuisine ? De la salle de bain ? Comment savez-vous que c'est un salon et non une salle à manger ? Utilisez [Protégé](https://protege.stanford.edu/) pour construire votre ontologie. diff --git a/translations/fr/lessons/3-NeuralNetworks/03-Perceptron/README.md b/translations/fr/lessons/3-NeuralNetworks/03-Perceptron/README.md index a9cf3a39..5e821499 100644 --- a/translations/fr/lessons/3-NeuralNetworks/03-Perceptron/README.md +++ b/translations/fr/lessons/3-NeuralNetworks/03-Perceptron/README.md @@ -1,12 +1,3 @@ - # Introduction aux réseaux neuronaux : Perceptron ## [Quiz pré-conférence](https://ff-quizzes.netlify.app/en/ai/quiz/5) @@ -15,7 +6,7 @@ L'une des premières tentatives pour mettre en œuvre quelque chose de similaire | | | |--------------|-----------| -|Frank Rosenblatt | Le Perceptron Mark 1| +|Frank Rosenblatt | Le Perceptron Mark 1| > Images [tirées de Wikipédia](https://en.wikipedia.org/wiki/Perceptron) @@ -34,7 +25,7 @@ y(x) = f(wTx) où f est une fonction d'activation en escalier. - + ## Entraînement du Perceptron diff --git a/translations/fr/lessons/3-NeuralNetworks/03-Perceptron/lab/README.md b/translations/fr/lessons/3-NeuralNetworks/03-Perceptron/lab/README.md index 51f55281..1dffd747 100644 --- a/translations/fr/lessons/3-NeuralNetworks/03-Perceptron/lab/README.md +++ b/translations/fr/lessons/3-NeuralNetworks/03-Perceptron/lab/README.md @@ -1,12 +1,3 @@ - # Classification multi-classes avec Perceptron Travail pratique tiré du [Curriculum AI pour les débutants](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/fr/lessons/3-NeuralNetworks/04-OwnFramework/README.md b/translations/fr/lessons/3-NeuralNetworks/04-OwnFramework/README.md index e09c80c9..6ac500ad 100644 --- a/translations/fr/lessons/3-NeuralNetworks/04-OwnFramework/README.md +++ b/translations/fr/lessons/3-NeuralNetworks/04-OwnFramework/README.md @@ -1,12 +1,3 @@ - # Introduction aux réseaux neuronaux. Perceptron multicouche Dans la section précédente, vous avez découvert le modèle de réseau neuronal le plus simple : le perceptron à une couche, un modèle de classification linéaire à deux classes. @@ -65,7 +56,7 @@ L'algorithme de descente de gradient reste le même, mais le calcul des gradient Notez que la partie la plus à gauche de toutes ces expressions est la même, et nous pouvons donc calculer efficacement les dérivées en commençant par la fonction de perte et en remontant "en arrière" à travers le graphe computationnel. Ainsi, la méthode d'entraînement d'un perceptron multicouche est appelée **rétropropagation**, ou 'backprop'. -graphe de calcul +graphe de calcul > TODO : citation de l'image diff --git a/translations/fr/lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md b/translations/fr/lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md index 5289ca4e..7e6371d1 100644 --- a/translations/fr/lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md +++ b/translations/fr/lessons/3-NeuralNetworks/04-OwnFramework/lab/README.md @@ -1,12 +1,3 @@ - # Classification MNIST avec notre propre framework Travail pratique issu du [Curriculum AI pour les débutants](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/fr/lessons/3-NeuralNetworks/05-Frameworks/README.md b/translations/fr/lessons/3-NeuralNetworks/05-Frameworks/README.md index 17716cbb..14041cc5 100644 --- a/translations/fr/lessons/3-NeuralNetworks/05-Frameworks/README.md +++ b/translations/fr/lessons/3-NeuralNetworks/05-Frameworks/README.md @@ -1,12 +1,3 @@ - # Frameworks de réseaux neuronaux Comme nous l'avons déjà appris, pour entraîner efficacement des réseaux neuronaux, nous devons faire deux choses : diff --git a/translations/fr/lessons/3-NeuralNetworks/05-Frameworks/lab/README.md b/translations/fr/lessons/3-NeuralNetworks/05-Frameworks/lab/README.md index d02135d6..aee385ba 100644 --- a/translations/fr/lessons/3-NeuralNetworks/05-Frameworks/lab/README.md +++ b/translations/fr/lessons/3-NeuralNetworks/05-Frameworks/lab/README.md @@ -1,12 +1,3 @@ - # Classification avec PyTorch/TensorFlow Devoir pratique tiré du [Curriculum AI for Beginners](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/fr/lessons/3-NeuralNetworks/README.md b/translations/fr/lessons/3-NeuralNetworks/README.md index 3edf05e9..225b2a72 100644 --- a/translations/fr/lessons/3-NeuralNetworks/README.md +++ b/translations/fr/lessons/3-NeuralNetworks/README.md @@ -1,12 +1,3 @@ - # Introduction aux réseaux neuronaux ![Résumé du contenu d'introduction aux réseaux neuronaux sous forme de dessin](../../../../translated_images/fr/ai-neuralnetworks.1c687ae40bc86e83.webp) diff --git a/translations/fr/lessons/4-ComputerVision/06-IntroCV/README.md b/translations/fr/lessons/4-ComputerVision/06-IntroCV/README.md index 6072ebb4..e4b059cc 100644 --- a/translations/fr/lessons/4-ComputerVision/06-IntroCV/README.md +++ b/translations/fr/lessons/4-ComputerVision/06-IntroCV/README.md @@ -1,12 +1,3 @@ - # Introduction à la vision par ordinateur [La vision par ordinateur](https://wikipedia.org/wiki/Computer_vision) est une discipline dont le but est de permettre aux ordinateurs de comprendre les images numériques à un niveau élevé. Cette définition est assez large, car *comprendre* peut signifier de nombreuses choses différentes, comme trouver un objet sur une image (**détection d'objet**), comprendre ce qui se passe (**détection d'événements**), décrire une image en texte ou reconstruire une scène en 3D. Il existe également des tâches spécifiques liées aux images humaines : estimation de l'âge et des émotions, détection et identification de visages, et estimation de la pose en 3D, pour n'en citer que quelques-unes. @@ -115,7 +106,7 @@ Lisez davantage sur le flux optique [dans ce tutoriel excellent](https://learnop Dans ce laboratoire, vous prendrez une vidéo avec des gestes simples, et votre objectif sera d'extraire les mouvements haut/bas/gauche/droite en utilisant le flux optique. -Image de mouvement de la paume +Image de mouvement de la paume --- diff --git a/translations/fr/lessons/4-ComputerVision/06-IntroCV/lab/README.md b/translations/fr/lessons/4-ComputerVision/06-IntroCV/lab/README.md index 2b66be89..8ae6c2b8 100644 --- a/translations/fr/lessons/4-ComputerVision/06-IntroCV/lab/README.md +++ b/translations/fr/lessons/4-ComputerVision/06-IntroCV/lab/README.md @@ -1,12 +1,3 @@ - # Détection de mouvements avec le flux optique Exercice pratique tiré du [programme AI for Beginners](https://aka.ms/ai-beginners). diff --git a/translations/fr/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md b/translations/fr/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md index dd24c70c..1e8bf4f7 100644 --- a/translations/fr/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md +++ b/translations/fr/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md @@ -1,12 +1,3 @@ - # Architectures CNN Bien Connues ### VGG-16 @@ -25,7 +16,7 @@ Comme vous pouvez le voir, VGG suit une architecture pyramidale traditionnelle, ResNet est une famille de modèles proposée par Microsoft Research en 2015. L'idée principale de ResNet est d'utiliser des **blocs résiduels** : - + > Image tirée de [cet article](https://arxiv.org/pdf/1512.03385.pdf) @@ -37,7 +28,7 @@ Vous pouvez également considérer ce réseau comme capable d'ajuster sa complex L'architecture Google Inception pousse cette idée un peu plus loin et construit chaque couche du réseau comme une combinaison de plusieurs chemins différents : - + > Image tirée de [Researchgate](https://www.researchgate.net/figure/Inception-module-with-dimension-reductions-left-and-schema-for-Inception-ResNet-v1_fig2_355547454) diff --git a/translations/fr/lessons/4-ComputerVision/07-ConvNets/README.md b/translations/fr/lessons/4-ComputerVision/07-ConvNets/README.md index 8994f3d6..9e1afe5e 100644 --- a/translations/fr/lessons/4-ComputerVision/07-ConvNets/README.md +++ b/translations/fr/lessons/4-ComputerVision/07-ConvNets/README.md @@ -1,12 +1,3 @@ - # Réseaux de Neurones Convolutionnels Nous avons vu précédemment que les réseaux de neurones sont assez performants pour traiter les images, et même un perceptron à une seule couche est capable de reconnaître les chiffres manuscrits du jeu de données MNIST avec une précision raisonnable. Cependant, le jeu de données MNIST est très particulier, car tous les chiffres y sont centrés dans l'image, ce qui simplifie la tâche. @@ -24,7 +15,7 @@ Pour extraire ces motifs, nous utiliserons la notion de **filtres convolutionnel Par exemple, si nous appliquons des filtres de bord vertical et horizontal 3x3 aux chiffres du MNIST, nous obtenons des zones mises en évidence (par exemple, des valeurs élevées) là où se trouvent des bords verticaux et horizontaux dans notre image originale. Ainsi, ces deux filtres peuvent être utilisés pour "chercher" des bords. De la même manière, nous pouvons concevoir différents filtres pour détecter d'autres motifs de bas niveau : - + > Image de [Leung-Malik Filter Bank](https://www.robots.ox.ac.uk/~vgg/research/texclass/filters.html) diff --git a/translations/fr/lessons/4-ComputerVision/07-ConvNets/lab/README.md b/translations/fr/lessons/4-ComputerVision/07-ConvNets/lab/README.md index bebf65fd..7dcaa8cd 100644 --- a/translations/fr/lessons/4-ComputerVision/07-ConvNets/lab/README.md +++ b/translations/fr/lessons/4-ComputerVision/07-ConvNets/lab/README.md @@ -1,12 +1,3 @@ - # Classification des visages d'animaux Travail pratique issu du [Curriculum AI pour débutants](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/fr/lessons/4-ComputerVision/08-TransferLearning/README.md b/translations/fr/lessons/4-ComputerVision/08-TransferLearning/README.md index 913c0cae..57101fe3 100644 --- a/translations/fr/lessons/4-ComputerVision/08-TransferLearning/README.md +++ b/translations/fr/lessons/4-ComputerVision/08-TransferLearning/README.md @@ -1,12 +1,3 @@ - # Réseaux pré-entraînés et apprentissage par transfert L'entraînement des CNN peut prendre beaucoup de temps et nécessite une grande quantité de données. Cependant, une grande partie du temps est consacrée à l'apprentissage des meilleurs filtres de bas niveau qu'un réseau peut utiliser pour extraire des motifs à partir d'images. Une question naturelle se pose : peut-on utiliser un réseau neuronal entraîné sur un ensemble de données et l'adapter pour classifier d'autres images sans nécessiter un processus d'entraînement complet ? diff --git a/translations/fr/lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md b/translations/fr/lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md index 4f39d6e0..458a2c09 100644 --- a/translations/fr/lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md +++ b/translations/fr/lessons/4-ComputerVision/08-TransferLearning/TrainingTricks.md @@ -1,12 +1,3 @@ - # Astuces pour l'entraînement en Deep Learning À mesure que les réseaux neuronaux deviennent plus profonds, leur entraînement devient de plus en plus complexe. Un problème majeur est celui des [gradients qui disparaissent](https://en.wikipedia.org/wiki/Vanishing_gradient_problem) ou des [gradients qui explosent](https://deepai.org/machine-learning-glossary-and-terms/exploding-gradient-problem#:~:text=Exploding%20gradients%20are%20a%20problem,updates%20are%20small%20and%20controlled.). [Cet article](https://towardsdatascience.com/the-vanishing-exploding-gradient-problem-in-deep-neural-networks-191358470c11) offre une bonne introduction à ces problèmes. diff --git a/translations/fr/lessons/4-ComputerVision/08-TransferLearning/lab/README.md b/translations/fr/lessons/4-ComputerVision/08-TransferLearning/lab/README.md index d5d978a3..8517f3dc 100644 --- a/translations/fr/lessons/4-ComputerVision/08-TransferLearning/lab/README.md +++ b/translations/fr/lessons/4-ComputerVision/08-TransferLearning/lab/README.md @@ -1,12 +1,3 @@ - # Classification des animaux d'Oxford à l'aide de l'apprentissage par transfert Devoir pratique tiré du [programme AI for Beginners](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/fr/lessons/4-ComputerVision/09-Autoencoders/README.md b/translations/fr/lessons/4-ComputerVision/09-Autoencoders/README.md index 3f921d4e..3373879a 100644 --- a/translations/fr/lessons/4-ComputerVision/09-Autoencoders/README.md +++ b/translations/fr/lessons/4-ComputerVision/09-Autoencoders/README.md @@ -1,12 +1,3 @@ - # Autoencodeurs Lors de l'entraînement des CNN, l'un des problèmes est que nous avons besoin de beaucoup de données annotées. Dans le cas de la classification d'images, il faut séparer les images en différentes classes, ce qui demande un effort manuel. @@ -46,7 +37,7 @@ En résumé : * Nous échantillonnons un vecteur `sample` à partir de la distribution N(zmean,exp(zlog\_sigma)) * Le décodeur essaie de décoder l'image originale en utilisant `sample` comme vecteur d'entrée - + > Image tirée de [cet article de blog](https://ijdykeman.github.io/ml/2016/12/21/cvae.html) par Isaak Dykeman @@ -57,13 +48,13 @@ Les autoencodeurs variationnels utilisent une fonction de perte complexe compos Un avantage important des VAE est qu'ils permettent de générer de nouvelles images relativement facilement, car nous savons de quelle distribution échantillonner les vecteurs latents. Par exemple, si nous entraînons un VAE avec un vecteur latent 2D sur MNIST, nous pouvons ensuite varier les composantes du vecteur latent pour obtenir différents chiffres : -vaemnist +vaemnist > Image par [Dmitry Soshnikov](http://soshnikov.com) Observez comment les images se fondent les unes dans les autres, à mesure que nous obtenons des vecteurs latents provenant de différentes parties de l'espace des paramètres latents. Nous pouvons également visualiser cet espace en 2D : -vaemnist cluster +vaemnist cluster > Image par [Dmitry Soshnikov](http://soshnikov.com) diff --git a/translations/fr/lessons/4-ComputerVision/10-GANs/README.md b/translations/fr/lessons/4-ComputerVision/10-GANs/README.md index e21a5adf..827dd6ac 100644 --- a/translations/fr/lessons/4-ComputerVision/10-GANs/README.md +++ b/translations/fr/lessons/4-ComputerVision/10-GANs/README.md @@ -1,12 +1,3 @@ - # Réseaux Adversaires Génératifs Dans la section précédente, nous avons appris à propos des **modèles génératifs** : des modèles capables de générer de nouvelles images similaires à celles du jeu de données d'entraînement. Le VAE était un bon exemple de modèle génératif. @@ -17,7 +8,7 @@ Cependant, si nous essayons de générer quelque chose de vraiment significatif, L'idée principale d'un GAN est d'avoir deux réseaux neuronaux qui seront entraînés l'un contre l'autre : - + > Image par [Dmitry Soshnikov](http://soshnikov.com) @@ -41,7 +32,7 @@ Un générateur est légèrement plus complexe. Vous pouvez le considérer comme > ✅ Étant donné que la couche de convolution est implémentée comme un filtre linéaire parcourant l'image, la déconvolution est essentiellement similaire à la convolution et peut être mise en œuvre en utilisant la même logique de couche. - + > Image par [Dmitry Soshnikov](http://soshnikov.com) diff --git a/translations/fr/lessons/4-ComputerVision/11-ObjectDetection/README.md b/translations/fr/lessons/4-ComputerVision/11-ObjectDetection/README.md index 11fc7287..66d4d6d4 100644 --- a/translations/fr/lessons/4-ComputerVision/11-ObjectDetection/README.md +++ b/translations/fr/lessons/4-ComputerVision/11-ObjectDetection/README.md @@ -1,12 +1,3 @@ - # Détection d'objets Les modèles de classification d'images que nous avons abordés jusqu'à présent prenaient une image et produisaient un résultat catégorique, comme la classe 'nombre' dans un problème MNIST. Cependant, dans de nombreux cas, nous ne voulons pas seulement savoir qu'une image représente des objets - nous voulons pouvoir déterminer leur emplacement précis. C'est précisément l'objectif de la **détection d'objets**. diff --git a/translations/fr/lessons/4-ComputerVision/11-ObjectDetection/lab/README.md b/translations/fr/lessons/4-ComputerVision/11-ObjectDetection/lab/README.md index 1c9a48e2..8d86e31b 100644 --- a/translations/fr/lessons/4-ComputerVision/11-ObjectDetection/lab/README.md +++ b/translations/fr/lessons/4-ComputerVision/11-ObjectDetection/lab/README.md @@ -1,12 +1,3 @@ - # Détection de têtes avec le dataset Hollywood Heads Travail pratique issu du [Curriculum AI pour les débutants](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/fr/lessons/4-ComputerVision/12-Segmentation/README.md b/translations/fr/lessons/4-ComputerVision/12-Segmentation/README.md index 4aafc081..351706ce 100644 --- a/translations/fr/lessons/4-ComputerVision/12-Segmentation/README.md +++ b/translations/fr/lessons/4-ComputerVision/12-Segmentation/README.md @@ -1,12 +1,3 @@ - # Segmentation Nous avons précédemment appris la détection d'objets, qui nous permet de localiser des objets dans une image en prédisant leurs *boîtes englobantes*. Cependant, pour certaines tâches, nous avons besoin d'une localisation d'objets plus précise que les simples boîtes englobantes. Cette tâche est appelée **segmentation**. @@ -20,7 +11,7 @@ La segmentation peut être vue comme une **classification de pixels**, où pour Par exemple, pour la segmentation par instance, ces moutons sont des objets distincts, tandis que pour la segmentation sémantique, tous les moutons sont représentés par une seule classe. - + > Image tirée de [cet article de blog](https://nirmalamurali.medium.com/image-classification-vs-semantic-segmentation-vs-instance-segmentation-625c33a08d50) @@ -29,7 +20,7 @@ Il existe différentes architectures neuronales pour la segmentation, mais elles * **Encodeur** : extrait les caractéristiques de l'image d'entrée. * **Décodeur** : transforme ces caractéristiques en une **image de masque**, ayant la même taille et un nombre de canaux correspondant au nombre de classes. - + > Image tirée de [cette publication](https://arxiv.org/pdf/2001.05566.pdf) @@ -43,7 +34,7 @@ Dans cette leçon, nous verrons la segmentation en action en entraînant un rés > ✅ Cette technique est particulièrement adaptée à ce type d'imagerie médicale, mais quelles autres applications réelles pourriez-vous imaginer ? -navi +navi > Image tirée de la base de données PH2 diff --git a/translations/fr/lessons/4-ComputerVision/12-Segmentation/lab/README.md b/translations/fr/lessons/4-ComputerVision/12-Segmentation/lab/README.md index 65b4b598..f0749bfd 100644 --- a/translations/fr/lessons/4-ComputerVision/12-Segmentation/lab/README.md +++ b/translations/fr/lessons/4-ComputerVision/12-Segmentation/lab/README.md @@ -1,12 +1,3 @@ - # Segmentation du Corps Humain Devoir de laboratoire tiré du [Curriculum AI pour Débutants](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/fr/lessons/4-ComputerVision/README.md b/translations/fr/lessons/4-ComputerVision/README.md index ab7b342c..9c7b7579 100644 --- a/translations/fr/lessons/4-ComputerVision/README.md +++ b/translations/fr/lessons/4-ComputerVision/README.md @@ -1,12 +1,3 @@ - # Vision par ordinateur ![Résumé du contenu sur la vision par ordinateur sous forme de dessin](../../../../translated_images/fr/ai-computervision.6506ebebac3fbf76.webp) diff --git a/translations/fr/lessons/5-NLP/13-TextRep/README.md b/translations/fr/lessons/5-NLP/13-TextRep/README.md index 1c636e73..07da277c 100644 --- a/translations/fr/lessons/5-NLP/13-TextRep/README.md +++ b/translations/fr/lessons/5-NLP/13-TextRep/README.md @@ -1,12 +1,3 @@ - # Représenter le texte sous forme de tenseurs ## [Quiz avant le cours](https://ff-quizzes.netlify.app/en/ai/quiz/25) @@ -25,7 +16,7 @@ Notre objectif sera de classer l'article dans l'une des catégories en fonction Pour résoudre des tâches de traitement du langage naturel (NLP) avec des réseaux neuronaux, nous devons trouver un moyen de représenter le texte sous forme de tenseurs. Les ordinateurs représentent déjà les caractères textuels sous forme de nombres qui correspondent aux polices affichées sur votre écran, en utilisant des encodages tels que ASCII ou UTF-8. -Image montrant un diagramme qui associe un caractère à une représentation ASCII et binaire +Image montrant un diagramme qui associe un caractère à une représentation ASCII et binaire > [Source de l'image](https://www.seobility.net/en/wiki/ASCII) @@ -48,7 +39,7 @@ Dans certains cas, nous pouvons envisager d'utiliser des tri-grams -- des combin Pour résoudre des tâches comme la classification de texte, nous devons être capables de représenter le texte par un vecteur de taille fixe, que nous utiliserons comme entrée pour le classificateur dense final. Une des façons les plus simples de le faire est de combiner toutes les représentations individuelles des mots, par exemple en les additionnant. Si nous additionnons les encodages one-hot de chaque mot, nous obtiendrons un vecteur de fréquences, montrant combien de fois chaque mot apparaît dans le texte. Une telle représentation du texte est appelée **sac de mots** (BoW). - + > Image par l'auteur diff --git a/translations/fr/lessons/5-NLP/13-TextRep/assignment.md b/translations/fr/lessons/5-NLP/13-TextRep/assignment.md index ed318f0b..c4824c80 100644 --- a/translations/fr/lessons/5-NLP/13-TextRep/assignment.md +++ b/translations/fr/lessons/5-NLP/13-TextRep/assignment.md @@ -1,12 +1,3 @@ - # Devoir : Notebooks En utilisant les notebooks associés à cette leçon (soit la version PyTorch, soit la version TensorFlow), relancez-les en utilisant votre propre jeu de données, peut-être un provenant de Kaggle, en respectant l'attribution. Réécrivez le notebook pour mettre en avant vos propres conclusions. Essayez des jeux de données innovants qui pourraient s'avérer surprenants, comme [celui-ci sur les observations d'OVNIs](https://www.kaggle.com/datasets/NUFORC/ufo-sightings) de NUFORC. diff --git a/translations/fr/lessons/5-NLP/14-Embeddings/README.md b/translations/fr/lessons/5-NLP/14-Embeddings/README.md index 9a2aad22..b159400a 100644 --- a/translations/fr/lessons/5-NLP/14-Embeddings/README.md +++ b/translations/fr/lessons/5-NLP/14-Embeddings/README.md @@ -1,12 +1,3 @@ - # Embeddings ## [Quiz avant le cours](https://ff-quizzes.netlify.app/en/ai/quiz/27) diff --git a/translations/fr/lessons/5-NLP/14-Embeddings/assignment.md b/translations/fr/lessons/5-NLP/14-Embeddings/assignment.md index 49860ab5..653be5e0 100644 --- a/translations/fr/lessons/5-NLP/14-Embeddings/assignment.md +++ b/translations/fr/lessons/5-NLP/14-Embeddings/assignment.md @@ -1,12 +1,3 @@ - # Devoir : Notebooks En utilisant les notebooks associés à cette leçon (soit la version PyTorch, soit la version TensorFlow), exécutez-les à nouveau avec votre propre jeu de données, peut-être un provenant de Kaggle, en respectant les attributions. Réécrivez le notebook pour mettre en avant vos propres observations. Essayez un type de jeu de données différent et documentez vos conclusions, en utilisant par exemple du texte comme [ces paroles des Beatles](https://www.kaggle.com/datasets/jenlooper/beatles-lyrics). diff --git a/translations/fr/lessons/5-NLP/15-LanguageModeling/README.md b/translations/fr/lessons/5-NLP/15-LanguageModeling/README.md index 4f726a37..51b4b8ee 100644 --- a/translations/fr/lessons/5-NLP/15-LanguageModeling/README.md +++ b/translations/fr/lessons/5-NLP/15-LanguageModeling/README.md @@ -1,12 +1,3 @@ - # Modélisation du langage Les embeddings sémantiques, tels que Word2Vec et GloVe, constituent en réalité une première étape vers la **modélisation du langage** - créer des modèles qui *comprennent* (ou *représentent*) d'une certaine manière la nature du langage. diff --git a/translations/fr/lessons/5-NLP/15-LanguageModeling/lab/README.md b/translations/fr/lessons/5-NLP/15-LanguageModeling/lab/README.md index e65b26e2..47091edf 100644 --- a/translations/fr/lessons/5-NLP/15-LanguageModeling/lab/README.md +++ b/translations/fr/lessons/5-NLP/15-LanguageModeling/lab/README.md @@ -1,12 +1,3 @@ - # Entraîner un modèle Skip-Gram Devoir de laboratoire tiré du [Curriculum AI pour débutants](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/fr/lessons/5-NLP/16-RNN/README.md b/translations/fr/lessons/5-NLP/16-RNN/README.md index 8104ce66..8081dbb5 100644 --- a/translations/fr/lessons/5-NLP/16-RNN/README.md +++ b/translations/fr/lessons/5-NLP/16-RNN/README.md @@ -1,12 +1,3 @@ - # Réseaux de Neurones Récurrents ## [Quiz avant le cours](https://ff-quizzes.netlify.app/en/ai/quiz/31) @@ -31,7 +22,7 @@ Voyons comment une cellule RNN simple est organisée. Elle accepte l'état préc Une cellule RNN simple contient deux matrices de poids : l'une transforme un symbole d'entrée (appelons-la W), et l'autre transforme un état d'entrée (H). Dans ce cas, la sortie du réseau est calculée comme σ(W×Xi+H×Si-1+b), où σ est la fonction d'activation et b est un biais supplémentaire. -Anatomie d'une cellule RNN +Anatomie d'une cellule RNN > Image par l'auteur diff --git a/translations/fr/lessons/5-NLP/16-RNN/assignment.md b/translations/fr/lessons/5-NLP/16-RNN/assignment.md index cef382f8..b6fa4eee 100644 --- a/translations/fr/lessons/5-NLP/16-RNN/assignment.md +++ b/translations/fr/lessons/5-NLP/16-RNN/assignment.md @@ -1,12 +1,3 @@ - # Devoir : Notebooks En utilisant les notebooks associés à cette leçon (soit la version PyTorch, soit la version TensorFlow), relancez-les en utilisant votre propre jeu de données, peut-être un provenant de Kaggle, utilisé avec attribution. Réécrivez le notebook pour mettre en évidence vos propres conclusions. Essayez un type de jeu de données différent et documentez vos observations, en utilisant un texte tel que [ce jeu de données de compétition Kaggle sur les tweets météo](https://www.kaggle.com/competitions/crowdflower-weather-twitter/data?select=train.csv). diff --git a/translations/fr/lessons/5-NLP/17-GenerativeNetworks/README.md b/translations/fr/lessons/5-NLP/17-GenerativeNetworks/README.md index e3247dc1..02c40d7f 100644 --- a/translations/fr/lessons/5-NLP/17-GenerativeNetworks/README.md +++ b/translations/fr/lessons/5-NLP/17-GenerativeNetworks/README.md @@ -1,12 +1,3 @@ - # Réseaux génératifs ## [Quiz pré-lecture](https://ff-quizzes.netlify.app/en/ai/quiz/33) @@ -36,7 +27,7 @@ Nous entraînerons ce RNN à générer du texte étape par étape. À chaque ét Lors de la génération de texte (pendant l'inférence), nous commençons par une **invite**, qui est passée à travers les cellules RNN pour générer son état intermédiaire, puis la génération commence à partir de cet état. Nous générons un caractère à la fois, et passons l'état et le caractère généré à une autre cellule RNN pour générer le suivant, jusqu'à ce que nous ayons généré suffisamment de caractères. - + > Image par l'auteur diff --git a/translations/fr/lessons/5-NLP/17-GenerativeNetworks/lab/README.md b/translations/fr/lessons/5-NLP/17-GenerativeNetworks/lab/README.md index 0d9132a1..1bf82dd3 100644 --- a/translations/fr/lessons/5-NLP/17-GenerativeNetworks/lab/README.md +++ b/translations/fr/lessons/5-NLP/17-GenerativeNetworks/lab/README.md @@ -1,12 +1,3 @@ - # Génération de texte au niveau des mots avec des RNN Devoir de laboratoire tiré du [Curriculum AI pour débutants](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/fr/lessons/5-NLP/18-Transformers/README.md b/translations/fr/lessons/5-NLP/18-Transformers/README.md index 608bdf20..f7746aa1 100644 --- a/translations/fr/lessons/5-NLP/18-Transformers/README.md +++ b/translations/fr/lessons/5-NLP/18-Transformers/README.md @@ -1,12 +1,3 @@ - # Mécanismes d'Attention et Transformers ## [Quiz avant le cours](https://ff-quizzes.netlify.app/en/ai/quiz/35) @@ -56,7 +47,7 @@ L'idée de l'encodage positionnel est la suivante : * Embedding entraînable, similaire à l'embedding des tokens. C'est l'approche que nous considérons ici. Nous appliquons des couches d'embedding à la fois sur les tokens et leurs positions, ce qui donne des vecteurs d'embedding de mêmes dimensions, que nous additionnons ensuite. * Fonction d'encodage positionnel fixe, comme proposé dans l'article original. - + > Image par l'auteur diff --git a/translations/fr/lessons/5-NLP/18-Transformers/READMEtransformers.md b/translations/fr/lessons/5-NLP/18-Transformers/READMEtransformers.md deleted file mode 100644 index 3afffed2..00000000 --- a/translations/fr/lessons/5-NLP/18-Transformers/READMEtransformers.md +++ /dev/null @@ -1,112 +0,0 @@ -# Mécanismes d'Attention et Transformateurs - -## [Quiz pré-conférence](https://ff-quizzes.netlify.app/en/ai/quiz/35) - -L'un des problèmes les plus importants dans le domaine du NLP est la **traduction automatique**, une tâche essentielle qui sous-tend des outils tels que Google Translate. Dans cette section, nous allons nous concentrer sur la traduction automatique, ou, plus généralement, sur toute tâche de *séquence à séquence* (qui est également appelée **transduction de phrases**). - -Avec les RNN, la séquence à séquence est mise en œuvre par deux réseaux récurrents, où un réseau, l'**encodeur**, réduit une séquence d'entrée à un état caché, tandis qu'un autre réseau, le **décodeur**, déroule cet état caché en un résultat traduit. Il y a quelques problèmes avec cette approche : - -* L'état final du réseau encodeur a du mal à se souvenir du début d'une phrase, ce qui entraîne une mauvaise qualité du modèle pour les longues phrases. -* Tous les mots d'une séquence ont le même impact sur le résultat. En réalité, cependant, certains mots spécifiques de la séquence d'entrée ont souvent plus d'impact sur les sorties séquentielles que d'autres. - -Les **Mécanismes d'Attention** fournissent un moyen de pondérer l'impact contextuel de chaque vecteur d'entrée sur chaque prédiction de sortie du RNN. La manière dont cela est mis en œuvre consiste à créer des raccourcis entre les états intermédiaires du RNN d'entrée et le RNN de sortie. De cette manière, lors de la génération du symbole de sortie yt, nous prendrons en compte tous les états cachés d'entrée hi, avec des coefficients de poids différents αt,i. - -![Image montrant un modèle encodeur/décodeur avec une couche d'attention additive](../../../../../translated_images/fr/encoder-decoder-attention.7a726296894fb567.webp) - -> Le modèle encodeur-décodeur avec mécanisme d'attention additive dans [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf), cité dans [cet article de blog](https://lilianweng.github.io/lil-log/2018/06/24/attention-attention.html) - -La matrice d'attention {αi,j} représenterait le degré auquel certains mots d'entrée jouent dans la génération d'un mot donné dans la séquence de sortie. Voici un exemple d'une telle matrice : - -![Image montrant un alignement d'exemple trouvé par RNNsearch-50, tiré de Bahdanau - arviz.org](../../../../../translated_images/fr/bahdanau-fig3.09ba2d37f202a6af.webp) - -> Figure de [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf) (Fig.3) - -Les mécanismes d'attention sont responsables de beaucoup de l'état actuel ou presque actuel de l'art dans le NLP. Cependant, l'ajout d'attention augmente considérablement le nombre de paramètres du modèle, ce qui a conduit à des problèmes d'échelle avec les RNN. Une contrainte clé de l'échelle des RNN est que la nature récurrente des modèles rend difficile le traitement par lots et la parallélisation de l'entraînement. Dans un RNN, chaque élément d'une séquence doit être traité dans l'ordre séquentiel, ce qui signifie qu'il ne peut pas être facilement parallélisé. - -![Encodeur Décodeur avec Attention](../../../../../lessons/5-NLP/18-Transformers/images/EncDecAttention.gif) - -> Figure de [Blog de Google](https://research.googleblog.com/2016/09/a-neural-network-for-machine.html) - -L'adoption des mécanismes d'attention combinée à cette contrainte a conduit à la création des modèles de transformateurs qui sont maintenant à la pointe de la technologie que nous connaissons et utilisons aujourd'hui, tels que BERT et Open-GPT3. - -## Modèles de Transformateurs - -L'une des principales idées derrière les transformateurs est d'éviter la nature séquentielle des RNN et de créer un modèle qui est parallélisable pendant l'entraînement. Cela est réalisé en mettant en œuvre deux idées : - -* codage positionnel -* utilisation du mécanisme d'auto-attention pour capturer des motifs au lieu des RNN (ou CNN) (c'est pourquoi l'article qui introduit les transformateurs s'appelle *[Attention is all you need](https://arxiv.org/abs/1706.03762)*) - -### Codage/Emballage Positionnel - -L'idée du codage positionnel est la suivante. -1. Lors de l'utilisation des RNN, la position relative des jetons est représentée par le nombre d'étapes, et n'a donc pas besoin d'être représentée explicitement. -2. Cependant, une fois que nous passons à l'attention, nous devons connaître les positions relatives des jetons dans une séquence. -3. Pour obtenir le codage positionnel, nous augmentons notre séquence de jetons avec une séquence de positions de jetons dans la séquence (c'est-à-dire, une séquence de nombres 0,1, ...). -4. Nous mélangeons ensuite la position du jeton avec un vecteur d'embedding de jeton. Pour transformer la position (entier) en un vecteur, nous pouvons utiliser différentes approches : - -* Embedding entraînable, similaire à l'embedding de jeton. C'est l'approche que nous considérons ici. Nous appliquons des couches d'embedding sur les jetons et leurs positions, ce qui donne des vecteurs d'embedding de mêmes dimensions, que nous additionnons ensuite. -* Fonction de codage de position fixe, comme proposé dans l'article original. - - - -> Image de l'auteur - -Le résultat que nous obtenons avec l'embedding positionnel intègre à la fois le jeton original et sa position dans une séquence. - -### Auto-Attention Multi-Tête - -Ensuite, nous devons capturer certains motifs au sein de notre séquence. Pour ce faire, les transformateurs utilisent un mécanisme d'**auto-attention**, qui est essentiellement une attention appliquée à la même séquence en tant qu'entrée et sortie. L'application de l'auto-attention nous permet de prendre en compte le **contexte** au sein de la phrase, et de voir quels mots sont inter-reliés. Par exemple, cela nous permet de voir quels mots sont référencés par des coréférences, telles que *il*, et également de prendre en compte le contexte : - -![](../../../../../translated_images/fr/CoreferenceResolution.861924d6d384a7d6.webp) - -> Image du [Blog de Google](https://research.googleblog.com/2017/08/transformer-novel-neural-network.html) - -Dans les transformateurs, nous utilisons l'**Attention Multi-Tête** afin de donner au réseau la capacité de capturer plusieurs types de dépendances différentes, par exemple les relations entre mots à long terme par rapport à celles à court terme, les co-références par rapport à autre chose, etc. - -[Notebook TensorFlow](../../../../../lessons/5-NLP/18-Transformers/TransformersTF.ipynb) contient plus de détails sur l'implémentation des couches de transformateurs. - -### Attention Encodeur-Décodeur - -Dans les transformateurs, l'attention est utilisée à deux endroits : - -* Pour capturer des motifs au sein du texte d'entrée en utilisant l'auto-attention -* Pour effectuer la traduction de séquence - c'est la couche d'attention entre l'encodeur et le décodeur. - -L'attention encodeur-décodeur est très similaire au mécanisme d'attention utilisé dans les RNN, comme décrit au début de cette section. Ce diagramme animé explique le rôle de l'attention encodeur-décodeur. - -![GIF animé montrant comment les évaluations sont effectuées dans les modèles de transformateurs.](../../../../../lessons/5-NLP/18-Transformers/images/transformer-animated-explanation.gif) - -Puisque chaque position d'entrée est mappée indépendamment à chaque position de sortie, les transformateurs peuvent mieux se paralléliser que les RNN, ce qui permet des modèles de langage beaucoup plus grands et plus expressifs. Chaque tête d'attention peut être utilisée pour apprendre différentes relations entre les mots, ce qui améliore les tâches de traitement du langage naturel en aval. - -## BERT - -**BERT** (Représentations Encodeur Bidirectionnelles à partir de Transformateurs) est un très grand réseau de transformateurs multicouches avec 12 couches pour *BERT-base*, et 24 pour *BERT-large*. Le modèle est d'abord pré-entraîné sur un grand corpus de données textuelles (WikiPedia + livres) en utilisant un entraînement non supervisé (prédiction de mots masqués dans une phrase). Pendant le pré-entraînement, le modèle absorbe des niveaux significatifs de compréhension linguistique qui peuvent ensuite être exploités avec d'autres ensembles de données en utilisant un ajustement fin. Ce processus est appelé **apprentissage par transfert**. - -![image de http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/fr/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp) - -> Image [source](http://jalammar.github.io/illustrated-bert/) - -## ✍️ Exercices : Transformateurs - -Poursuivez votre apprentissage dans les notebooks suivants : - -* [Transformateurs en PyTorch](../../../../../lessons/5-NLP/18-Transformers/TransformersPyTorch.ipynb) -* [Transformateurs en TensorFlow](../../../../../lessons/5-NLP/18-Transformers/TransformersTF.ipynb) - -## Conclusion - -Dans cette leçon, vous avez appris sur les Transformateurs et les Mécanismes d'Attention, tous des outils essentiels dans la boîte à outils du NLP. Il existe de nombreuses variations d'architectures de Transformateurs, y compris BERT, DistilBERT, BigBird, OpenGPT3 et plus encore, qui peuvent être ajustées. Le [package HuggingFace](https://github.com/huggingface/) fournit un dépôt pour entraîner bon nombre de ces architectures avec PyTorch et TensorFlow. - -## 🚀 Défi - -## [Quiz post-conférence](https://ff-quizzes.netlify.app/en/ai/quiz/36) - -## Revue & Auto-étude - -* [Article de blog](https://mchromiak.github.io/articles/2017/Sep/12/Transformer-Attention-is-all-you-need/), expliquant l'article classique [Attention is all you need](https://arxiv.org/abs/1706.03762) sur les transformateurs. -* [Une série d'articles de blog](https://towardsdatascience.com/transformers-explained-visually-part-1-overview-of-functionality-95a6dd460452) sur les transformateurs, expliquant l'architecture en détail. - -## [Devoir](assignment.md) - -**Avertissement** : -Ce document a été traduit à l'aide de services de traduction automatique basés sur l'IA. Bien que nous nous efforçons d'assurer l'exactitude, veuillez noter que les traductions automatiques peuvent contenir des erreurs ou des inexactitudes. Le document original dans sa langue native doit être considéré comme la source autoritaire. Pour des informations critiques, une traduction professionnelle par un humain est recommandée. Nous ne sommes pas responsables des malentendus ou des interprétations erronées résultant de l'utilisation de cette traduction. \ No newline at end of file diff --git a/translations/fr/lessons/5-NLP/18-Transformers/assignment.md b/translations/fr/lessons/5-NLP/18-Transformers/assignment.md index c7a11568..3694f859 100644 --- a/translations/fr/lessons/5-NLP/18-Transformers/assignment.md +++ b/translations/fr/lessons/5-NLP/18-Transformers/assignment.md @@ -1,12 +1,3 @@ - # Devoir : Transformers Expérimentez avec les Transformers sur HuggingFace ! Essayez certains des scripts qu'ils proposent pour travailler avec les différents modèles disponibles sur leur site : https://huggingface.co/docs/transformers/run_scripts. Testez l'un de leurs jeux de données, puis importez-en un de ce programme ou de Kaggle et voyez si vous pouvez générer des textes intéressants. Produisez un notebook avec vos conclusions. diff --git a/translations/fr/lessons/5-NLP/19-NER/README.md b/translations/fr/lessons/5-NLP/19-NER/README.md index 8205a377..e23e2252 100644 --- a/translations/fr/lessons/5-NLP/19-NER/README.md +++ b/translations/fr/lessons/5-NLP/19-NER/README.md @@ -1,12 +1,3 @@ - # Reconnaissance d'entités nommées Jusqu'à présent, nous nous sommes principalement concentrés sur une tâche de traitement du langage naturel (NLP) : la classification. Cependant, il existe d'autres tâches NLP qui peuvent être réalisées avec des réseaux neuronaux. L'une de ces tâches est la **[Reconnaissance d'entités nommées](https://wikipedia.org/wiki/Named-entity_recognition)** (NER), qui consiste à identifier des entités spécifiques dans un texte, telles que des lieux, des noms de personnes, des intervalles de date-heure, des formules chimiques, etc. @@ -17,7 +8,7 @@ Jusqu'à présent, nous nous sommes principalement concentrés sur une tâche de Supposons que vous souhaitiez développer un chatbot en langage naturel, similaire à Amazon Alexa ou Google Assistant. Les chatbots intelligents fonctionnent en *comprenant* ce que l'utilisateur veut grâce à la classification de texte sur la phrase d'entrée. Le résultat de cette classification est ce qu'on appelle **l'intention**, qui détermine ce que le chatbot doit faire. -Bot NER +Bot NER > Image par l'auteur diff --git a/translations/fr/lessons/5-NLP/19-NER/lab/README.md b/translations/fr/lessons/5-NLP/19-NER/lab/README.md index 57102e1a..1f8ec14f 100644 --- a/translations/fr/lessons/5-NLP/19-NER/lab/README.md +++ b/translations/fr/lessons/5-NLP/19-NER/lab/README.md @@ -1,12 +1,3 @@ - # NER Travail pratique issu du [programme AI for Beginners](https://github.com/microsoft/ai-for-beginners). diff --git a/translations/fr/lessons/5-NLP/20-LangModels/README.md b/translations/fr/lessons/5-NLP/20-LangModels/README.md index 10e555bb..19301624 100644 --- a/translations/fr/lessons/5-NLP/20-LangModels/README.md +++ b/translations/fr/lessons/5-NLP/20-LangModels/README.md @@ -1,12 +1,3 @@ - # Modèles de Langage Pré-entraînés Dans toutes nos tâches précédentes, nous entraînions un réseau neuronal pour accomplir une tâche spécifique en utilisant un ensemble de données étiquetées. Avec les grands modèles de transformateurs, tels que BERT, nous utilisons la modélisation linguistique de manière auto-supervisée pour construire un modèle de langage, qui est ensuite spécialisé pour une tâche spécifique grâce à un entraînement supplémentaire adapté au domaine. Cependant, il a été démontré que les grands modèles de langage peuvent également résoudre de nombreuses tâches sans AUCUN entraînement spécifique au domaine. Une famille de modèles capables de faire cela est appelée **GPT** : Generative Pre-Trained Transformer. diff --git a/translations/fr/lessons/5-NLP/20-LangModels/READMELargeLang.md b/translations/fr/lessons/5-NLP/20-LangModels/READMELargeLang.md deleted file mode 100644 index fb0e96f9..00000000 --- a/translations/fr/lessons/5-NLP/20-LangModels/READMELargeLang.md +++ /dev/null @@ -1,56 +0,0 @@ -# Modèles de Langage de Grande Taille Pré-entrainés - -Dans toutes nos tâches précédentes, nous avons entraîné un réseau de neurones pour effectuer une tâche spécifique en utilisant un ensemble de données étiquetées. Avec de grands modèles de transformateurs, tels que BERT, nous utilisons la modélisation du langage de manière auto-supervisée pour construire un modèle de langage, qui est ensuite spécialisé pour une tâche spécifique en aval grâce à un entraînement supplémentaire spécifique au domaine. Cependant, il a été démontré que de grands modèles de langage peuvent également résoudre de nombreuses tâches sans AUCUN entraînement spécifique au domaine. Une famille de modèles capables de le faire est appelée **GPT** : Transformateur Génératif Pré-entrainé. - -## [Quiz pré-conférence](https://ff-quizzes.netlify.app/en/ai/quiz/39) - -## Génération de Texte et Perplexité - -L'idée qu'un réseau de neurones puisse accomplir des tâches générales sans entraînement en aval est présentée dans l'article [Les Modèles de Langage sont des Apprenants Multitâches Non Supervisés](https://cdn.openai.com/better-language-models/language_models_are_unsupervised_multitask_learners.pdf). L'idée principale est que de nombreuses autres tâches peuvent être modélisées en utilisant **la génération de texte**, car comprendre le texte signifie essentiellement être capable de le produire. Comme le modèle est entraîné sur une énorme quantité de texte qui englobe les connaissances humaines, il devient également informé sur une grande variété de sujets. - -> Comprendre et être capable de produire du texte implique également de connaître quelque chose sur le monde qui nous entoure. Les gens apprennent également beaucoup en lisant, et le réseau GPT est similaire à cet égard. - -Les réseaux de génération de texte fonctionnent en prédisant la probabilité du mot suivant $$P(w_N)$$. Cependant, la probabilité inconditionnelle du mot suivant est égale à la fréquence de ce mot dans le corpus de texte. GPT est capable de nous donner la **probabilité conditionnelle** du mot suivant, étant donné les précédents : $$P(w_N | w_{n-1}, ..., w_0)$$ - -> Vous pouvez en savoir plus sur les probabilités dans notre [Curriculum de Science des Données pour Débutants](https://github.com/microsoft/Data-Science-For-Beginners/tree/main/1-Introduction/04-stats-and-probability) - -La qualité d'un modèle de génération de langage peut être définie à l'aide de la **perplexité**. C'est une métrique intrinsèque qui nous permet de mesurer la qualité du modèle sans aucun ensemble de données spécifique à une tâche. Elle repose sur la notion de *probabilité d'une phrase* - le modèle attribue une forte probabilité à une phrase qui est susceptible d'être réelle (c'est-à-dire que le modèle n'est pas **perplexe** par rapport à elle), et une faible probabilité aux phrases qui ont moins de sens (par exemple, *Peut-il faire quoi ?*). Lorsque nous donnons à notre modèle des phrases provenant d'un corpus de texte réel, nous nous attendons à ce qu'elles aient une forte probabilité et une faible **perplexité**. Mathématiquement, elle est définie comme la probabilité inverse normalisée de l'ensemble de test : -$$ -\mathrm{Perplexity}(W) = \sqrt[N]{1\over P(W_1,...,W_N)} -$$ - -**Vous pouvez expérimenter avec la génération de texte en utilisant [l'éditeur de texte alimenté par GPT de Hugging Face](https://transformer.huggingface.co/doc/gpt2-large)**. Dans cet éditeur, vous commencez à écrire votre texte, et en appuyant sur **[TAB]**, plusieurs options de complétion vous seront proposées. Si elles sont trop courtes, ou si vous n'êtes pas satisfait, appuyez à nouveau sur [TAB], et vous aurez plus d'options, y compris des morceaux de texte plus longs. - -## GPT est une Famille - -GPT n'est pas un modèle unique, mais plutôt une collection de modèles développés et entraînés par [OpenAI](https://openai.com). - -Sous les modèles GPT, nous avons : - -| [GPT-2](https://huggingface.co/docs/transformers/model_doc/gpt2#openai-gpt2) | [GPT 3](https://openai.com/research/language-models-are-few-shot-learners) | [GPT-4](https://openai.com/gpt-4) | -| -- | -- | -- | -| Modèle de langage avec jusqu'à 1,5 milliard de paramètres. | Modèle de langage avec jusqu'à 175 milliards de paramètres | 100T paramètres et accepte à la fois des entrées d'image et de texte et produit du texte. | - - -Les modèles GPT-3 et GPT-4 sont disponibles [en tant que service cognitif de Microsoft Azure](https://azure.microsoft.com/en-us/services/cognitive-services/openai-service/#overview?WT.mc_id=academic-77998-cacaste), et en tant qu'[API OpenAI](https://openai.com/api/). - -## Ingénierie des Prompts - -Parce que GPT a été entraîné sur d'énormes volumes de données pour comprendre le langage et le code, il fournit des sorties en réponse à des entrées (prompts). Les prompts sont des entrées ou des requêtes GPT par lesquelles on fournit des instructions aux modèles sur les tâches qu'ils doivent ensuite accomplir. Pour obtenir un résultat souhaité, vous avez besoin du prompt le plus efficace, ce qui implique de sélectionner les bons mots, formats, phrases ou même symboles. Cette approche est appelée [Ingénierie des Prompts](https://learn.microsoft.com/en-us/shows/ai-show/the-basics-of-prompt-engineering-with-azure-openai-service?WT.mc_id=academic-77998-bethanycheum). - -[Cette documentation](https://learn.microsoft.com/en-us/semantic-kernel/prompt-engineering/?WT.mc_id=academic-77998-bethanycheum) vous fournit plus d'informations sur l'ingénierie des prompts. - -## ✍️ Exemple de Notebook : [Jouer avec OpenAI-GPT](../../../../../lessons/5-NLP/20-LangModels/GPT-PyTorch.ipynb) - -Continuez votre apprentissage dans les notebooks suivants : - -* [Générer du texte avec OpenAI-GPT et Hugging Face Transformers](../../../../../lessons/5-NLP/20-LangModels/GPT-PyTorch.ipynb) - -## Conclusion - -Les nouveaux modèles de langage pré-entraînés généraux ne modélisent pas seulement la structure du langage, mais contiennent également une vaste quantité de langage naturel. Ainsi, ils peuvent être utilisés efficacement pour résoudre certaines tâches de traitement du langage naturel dans des contextes zéro-shot ou few-shot. - -## [Quiz post-conférence](https://ff-quizzes.netlify.app/en/ai/quiz/40) - -**Avertissement** : -Ce document a été traduit à l'aide de services de traduction automatique basés sur l'IA. Bien que nous nous efforçons d'assurer l'exactitude, veuillez noter que les traductions automatiques peuvent contenir des erreurs ou des inexactitudes. Le document original dans sa langue native doit être considéré comme la source autoritaire. Pour des informations critiques, une traduction humaine professionnelle est recommandée. Nous ne sommes pas responsables des malentendus ou des interprétations erronées résultant de l'utilisation de cette traduction. \ No newline at end of file diff --git a/translations/fr/lessons/5-NLP/README.md b/translations/fr/lessons/5-NLP/README.md index 7d88a2a6..233e2add 100644 --- a/translations/fr/lessons/5-NLP/README.md +++ b/translations/fr/lessons/5-NLP/README.md @@ -1,12 +1,3 @@ - # Traitement du Langage Naturel ![Résumé des tâches NLP dans un croquis](../../../../translated_images/fr/ai-nlp.b22dcb8ca4707cea.webp) diff --git a/translations/fr/lessons/6-Other/21-GeneticAlgorithms/README.md b/translations/fr/lessons/6-Other/21-GeneticAlgorithms/README.md index 269a4b61..e83d2706 100644 --- a/translations/fr/lessons/6-Other/21-GeneticAlgorithms/README.md +++ b/translations/fr/lessons/6-Other/21-GeneticAlgorithms/README.md @@ -1,12 +1,3 @@ - # Algorithmes Génétiques ## [Quiz avant le cours](https://ff-quizzes.netlify.app/en/ai/quiz/41) diff --git a/translations/fr/lessons/6-Other/22-DeepRL/README.md b/translations/fr/lessons/6-Other/22-DeepRL/README.md index 94bde1ba..d9262c80 100644 --- a/translations/fr/lessons/6-Other/22-DeepRL/README.md +++ b/translations/fr/lessons/6-Other/22-DeepRL/README.md @@ -1,12 +1,3 @@ - # Apprentissage par renforcement profond L'apprentissage par renforcement (RL) est considéré comme l'un des paradigmes fondamentaux de l'apprentissage automatique, aux côtés de l'apprentissage supervisé et non supervisé. Alors que l'apprentissage supervisé repose sur un ensemble de données avec des résultats connus, le RL est basé sur **l'apprentissage par l'expérience**. Par exemple, lorsque nous découvrons un jeu vidéo pour la première fois, nous commençons à jouer, même sans connaître les règles, et nous améliorons rapidement nos compétences simplement en jouant et en ajustant notre comportement. @@ -34,7 +25,7 @@ Vous avez probablement déjà vu des dispositifs modernes d'équilibrage tels qu Une version simplifiée de l'équilibrage est connue sous le nom de problème **CartPole**. Dans le monde du CartPole, nous avons un curseur horizontal qui peut se déplacer à gauche ou à droite, et l'objectif est d'équilibrer un poteau vertical au-dessus du curseur pendant qu'il se déplace. -un CartPole +un CartPole Pour créer et utiliser cet environnement, nous avons besoin de quelques lignes de code Python : diff --git a/translations/fr/lessons/6-Other/22-DeepRL/lab/README.md b/translations/fr/lessons/6-Other/22-DeepRL/lab/README.md index acea4dea..11f68d58 100644 --- a/translations/fr/lessons/6-Other/22-DeepRL/lab/README.md +++ b/translations/fr/lessons/6-Other/22-DeepRL/lab/README.md @@ -1,12 +1,3 @@ - ## L'Environnement L'environnement Mountain Car consiste en une voiture piégée dans une vallée. Votre objectif est de sortir de la vallée et d'atteindre le drapeau. Les actions que vous pouvez effectuer sont : accélérer vers la gauche, accélérer vers la droite, ou ne rien faire. Vous pouvez observer la position de la voiture sur l'axe x, ainsi que sa vitesse. diff --git a/translations/fr/lessons/6-Other/23-MultiagentSystems/README.md b/translations/fr/lessons/6-Other/23-MultiagentSystems/README.md index 3b36f698..2e7325c2 100644 --- a/translations/fr/lessons/6-Other/23-MultiagentSystems/README.md +++ b/translations/fr/lessons/6-Other/23-MultiagentSystems/README.md @@ -1,12 +1,3 @@ - # Systèmes Multi-Agents Une des approches possibles pour atteindre l'intelligence est l'approche dite **émergente** (ou **synergétique**), qui repose sur le fait que le comportement combiné de nombreux agents relativement simples peut aboutir à un comportement global plus complexe (ou intelligent) du système dans son ensemble. Théoriquement, cela repose sur les principes de l'[Intelligence Collective](https://fr.wikipedia.org/wiki/Intelligence_collective), de l'[Émergentisme](https://fr.wikipedia.org/wiki/Cerveau_global) et de la [Cybernétique Évolutionnaire](https://fr.wikipedia.org/wiki/Cerveau_global), qui stipulent que les systèmes de niveau supérieur acquièrent une certaine valeur ajoutée lorsqu'ils sont correctement combinés à partir de systèmes de niveau inférieur (le *principe de transition méta-systémique*). @@ -60,7 +51,7 @@ Vous pouvez [télécharger](https://ccl.northwestern.edu/netlogo/download.shtml) Un des grands avantages de NetLogo est qu'il contient une bibliothèque de modèles fonctionnels que vous pouvez essayer. Allez dans **File → Models Library**, et vous trouverez de nombreuses catégories de modèles à explorer. -Bibliothèque de modèles NetLogo +Bibliothèque de modèles NetLogo > Une capture d'écran de la bibliothèque de modèles par Dmitry Soshnikov diff --git a/translations/fr/lessons/6-Other/23-MultiagentSystems/assignment.md b/translations/fr/lessons/6-Other/23-MultiagentSystems/assignment.md index 2d494ea0..7b85c400 100644 --- a/translations/fr/lessons/6-Other/23-MultiagentSystems/assignment.md +++ b/translations/fr/lessons/6-Other/23-MultiagentSystems/assignment.md @@ -1,12 +1,3 @@ - # Devoir NetLogo Prenez l'un des modèles de la bibliothèque de NetLogo et utilisez-le pour simuler une situation réelle aussi fidèlement que possible. Un bon exemple serait de modifier le modèle Virus dans le dossier Alternative Visualizations pour montrer comment il peut être utilisé pour modéliser la propagation du COVID-19. Pouvez-vous créer un modèle qui imite la propagation réelle d'un virus ? diff --git a/translations/fr/lessons/7-Ethics/README.md b/translations/fr/lessons/7-Ethics/README.md index a49dafb1..1ea690d7 100644 --- a/translations/fr/lessons/7-Ethics/README.md +++ b/translations/fr/lessons/7-Ethics/README.md @@ -1,12 +1,3 @@ - # IA Éthique et Responsable Vous avez presque terminé ce cours, et j'espère qu'à ce stade, vous comprenez clairement que l'IA repose sur un certain nombre de méthodes mathématiques formelles qui nous permettent de trouver des relations dans les données et de former des modèles pour reproduire certains aspects du comportement humain. À ce moment de l'histoire, nous considérons l'IA comme un outil très puissant pour extraire des motifs des données et appliquer ces motifs pour résoudre de nouveaux problèmes. diff --git a/translations/fr/lessons/README.md b/translations/fr/lessons/README.md index 72bc1a7b..d293ad60 100644 --- a/translations/fr/lessons/README.md +++ b/translations/fr/lessons/README.md @@ -1,12 +1,3 @@ - # Vue d'ensemble ![Vue d'ensemble dans un croquis](../../../translated_images/fr/ai-overview.0857791951d19500.webp) diff --git a/translations/fr/lessons/X-Extras/X1-MultiModal/README.md b/translations/fr/lessons/X-Extras/X1-MultiModal/README.md index 3baf1218..089dcca4 100644 --- a/translations/fr/lessons/X-Extras/X1-MultiModal/README.md +++ b/translations/fr/lessons/X-Extras/X1-MultiModal/README.md @@ -1,12 +1,3 @@ - # Réseaux Multi-Modalités Après le succès des modèles transformateurs pour résoudre des tâches de traitement du langage naturel (NLP), les mêmes architectures ou des architectures similaires ont été appliquées aux tâches de vision par ordinateur. Il y a un intérêt croissant pour la création de modèles qui *combinent* les capacités de vision et de langage naturel. Une des tentatives dans ce domaine a été réalisée par OpenAI, et elle s'appelle CLIP et DALL.E. diff --git a/translations/fr/lessons/sketchnotes/LICENSE.md b/translations/fr/lessons/sketchnotes/LICENSE.md index 0a8e1d25..a388642d 100644 --- a/translations/fr/lessons/sketchnotes/LICENSE.md +++ b/translations/fr/lessons/sketchnotes/LICENSE.md @@ -1,12 +1,3 @@ - Attribution-ShareAlike 4.0 International ======================================================================= diff --git a/translations/fr/lessons/sketchnotes/README.md b/translations/fr/lessons/sketchnotes/README.md index e9b93c07..06492639 100644 --- a/translations/fr/lessons/sketchnotes/README.md +++ b/translations/fr/lessons/sketchnotes/README.md @@ -1,12 +1,3 @@ - Tous les sketchnotes du programme peuvent être téléchargés ici. 🎨 Créé par : Tomomi Imura (Twitter : [@girlie_mac](https://twitter.com/girlie_mac), GitHub : [girliemac](https://github.com/girliemac)) diff --git a/translations/fr/troubleshoot.md b/translations/fr/troubleshoot.md index 1bfab3e6..e5915857 100644 --- a/translations/fr/troubleshoot.md +++ b/translations/fr/troubleshoot.md @@ -1,12 +1,3 @@ - # Guide de dépannage pour AI-For-Beginners Ce guide vous aide à résoudre les problèmes courants rencontrés lors de l'utilisation ou de la contribution au dépôt [AI-For-Beginners](https://github.com/microsoft/AI-For-Beginners). Chaque problème inclut des informations de base, des symptômes, des explications et des solutions étape par étape. diff --git a/translations/hk/README.md b/translations/hk/README.md index 3372aeaa..e6ab0115 100644 --- a/translations/hk/README.md +++ b/translations/hk/README.md @@ -23,7 +23,7 @@ CO_OP_TRANSLATOR_METADATA: # 人工智能入門課程 -|![Sketchnote by @girlie_mac https://twitter.com/girlie_mac](../../../../translated_images/hk/ai-overview.0857791951d19500.webp)| +|![Sketchnote by @girlie_mac https://twitter.com/girlie_mac](../../../../translated_images/zh-HK/ai-overview.0857791951d19500.webp)| |:---:| | AI For Beginners - _手繪筆記由 [@girlie_mac](https://twitter.com/girlie_mac) 製作_ | diff --git a/translations/hk/lessons/1-Intro/README.md b/translations/hk/lessons/1-Intro/README.md index 29df50e5..4fdda6bb 100644 --- a/translations/hk/lessons/1-Intro/README.md +++ b/translations/hk/lessons/1-Intro/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 人工智能簡介 -![人工智能簡介內容的手繪圖](../../../../translated_images/hk/ai-intro.bf28d1ac4235881c.webp) +![人工智能簡介內容的手繪圖](../../../../translated_images/zh-HK/ai-intro.bf28d1ac4235881c.webp) > 手繪筆記由 [Tomomi Imura](https://twitter.com/girlie_mac) 提供 @@ -19,7 +19,7 @@ CO_OP_TRANSLATOR_METADATA: 最初，電腦是由 [Charles Babbage](https://en.wikipedia.org/wiki/Charles_Babbage) 發明的，用於按照明確定義的程序（即算法）處理數字。現代電腦雖然比19世紀提出的原型先進得多，但仍然遵循受控計算的理念。因此，如果我們知道實現目標所需的精確步驟，就可以編程讓電腦完成某些事情。 -![一個人的照片](../../../../translated_images/hk/dsh_age.d212a30d4e54fb5f.webp) +![一個人的照片](../../../../translated_images/zh-HK/dsh_age.d212a30d4e54fb5f.webp) > 照片由 [Vickie Soshnikova](http://twitter.com/vickievalerie) 提供 @@ -46,7 +46,7 @@ CO_OP_TRANSLATOR_METADATA: 當討論 **[智能](https://en.wikipedia.org/wiki/Intelligence)** 這個術語時，問題之一是我們對這個術語沒有明確的定義。有人認為智能與 **抽象思維** 或 **自我意識** 有關，但我們無法準確定義它。 -![一隻貓的照片](../../../../translated_images/hk/photo-cat.8c8e8fb760ffe457.webp) +![一隻貓的照片](../../../../translated_images/zh-HK/photo-cat.8c8e8fb760ffe457.webp) > [照片](https://unsplash.com/photos/75715CVEJhI) 由 [Amber Kipp](https://unsplash.com/@sadmax) 提供，來自 Unsplash @@ -98,13 +98,13 @@ CO_OP_TRANSLATOR_METADATA: > | 那麼機器學習呢？ | | > |--------------|-----------| -> | 基於電腦通過一些數據學習解決問題的人工智能部分被稱為 **機器學習**。我們不會在本課程中考慮經典的機器學習——我們建議你參考單獨的 [機器學習初學者課程](http://aka.ms/ml-beginners)。 | ![機器學習初學者課程](../../../../translated_images/hk/ml-for-beginners.9e4fed176fd5817d.webp) | +> | 基於電腦通過一些數據學習解決問題的人工智能部分被稱為 **機器學習**。我們不會在本課程中考慮經典的機器學習——我們建議你參考單獨的 [機器學習初學者課程](http://aka.ms/ml-beginners)。 | ![機器學習初學者課程](../../../../translated_images/zh-HK/ml-for-beginners.9e4fed176fd5817d.webp) | ## 人工智能的簡史 人工智能作為一個領域始於20世紀中葉。最初，符號推理是一種流行的方法，並且它帶來了一些重要的成功，例如專家系統——能夠在某些有限問題領域中充當專家的電腦程序。然而，很快就發現這種方法並不適用於大規模應用。從專家那裡提取知識、將其表示在電腦中並保持知識庫的準確性，事實證明這是一項非常複雜且在許多情況下成本過高的任務。這導致了20世紀70年代所謂的 [人工智能寒冬](https://en.wikipedia.org/wiki/AI_winter)。 -人工智能簡史 +人工智能簡史 > 圖片由 [Dmitry Soshnikov](http://soshnikov.com) 提供 @@ -124,7 +124,7 @@ CO_OP_TRANSLATOR_METADATA: * 現代助手，例如 Cortana、Siri 或 Google Assistant，都是混合系統，使用神經網絡將語音轉換為文本並識別我們的意圖，然後使用一些推理或明確的算法執行所需的操作。 * 未來，我們可能會期待一個完全基於神經網絡的模型能夠自行處理對話。最近的 GPT 和 [Turing-NLG](https://www.microsoft.com/research/blog/turing-nlg-a-17-billion-parameter-language-model-by-microsoft) 系列神經網絡在這方面表現出色。 -圖靈測試的演變 +圖靈測試的演變 > 圖片由 Dmitry Soshnikov 提供，[照片](https://unsplash.com/photos/r8LmVbUKgns) 由 [Marina Abrosimova](https://unsplash.com/@abrosimova_marina_foto) 提供，Unsplash ## 最近的人工智能研究 diff --git a/translations/hk/lessons/2-Symbolic/Animals.ipynb b/translations/hk/lessons/2-Symbolic/Animals.ipynb index 11e03dc5..4a18b97a 100644 --- a/translations/hk/lessons/2-Symbolic/Animals.ipynb +++ b/translations/hk/lessons/2-Symbolic/Animals.ipynb @@ -12,7 +12,7 @@ "\n", "在此範例中，我們將實作一個簡單的基於知識的系統，以根據一些物理特徵判斷動物。該系統可由以下 AND-OR 樹表示（這是整棵樹的一部分，我們可以輕易新增更多規則）：\n", "\n", - "![](../../../../../../translated_images/hk/AND-OR-Tree.5592d2c70187f283.webp)\n" + "![](../../../../../../translated_images/zh-HK/AND-OR-Tree.5592d2c70187f283.webp)\n" ] }, { diff --git a/translations/hk/lessons/2-Symbolic/README.md b/translations/hk/lessons/2-Symbolic/README.md index d15f51f2..7eaba2b0 100644 --- a/translations/hk/lessons/2-Symbolic/README.md +++ b/translations/hk/lessons/2-Symbolic/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 知識表徵與專家系統 -![Symbolic AI 內容摘要](../../../../../../translated_images/hk/ai-symbolic.715a30cb610411a6.webp) +![Symbolic AI 內容摘要](../../../../../../translated_images/zh-HK/ai-symbolic.715a30cb610411a6.webp) > 速寫筆記作者：[Tomomi Imura](https://twitter.com/girlie_mac) @@ -35,13 +35,13 @@ CO_OP_TRANSLATOR_METADATA: * **知識**是信息被整合進我們的世界模型。例如，一旦了解電腦的概念，我們開始知道它的運作方式、成本和用途。這一互相關聯的概念網形成了我們的知識。 * **智慧**則是對世界理解的更高層次，代表*元知識*，例如什麼時候以及如何使用該知識的概念。 - + *圖片來源：[維基百科](https://commons.wikimedia.org/w/index.php?curid=37705247)，作者 Longlivetheux，CC BY-SA 4.0* 因此，**知識表徵**問題就是尋找某種有效方式在計算機中以數據形式表示知識，讓它自動可用。這可以被視為一個光譜： -![知識表徵光譜](../../../../../../translated_images/hk/knowledge-spectrum.b60df631852c0217.webp) +![知識表徵光譜](../../../../../../translated_images/zh-HK/knowledge-spectrum.b60df631852c0217.webp) > 圖片作者：[Dmitry Soshnikov](http://soshnikov.com) @@ -94,7 +94,7 @@ Python | 區塊語法 | 縮排 符號 AI 的早期成功之一是所謂的**專家系統**－設計用於有限問題領域做專家決策的計算機系統。基於從一位或多位專家手動提取的**知識庫**，並包含執行推理的**推理引擎**。 -![人類架構](../../../../../../translated_images/hk/arch-human.5d4d35f1bba3ab1c.webp) | ![知識系統架構](../../../../../../translated_images/hk/arch-kbs.3ec5c150b09fa8da.webp) +![人類架構](../../../../../../translated_images/zh-HK/arch-human.5d4d35f1bba3ab1c.webp) | ![知識系統架構](../../../../../../translated_images/zh-HK/arch-kbs.3ec5c150b09fa8da.webp) ------------------------------|------------------------------ 簡化的人類神經系統結構 | 知識系統架構 @@ -106,7 +106,7 @@ Python | 區塊語法 | 縮排 舉例，考慮以下依據動物物理特徵判斷的專家系統： -![AND-OR 樹](../../../../../../translated_images/hk/AND-OR-Tree.5592d2c70187f283.webp) +![AND-OR 樹](../../../../../../translated_images/zh-HK/AND-OR-Tree.5592d2c70187f283.webp) > 圖片作者：[Dmitry Soshnikov](http://soshnikov.com) @@ -175,7 +175,7 @@ THEN the animal is a carnivore 在語意網中，所有表示皆基於三元組。每個物件和每個關係都由 URI 唯一標識。例如，如果我們想表述這個 AI 課程是由 Dmitry Soshnikov 於 2022 年 1 月 1 日開發的——以下是我們可以使用的三元組： - + ``` http://github.com/microsoft/ai-for-beginners http://www.example.com/terms/creation-date “Jan 1, 2022” @@ -186,7 +186,7 @@ http://github.com/microsoft/ai-for-beginners http://purl.org/dc/elements/1.1/cre 在更加複雜的情況下，如果我們想定義一個創建者列表，可以使用 RDF 中定義的一些資料結構。 - + > 以上圖表由 [Dmitry Soshnikov](http://soshnikov.com) 提供 @@ -210,7 +210,7 @@ GROUP BY ?eyeColorLabel > ✅ 如果你想嘗試建立自己的本體，或打開現有本體，有一個很棒的視覺本體編輯器叫做 [Protégé](https://protege.stanford.edu/)。下載它，或在線使用。 - + *Web Protégé 編輯器開啟羅曼諾夫家族本體。截圖由 Dmitry Soshnikov 提供* diff --git a/translations/hk/lessons/3-NeuralNetworks/05-Frameworks/README.md b/translations/hk/lessons/3-NeuralNetworks/05-Frameworks/README.md index 094c6e07..1b835f61 100644 --- a/translations/hk/lessons/3-NeuralNetworks/05-Frameworks/README.md +++ b/translations/hk/lessons/3-NeuralNetworks/05-Frameworks/README.md @@ -58,7 +58,7 @@ CO_OP_TRANSLATOR_METADATA: 考慮以下近似 5 個點（圖中的 `x`）的問題： -![線性模型](../../../../../translated_images/hk/overfit1.f24b71c6f652e59e.webp) | ![過擬合模型](../../../../../translated_images/hk/overfit2.131f5800ae10ca5e.webp) +![線性模型](../../../../../translated_images/zh-HK/overfit1.f24b71c6f652e59e.webp) | ![過擬合模型](../../../../../translated_images/zh-HK/overfit2.131f5800ae10ca5e.webp) -------------------------|-------------------------- **線性模型，2 個參數** | **非線性模型，7 個參數** 訓練誤差 = 5.3 | 訓練誤差 = 0 @@ -79,7 +79,7 @@ CO_OP_TRANSLATOR_METADATA: 如上圖所示，過擬合可以通過非常低的訓練誤差和非常高的驗證誤差來檢測。通常在訓練過程中，我們會看到訓練和驗證誤差都開始下降，然後在某個時候驗證誤差可能停止下降並開始上升。這就是過擬合的跡象，表明我們可能應該停止訓練（或者至少保存模型的快照）。 -![過擬合](../../../../../translated_images/hk/Overfitting.408ad91cd90b4371.webp) +![過擬合](../../../../../translated_images/zh-HK/Overfitting.408ad91cd90b4371.webp) ## 如何防止過擬合 diff --git a/translations/hk/lessons/3-NeuralNetworks/README.md b/translations/hk/lessons/3-NeuralNetworks/README.md index 0435a69a..c35eb194 100644 --- a/translations/hk/lessons/3-NeuralNetworks/README.md +++ b/translations/hk/lessons/3-NeuralNetworks/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 神經網絡簡介 -![神經網絡簡介內容的手繪圖](../../../../translated_images/hk/ai-neuralnetworks.1c687ae40bc86e83.webp) +![神經網絡簡介內容的手繪圖](../../../../translated_images/zh-HK/ai-neuralnetworks.1c687ae40bc86e83.webp) 正如我們在介紹中所討論的，實現智能的一種方法是訓練一個**計算機模型**或**人工大腦**。自20世紀中期以來，研究人員嘗試了不同的數學模型，直到近年來這一方向證明非常成功。這些模仿大腦的數學模型被稱為**神經網絡**。 @@ -36,13 +36,13 @@ CO_OP_TRANSLATOR_METADATA: 從生物學中我們知道，大腦由神經細胞（神經元）組成，每個神經元有多個“輸入”（樹突）和一個“輸出”（軸突）。樹突和軸突都可以傳導電信號，而它們之間的連接——稱為突觸——可以表現出不同程度的導電性，這由神經遞質調節。 -![神經元模型](../../../../translated_images/hk/synapse-wikipedia.ed20a9e4726ea1c6.webp) | ![神經元模型](../../../../translated_images/hk/artneuron.1a5daa88d20ebe6f.webp) +![神經元模型](../../../../translated_images/zh-HK/synapse-wikipedia.ed20a9e4726ea1c6.webp) | ![神經元模型](../../../../translated_images/zh-HK/artneuron.1a5daa88d20ebe6f.webp) ----|---- 真實神經元 *（[圖片](https://en.wikipedia.org/wiki/Synapse#/media/File:SynapseSchematic_lines.svg) 來自維基百科）* | 人工神經元 *（作者提供圖片）* 因此，神經元的最簡單數學模型包含若干輸入 X1, ..., XN 和一個輸出 Y，以及一系列權重 W1, ..., WN。輸出計算公式為： -Y = f\left(\sum_{i=1}^N X_iW_i\right) +Y = f\left(\sum_{i=1}^N X_iW_i\right) 其中 f 是某種非線性的**激活函數**。 diff --git a/translations/hk/lessons/4-ComputerVision/06-IntroCV/README.md b/translations/hk/lessons/4-ComputerVision/06-IntroCV/README.md index ca2e3bcb..54b86db8 100644 --- a/translations/hk/lessons/4-ComputerVision/06-IntroCV/README.md +++ b/translations/hk/lessons/4-ComputerVision/06-IntroCV/README.md @@ -75,14 +75,14 @@ im = cv2.cvtColor(im,cv2.COLOR_BGR2RGB) * **預處理盲文書籍的照片**。我們專注於如何使用閾值處理、特徵檢測、透視變換和 NumPy 操作來分離單個盲文符號，以便神經網絡進一步分類。 -![盲文圖像](../../../../../translated_images/hk/braille.341962ff76b1bd70.webp) | ![盲文圖像預處理](../../../../../translated_images/hk/braille-result.46530fea020b03c7.webp) | ![盲文符號](../../../../../translated_images/hk/braille-symbols.0159185ab69d5339.webp) +![盲文圖像](../../../../../translated_images/zh-HK/braille.341962ff76b1bd70.webp) | ![盲文圖像預處理](../../../../../translated_images/zh-HK/braille-result.46530fea020b03c7.webp) | ![盲文符號](../../../../../translated_images/zh-HK/braille-symbols.0159185ab69d5339.webp) ----|-----|----- > 圖片來自 [OpenCV.ipynb](OpenCV.ipynb) * **使用幀差檢測視頻中的運動**。如果攝像機是固定的，那麼來自攝像機的幀應該彼此非常相似。由於幀表示為數組，只需對兩個連續幀的數組進行減法運算，我們就能得到像素差異，靜態幀的差異應該很小，而當圖像中有顯著運動時，差異會變大。 -![視頻幀和幀差的圖像](../../../../../translated_images/hk/frame-difference.706f805491a0883c.webp) +![視頻幀和幀差的圖像](../../../../../translated_images/zh-HK/frame-difference.706f805491a0883c.webp) > 圖片來自 [OpenCV.ipynb](OpenCV.ipynb) @@ -91,7 +91,7 @@ im = cv2.cvtColor(im,cv2.COLOR_BGR2RGB) - **密集光流** 計算顯示每個像素移動方向的向量場 - **稀疏光流** 基於圖像中的一些顯著特徵（例如邊緣），並從幀到幀構建它們的軌跡。 -![光流圖像](../../../../../translated_images/hk/optical.1f4a94464579a83a.webp) +![光流圖像](../../../../../translated_images/zh-HK/optical.1f4a94464579a83a.webp) > 圖片來自 [OpenCV.ipynb](OpenCV.ipynb) diff --git a/translations/hk/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md b/translations/hk/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md index 4e4be43e..927ed10a 100644 --- a/translations/hk/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md +++ b/translations/hk/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md @@ -13,11 +13,11 @@ CO_OP_TRANSLATOR_METADATA: VGG-16 是一個在 2014 年 ImageNet top-5 分類中達到 92.7% 準確率的網絡。它的層結構如下： -![ImageNet Layers](../../../../../translated_images/hk/vgg-16-arch1.d901a5583b3a51ba.webp) +![ImageNet Layers](../../../../../translated_images/zh-HK/vgg-16-arch1.d901a5583b3a51ba.webp) 如你所見，VGG 採用了傳統的金字塔架構，即一系列的卷積-池化層。 -![ImageNet Pyramid](../../../../../translated_images/hk/vgg-16-arch.64ff2137f50dd49f.webp) +![ImageNet Pyramid](../../../../../translated_images/zh-HK/vgg-16-arch.64ff2137f50dd49f.webp) > 圖片來源：[Researchgate](https://www.researchgate.net/figure/Vgg16-model-structure-To-get-the-VGG-NIN-model-we-replace-the-2-nd-4-th-6-th-7-th_fig2_335194493) diff --git a/translations/hk/lessons/4-ComputerVision/07-ConvNets/README.md b/translations/hk/lessons/4-ComputerVision/07-ConvNets/README.md index 19c319be..3b189eb3 100644 --- a/translations/hk/lessons/4-ComputerVision/07-ConvNets/README.md +++ b/translations/hk/lessons/4-ComputerVision/07-ConvNets/README.md @@ -17,7 +17,7 @@ CO_OP_TRANSLATOR_METADATA: 為了提取模式，我們將使用**卷積濾波器**的概念。正如你所知，圖像是由一個二維矩陣或帶有色彩深度的三維張量表示的。應用濾波器意味著我們取一個相對較小的**濾波核**矩陣，並對原始圖像中的每個像素與其鄰近點進行加權平均。我們可以將其視為一個小窗口在整個圖像上滑動，並根據濾波核矩陣中的權重對所有像素進行平均。 -![垂直邊緣濾波器](../../../../../translated_images/hk/filter-vert.b7148390ca0bc356.webp) | ![水平邊緣濾波器](../../../../../translated_images/hk/filter-horiz.59b80ed4feb946ef.webp) +![垂直邊緣濾波器](../../../../../translated_images/zh-HK/filter-vert.b7148390ca0bc356.webp) | ![水平邊緣濾波器](../../../../../translated_images/zh-HK/filter-horiz.59b80ed4feb946ef.webp) ----|---- > 圖片由 Dmitry Soshnikov 提供 @@ -38,7 +38,7 @@ CNN 的工作方式基於以下重要思想： * 我們可以設計網絡，使濾波器能夠自動訓練 * 我們可以使用相同的方法來在高層次特徵中尋找模式，而不僅僅是在原始圖像中。因此，CNN 的特徵提取在特徵層次上工作，從低層次的像素組合開始，到更高層次的圖像部分組合。 -![層次特徵提取](../../../../../translated_images/hk/FeatureExtractionCNN.d9b456cbdae7cb64.webp) +![層次特徵提取](../../../../../translated_images/zh-HK/FeatureExtractionCNN.d9b456cbdae7cb64.webp) > 圖片來自 [Hislop-Lynch 的論文](https://www.semanticscholar.org/paper/Computer-vision-based-pedestrian-trajectory-Hislop-Lynch/26e6f74853fc9bbb7487b06dc2cf095d36c9021d)，基於[他們的研究](https://dl.acm.org/doi/abs/10.1145/1553374.1553453) @@ -55,9 +55,9 @@ CNN 的工作方式基於以下重要思想： 例如，讓我們看看 VGG-16 的架構，這是一個在 2014 年 ImageNet 的前五名分類中達到 92.7% 準確率的網絡： -![ImageNet 層](../../../../../translated_images/hk/vgg-16-arch1.d901a5583b3a51ba.webp) +![ImageNet 層](../../../../../translated_images/zh-HK/vgg-16-arch1.d901a5583b3a51ba.webp) -![ImageNet 金字塔](../../../../../translated_images/hk/vgg-16-arch.64ff2137f50dd49f.webp) +![ImageNet 金字塔](../../../../../translated_images/zh-HK/vgg-16-arch.64ff2137f50dd49f.webp) > 圖片來自 [Researchgate](https://www.researchgate.net/figure/Vgg16-model-structure-To-get-the-VGG-NIN-model-we-replace-the-2-nd-4-th-6-th-7-th_fig2_335194493) diff --git a/translations/hk/lessons/4-ComputerVision/07-ConvNets/lab/README.md b/translations/hk/lessons/4-ComputerVision/07-ConvNets/lab/README.md index 7102677e..055d58ae 100644 --- a/translations/hk/lessons/4-ComputerVision/07-ConvNets/lab/README.md +++ b/translations/hk/lessons/4-ComputerVision/07-ConvNets/lab/README.md @@ -21,7 +21,7 @@ CO_OP_TRANSLATOR_METADATA: 我們將使用 [Oxford-IIIT Pet Dataset](https://www.robots.ox.ac.uk/~vgg/data/pets/)，該數據集包含37種不同品種的狗和貓的圖片。 -![我們將處理的數據集](../../../../../../translated_images/hk/data.50b2a9d5484bdbf0.webp) +![我們將處理的數據集](../../../../../../translated_images/zh-HK/data.50b2a9d5484bdbf0.webp) 要下載數據集，請使用以下代碼片段： diff --git a/translations/hk/lessons/4-ComputerVision/08-TransferLearning/README.md b/translations/hk/lessons/4-ComputerVision/08-TransferLearning/README.md index b5fb9a51..d172369d 100644 --- a/translations/hk/lessons/4-ComputerVision/08-TransferLearning/README.md +++ b/translations/hk/lessons/4-ComputerVision/08-TransferLearning/README.md @@ -29,7 +29,7 @@ Keras 和 PyTorch 都包含了方便的函數，可以輕鬆加載一些常見 以下是 VGG-16 網絡從一張貓的圖片中提取的特徵示例： -![VGG-16 提取的特徵](../../../../../translated_images/hk/features.6291f9c7ba3a0b95.webp) +![VGG-16 提取的特徵](../../../../../translated_images/zh-HK/features.6291f9c7ba3a0b95.webp) ## 貓與狗數據集 @@ -48,19 +48,19 @@ Keras 和 PyTorch 都包含了方便的函數，可以輕鬆加載一些常見 我們可以採取的一種方法是從一張隨機圖像開始，然後嘗試使用**梯度下降優化**技術調整該圖像，使得網絡認為它是一隻貓。 -![圖像優化循環](../../../../../translated_images/hk/ideal-cat-loop.999fbb8ff306e044.webp) +![圖像優化循環](../../../../../translated_images/zh-HK/ideal-cat-loop.999fbb8ff306e044.webp) 然而，如果我們這樣做，我們會得到一些非常接近隨機噪聲的東西。這是因為*有很多方法可以讓網絡認為輸入圖像是一隻貓*，其中一些方法在視覺上並不合理。雖然這些圖像包含了許多典型於貓的模式，但並沒有任何約束使它們在視覺上具有辨識度。 為了改善結果，我們可以在損失函數中添加另一個項，稱為**變化損失**。這是一種衡量圖像中相鄰像素相似程度的指標。最小化變化損失可以使圖像更平滑，並消除噪聲——從而揭示出更具視覺吸引力的模式。以下是一些這樣的「理想」圖像的例子，它們被高概率分類為貓和斑馬： -![理想貓](../../../../../translated_images/hk/ideal-cat.203dd4597643d6b0.webp) | ![理想斑馬](../../../../../translated_images/hk/ideal-zebra.7f70e8b54ee15a7a.webp) +![理想貓](../../../../../translated_images/zh-HK/ideal-cat.203dd4597643d6b0.webp) | ![理想斑馬](../../../../../translated_images/zh-HK/ideal-zebra.7f70e8b54ee15a7a.webp) -----|----- *理想貓* | *理想斑馬* 類似的方法可以用於對神經網絡進行所謂的**對抗性攻擊**。假設我們想要欺騙神經網絡，讓一隻狗看起來像一隻貓。如果我們拿一張狗的圖片，該圖片被網絡識別為狗，然後稍微調整它，使用梯度下降優化，直到網絡開始將其分類為貓： -![狗的圖片](../../../../../translated_images/hk/original-dog.8f68a67d2fe0911f.webp) | ![被分類為貓的狗圖片](../../../../../translated_images/hk/adversarial-dog.d9fc7773b0142b89.webp) +![狗的圖片](../../../../../translated_images/zh-HK/original-dog.8f68a67d2fe0911f.webp) | ![被分類為貓的狗圖片](../../../../../translated_images/zh-HK/adversarial-dog.d9fc7773b0142b89.webp) -----|----- *原始狗圖片* | *被分類為貓的狗圖片* diff --git a/translations/hk/lessons/4-ComputerVision/09-Autoencoders/README.md b/translations/hk/lessons/4-ComputerVision/09-Autoencoders/README.md index 5185151e..fd04ffb2 100644 --- a/translations/hk/lessons/4-ComputerVision/09-Autoencoders/README.md +++ b/translations/hk/lessons/4-ComputerVision/09-Autoencoders/README.md @@ -19,7 +19,7 @@ CO_OP_TRANSLATOR_METADATA: 由於我們訓練自動編碼器以捕捉原始圖像中的盡可能多的信息以進行準確重建，網絡會嘗試找到最佳的**嵌入**方式來捕捉輸入圖像的含義。 -![自動編碼器示意圖](../../../../../translated_images/hk/autoencoder_schema.5e6fc9ad98a5eb61.webp) +![自動編碼器示意圖](../../../../../translated_images/zh-HK/autoencoder_schema.5e6fc9ad98a5eb61.webp) > 圖片來源：[Keras 博客](https://blog.keras.io/building-autoencoders-in-keras.html) diff --git a/translations/hk/lessons/4-ComputerVision/11-ObjectDetection/README.md b/translations/hk/lessons/4-ComputerVision/11-ObjectDetection/README.md index cb8a979e..1338f160 100644 --- a/translations/hk/lessons/4-ComputerVision/11-ObjectDetection/README.md +++ b/translations/hk/lessons/4-ComputerVision/11-ObjectDetection/README.md @@ -13,7 +13,7 @@ CO_OP_TRANSLATOR_METADATA: ## [課前測驗](https://ff-quizzes.netlify.app/en/ai/quiz/21) -![物件偵測](../../../../../translated_images/hk/Screen_Shot_2016-11-17_at_11.14.54_AM.b4bb3769353287be.webp) +![物件偵測](../../../../../translated_images/zh-HK/Screen_Shot_2016-11-17_at_11.14.54_AM.b4bb3769353287be.webp) > 圖片來源：[YOLO v2 官方網站](https://pjreddie.com/darknet/yolov2/) @@ -25,7 +25,7 @@ CO_OP_TRANSLATOR_METADATA: 2. 對每個區塊進行影像分類。 3. 將分類結果中激活值足夠高的區塊視為包含目標物件的區域。 -![簡單的物件偵測](../../../../../translated_images/hk/naive-detection.e7f1ba220ccd08c6.webp) +![簡單的物件偵測](../../../../../translated_images/zh-HK/naive-detection.e7f1ba220ccd08c6.webp) > *圖片來源：[練習筆記本](ObjectDetection-TF.ipynb)* @@ -42,7 +42,7 @@ CO_OP_TRANSLATOR_METADATA: * [PASCAL VOC](http://host.robots.ox.ac.uk/pascal/VOC/) - 包含 20 個類別 * [COCO](http://cocodataset.org/#home) - 常見物件上下文數據集，包含 80 個類別、邊界框和分割遮罩 -![COCO](../../../../../translated_images/hk/coco-examples.71bc60380fa6cceb.webp) +![COCO](../../../../../translated_images/zh-HK/coco-examples.71bc60380fa6cceb.webp) ## 物件偵測的評估指標 @@ -50,7 +50,7 @@ CO_OP_TRANSLATOR_METADATA: 對於影像分類來說，衡量算法表現的方式很簡單，但對於物件偵測，我們需要同時衡量類別的正確性以及推測邊界框位置的精確性。後者使用所謂的**交集比聯集** (IoU) 來衡量，這是一種用來測量兩個框（或任意兩個區域）重疊程度的方法。 -![IoU](../../../../../translated_images/hk/iou_equation.9a4751d40fff4e11.webp) +![IoU](../../../../../translated_images/zh-HK/iou_equation.9a4751d40fff4e11.webp) > *圖片來源：[這篇優秀的 IoU 部落格文章](https://pyimagesearch.com/2016/11/07/intersection-over-union-iou-for-object-detection/)* @@ -97,11 +97,11 @@ $$ [R-CNN](http://islab.ulsan.ac.kr/files/announcement/513/rcnn_pami.pdf) 使用 [Selective Search](http://www.huppelen.nl/publications/selectiveSearchDraft.pdf) 生成 ROI 區域的層次結構，然後通過 CNN 特徵提取器和 SVM 分類器來確定物件類別，並通過線性回歸確定*邊界框*座標。[官方論文](https://arxiv.org/pdf/1506.01497v1.pdf) -![RCNN](../../../../../translated_images/hk/rcnn1.cae407020dfb1d1f.webp) +![RCNN](../../../../../translated_images/zh-HK/rcnn1.cae407020dfb1d1f.webp) > *圖片來源：van de Sande et al. ICCV’11* -![RCNN-1](../../../../../translated_images/hk/rcnn2.2d9530bb83516484.webp) +![RCNN-1](../../../../../translated_images/zh-HK/rcnn2.2d9530bb83516484.webp) > *圖片來源：[這篇部落格](https://towardsdatascience.com/r-cnn-fast-r-cnn-faster-r-cnn-yolo-object-detection-algorithms-36d53571365e)* @@ -109,7 +109,7 @@ $$ 這種方法與 R-CNN 類似，但區域是在卷積層應用後定義的。 -![FRCNN](../../../../../translated_images/hk/f-rcnn.3cda6d9bb4188875.webp) +![FRCNN](../../../../../translated_images/zh-HK/f-rcnn.3cda6d9bb4188875.webp) > 圖片來源：[官方論文](https://www.cv-foundation.org/openaccess/content_iccv_2015/papers/Girshick_Fast_R-CNN_ICCV_2015_paper.pdf)，[arXiv](https://arxiv.org/pdf/1504.08083.pdf)，2015 @@ -117,7 +117,7 @@ $$ 這種方法的主要思想是使用神經網絡來預測 ROI，即所謂的*區域提議網絡* (Region Proposal Network)。[論文](https://arxiv.org/pdf/1506.01497.pdf)，2016 -![FasterRCNN](../../../../../translated_images/hk/faster-rcnn.8d46c099b87ef30a.webp) +![FasterRCNN](../../../../../translated_images/zh-HK/faster-rcnn.8d46c099b87ef30a.webp) > 圖片來源：[官方論文](https://arxiv.org/pdf/1506.01497.pdf) @@ -129,7 +129,7 @@ $$ 2. 特徵經過**位置敏感得分圖** (Position-Sensitive Score Map) 處理。每個來自 $C$ 類別的物件被劃分為 $k\times k$ 區域，我們訓練網絡來預測物件的部分。 3. 對於 $k\times k$ 區域中的每個部分，所有網絡對物件類別進行投票，選擇得票最多的物件類別。 -![r-fcn 圖片](../../../../../translated_images/hk/r-fcn.13eb88158b99a3da.webp) +![r-fcn 圖片](../../../../../translated_images/zh-HK/r-fcn.13eb88158b99a3da.webp) > 圖片來源：[官方論文](https://arxiv.org/abs/1605.06409) @@ -140,7 +140,7 @@ YOLO 是一種實時單次通過算法。主要思想如下： * 將圖片劃分為 $S\times S$ 區域。 * 對於每個區域，**CNN** 預測 $n$ 個可能的物件、*邊界框*座標以及*置信度*=*概率* * IoU。 - ![YOLO](../../../../../translated_images/hk/yolo.a2648ec82ee8bb4e.webp) + ![YOLO](../../../../../translated_images/zh-HK/yolo.a2648ec82ee8bb4e.webp) > 圖片來源：[官方論文](https://arxiv.org/abs/1506.02640) diff --git a/translations/hk/lessons/4-ComputerVision/README.md b/translations/hk/lessons/4-ComputerVision/README.md index 58212b66..063dfbab 100644 --- a/translations/hk/lessons/4-ComputerVision/README.md +++ b/translations/hk/lessons/4-ComputerVision/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 電腦視覺 -![電腦視覺內容摘要的手繪圖](../../../../translated_images/hk/ai-computervision.6506ebebac3fbf76.webp) +![電腦視覺內容摘要的手繪圖](../../../../translated_images/zh-HK/ai-computervision.6506ebebac3fbf76.webp) 在這部分，我們將學習以下內容： diff --git a/translations/hk/lessons/5-NLP/14-Embeddings/README.md b/translations/hk/lessons/5-NLP/14-Embeddings/README.md index 9aa34d06..e6367b5c 100644 --- a/translations/hk/lessons/5-NLP/14-Embeddings/README.md +++ b/translations/hk/lessons/5-NLP/14-Embeddings/README.md @@ -19,7 +19,7 @@ CO_OP_TRANSLATOR_METADATA: 通過在分類器網絡中使用嵌入層作為第一層，我們可以從詞袋模型切換到 **嵌入袋** 模型。在嵌入袋模型中，我們首先將文本中的每個詞轉換為相應的嵌入，然後對所有嵌入計算某種聚合函數，例如 `sum`、`average` 或 `max`。 -![展示五個序列詞嵌入分類器的圖片。](../../../../../translated_images/hk/embedding-classifier-example.b77f021a7ee67eee.webp) +![展示五個序列詞嵌入分類器的圖片。](../../../../../translated_images/zh-HK/embedding-classifier-example.b77f021a7ee67eee.webp) > 圖片由作者提供 @@ -40,7 +40,7 @@ CO_OP_TRANSLATOR_METADATA: CBoW 的速度更快，而 Skip-Gram 雖然較慢，但在表示不常見詞方面效果更好。 -![展示 CBoW 和 Skip-Gram 將詞轉換為向量的算法圖片。](../../../../../translated_images/hk/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) +![展示 CBoW 和 Skip-Gram 將詞轉換為向量的算法圖片。](../../../../../translated_images/zh-HK/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) > 圖片來源：[這篇論文](https://arxiv.org/pdf/1301.3781.pdf) diff --git a/translations/hk/lessons/5-NLP/15-LanguageModeling/README.md b/translations/hk/lessons/5-NLP/15-LanguageModeling/README.md index 3dc4dfe4..8b62e594 100644 --- a/translations/hk/lessons/5-NLP/15-LanguageModeling/README.md +++ b/translations/hk/lessons/5-NLP/15-LanguageModeling/README.md @@ -23,7 +23,7 @@ CO_OP_TRANSLATOR_METADATA: * **連續詞袋模型** (CBoW)，通過預測詞元序列 $W_{-N}$, ..., $W_N$ 中的中間詞元 $W_0$。 * **Skip-gram**，通過中間詞元 $W_0$ 預測一組相鄰詞元 {$W_{-N},\dots, W_{-1}, W_1,\dots, W_N$}。 -![來自論文的將詞語轉換為向量的算法示例](../../../../../translated_images/hk/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) +![來自論文的將詞語轉換為向量的算法示例](../../../../../translated_images/zh-HK/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) > 圖片來源：[這篇論文](https://arxiv.org/pdf/1301.3781.pdf) diff --git a/translations/hk/lessons/5-NLP/16-RNN/README.md b/translations/hk/lessons/5-NLP/16-RNN/README.md index 25e3deb6..0748835d 100644 --- a/translations/hk/lessons/5-NLP/16-RNN/README.md +++ b/translations/hk/lessons/5-NLP/16-RNN/README.md @@ -15,7 +15,7 @@ CO_OP_TRANSLATOR_METADATA: 為了捕捉文本序列的意義，我們需要使用另一種神經網絡架構，稱為**循環神經網絡**（Recurrent Neural Network，簡稱 RNN）。在 RNN 中，我們將句子逐個符號地輸入網絡，網絡會生成某種**狀態**，然後將該狀態與下一個符號一起再次輸入網絡。 -![RNN](../../../../../translated_images/hk/rnn.27f5c29c53d727b5.webp) +![RNN](../../../../../translated_images/zh-HK/rnn.27f5c29c53d727b5.webp) > 圖片由作者提供 @@ -61,7 +61,7 @@ LSTM 網絡的組織方式與 RNN 類似，但有兩個狀態會從層到層傳 循環網絡，無論是單向還是雙向，都能捕捉序列中的某些模式，並將其存儲到狀態向量中或傳遞到輸出中。與卷積網絡類似，我們可以在第一層之上構建另一個循環層，以捕捉更高層次的模式，並基於第一層提取的低層次模式進行構建。這引出了**多層 RNN**的概念，它由兩個或更多循環網絡組成，其中前一層的輸出作為下一層的輸入。 -![顯示多層長短期記憶 RNN 的圖片](../../../../../translated_images/hk/multi-layer-lstm.dd975e29bb2a59fe.webp) +![顯示多層長短期記憶 RNN 的圖片](../../../../../translated_images/zh-HK/multi-layer-lstm.dd975e29bb2a59fe.webp) *圖片來自 [這篇精彩文章](https://towardsdatascience.com/from-a-lstm-cell-to-a-multilayer-lstm-network-with-pytorch-2899eb5696f3) 作者 Fernando López* diff --git a/translations/hk/lessons/5-NLP/17-GenerativeNetworks/README.md b/translations/hk/lessons/5-NLP/17-GenerativeNetworks/README.md index 87b26be2..66e271e1 100644 --- a/translations/hk/lessons/5-NLP/17-GenerativeNetworks/README.md +++ b/translations/hk/lessons/5-NLP/17-GenerativeNetworks/README.md @@ -19,7 +19,7 @@ CO_OP_TRANSLATOR_METADATA: 這使得不同的神經架構成為可能，如下圖所示： -![展示常見循環神經網絡模式的圖片。](../../../../../translated_images/hk/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) +![展示常見循環神經網絡模式的圖片。](../../../../../translated_images/zh-HK/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) > 圖片來自 [Andrej Karpaty](http://karpathy.github.io/) 的博客文章 [Unreasonable Effectiveness of Recurrent Neural Networks](http://karpathy.github.io/2015/05/21/rnn-effectiveness/) @@ -32,7 +32,7 @@ CO_OP_TRANSLATOR_METADATA: 我們將訓練這個 RNN 逐步生成文本。在每一步中，我們將取一個長度為 `nchars` 的字符序列，並要求網絡為每個輸入字符生成下一個輸出字符： -![展示 RNN 生成單詞 'HELLO' 的示例圖片。](../../../../../translated_images/hk/rnn-generate.56c54afb52f9781d.webp) +![展示 RNN 生成單詞 'HELLO' 的示例圖片。](../../../../../translated_images/zh-HK/rnn-generate.56c54afb52f9781d.webp) 在生成文本（推理過程中）時，我們從某個**提示**開始，將其通過 RNN 單元生成中間狀態，然後從該狀態開始生成。我們一次生成一個字符，並將狀態和生成的字符傳遞給另一個 RNN 單元以生成下一個字符，直到生成足夠的字符。 diff --git a/translations/hk/lessons/5-NLP/18-Transformers/README.md b/translations/hk/lessons/5-NLP/18-Transformers/README.md index 7c683bff..55821740 100644 --- a/translations/hk/lessons/5-NLP/18-Transformers/README.md +++ b/translations/hk/lessons/5-NLP/18-Transformers/README.md @@ -20,13 +20,13 @@ CO_OP_TRANSLATOR_METADATA: **注意力機制**提供了一種方法，能夠對每個輸入向量對RNN輸出預測的上下文影響進行加權。其實現方式是通過在輸入RNN的中間狀態與輸出RNN之間創建捷徑。這樣，在生成輸出符號yt時，我們會考慮所有輸入隱藏狀態hi，並賦予不同的權重係數αt,i。 -![顯示帶有加性注意力層的編碼器/解碼器模型的圖片](../../../../../translated_images/hk/encoder-decoder-attention.7a726296894fb567.webp) +![顯示帶有加性注意力層的編碼器/解碼器模型的圖片](../../../../../translated_images/zh-HK/encoder-decoder-attention.7a726296894fb567.webp) > [Bahdanau等人, 2015](https://arxiv.org/pdf/1409.0473.pdf)中的加性注意力機制編碼器-解碼器模型，圖片來源於[這篇博客文章](https://lilianweng.github.io/lil-log/2018/06/24/attention-attention.html) 注意力矩陣{αi,j}表示某些輸入詞在生成給定輸出序列中的某個詞時所起的作用程度。以下是一個這樣的矩陣示例： -![顯示RNNsearch-50找到的示例對齊的圖片，取自Bahdanau - arviz.org](../../../../../translated_images/hk/bahdanau-fig3.09ba2d37f202a6af.webp) +![顯示RNNsearch-50找到的示例對齊的圖片，取自Bahdanau - arviz.org](../../../../../translated_images/zh-HK/bahdanau-fig3.09ba2d37f202a6af.webp) > 圖片來自[Bahdanau等人, 2015](https://arxiv.org/pdf/1409.0473.pdf)（圖3） @@ -66,7 +66,7 @@ Transformer的主要思想之一是避免RNN的序列性質，並創建一個在 接下來，我們需要捕捉序列中的一些模式。為此，Transformer使用了**自注意力**機制，這本質上是將注意力應用於相同的輸入和輸出序列。應用自注意力使我們能夠考慮句子中的**上下文**，並查看哪些詞是相互關聯的。例如，它使我們能夠看到哪些詞由指代詞（如*it*）指代，並考慮上下文： -![](../../../../../translated_images/hk/CoreferenceResolution.861924d6d384a7d6.webp) +![](../../../../../translated_images/zh-HK/CoreferenceResolution.861924d6d384a7d6.webp) > 圖片來自[Google博客](https://research.googleblog.com/2017/08/transformer-novel-neural-network.html) @@ -91,7 +91,7 @@ Transformer的主要思想之一是避免RNN的序列性質，並創建一個在 **BERT**（Bidirectional Encoder Representations from Transformers）是一個非常大的多層Transformer網絡，*BERT-base*有12層，*BERT-large*有24層。該模型首先在大規模文本數據語料庫（維基百科+書籍）上進行無監督預訓練（預測句子中的被遮蔽詞）。在預訓練過程中，模型吸收了大量的語言理解能力，這些能力可以通過微調其他數據集來利用。這個過程被稱為**遷移學習**。 -![圖片來自http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/hk/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp) +![圖片來自http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/zh-HK/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp) > 圖片[來源](http://jalammar.github.io/illustrated-bert/) diff --git a/translations/hk/lessons/5-NLP/19-NER/README.md b/translations/hk/lessons/5-NLP/19-NER/README.md index a93b9689..30737db9 100644 --- a/translations/hk/lessons/5-NLP/19-NER/README.md +++ b/translations/hk/lessons/5-NLP/19-NER/README.md @@ -57,7 +57,7 @@ NER 模型本質上是 **標記分類模型**，因為對於每個輸入標記 由於我們需要在標記和類別之間建立一對一的對應關係，我們可以從這張圖中訓練一個最右側的 **多對多** 神經網絡模型： -![展示常見循環神經網絡模式的圖片。](../../../../../translated_images/hk/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) +![展示常見循環神經網絡模式的圖片。](../../../../../translated_images/zh-HK/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) > *圖片來自 [這篇博客文章](http://karpathy.github.io/2015/05/21/rnn-effectiveness/) 作者 [Andrej Karpathy](http://karpathy.github.io/)。NER 標記分類模型對應於圖片中最右側的網絡架構。* diff --git a/translations/hk/lessons/5-NLP/README.md b/translations/hk/lessons/5-NLP/README.md index 472e4581..2effcd9d 100644 --- a/translations/hk/lessons/5-NLP/README.md +++ b/translations/hk/lessons/5-NLP/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 自然語言處理 -![NLP 任務的手繪圖概述](../../../../translated_images/hk/ai-nlp.b22dcb8ca4707cea.webp) +![NLP 任務的手繪圖概述](../../../../translated_images/zh-HK/ai-nlp.b22dcb8ca4707cea.webp) 在本節中，我們將專注於使用神經網絡來處理與**自然語言處理 (NLP)** 相關的任務。我們希望計算機能夠解決許多 NLP 問題： diff --git a/translations/hk/lessons/6-Other/23-MultiagentSystems/README.md b/translations/hk/lessons/6-Other/23-MultiagentSystems/README.md index f72bbfae..702b966b 100644 --- a/translations/hk/lessons/6-Other/23-MultiagentSystems/README.md +++ b/translations/hk/lessons/6-Other/23-MultiagentSystems/README.md @@ -70,7 +70,7 @@ NetLogo的一大優勢是它包含一個可供試用的工作模型庫。進入* 打開模型後，你會進入NetLogo的主屏幕。以下是一個描述狼和羊的種群模型，給定有限資源（草地）。 -![NetLogo主屏幕](../../../../../translated_images/hk/NetLogo-Main.32653711ec1a01b3.webp) +![NetLogo主屏幕](../../../../../translated_images/zh-HK/NetLogo-Main.32653711ec1a01b3.webp) > Dmitry Soshnikov提供的截圖 diff --git a/translations/hk/lessons/README.md b/translations/hk/lessons/README.md index 5f65959b..e491f890 100644 --- a/translations/hk/lessons/README.md +++ b/translations/hk/lessons/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 概覽 -![概覽的手繪圖](../../../translated_images/hk/ai-overview.0857791951d19500.webp) +![概覽的手繪圖](../../../translated_images/zh-HK/ai-overview.0857791951d19500.webp) > 手繪筆記由 [Tomomi Imura](https://twitter.com/girlie_mac) 提供 diff --git a/translations/hk/lessons/X-Extras/X1-MultiModal/README.md b/translations/hk/lessons/X-Extras/X1-MultiModal/README.md index 5ed8ed79..508f9d2e 100644 --- a/translations/hk/lessons/X-Extras/X1-MultiModal/README.md +++ b/translations/hk/lessons/X-Extras/X1-MultiModal/README.md @@ -15,7 +15,7 @@ CO_OP_TRANSLATOR_METADATA: CLIP 的主要理念是能夠比較文本提示與圖像，並判斷圖像與提示的匹配程度。 -![CLIP 架構](../../../../../translated_images/hk/clip-arch.b3dbf20b4e8ed8be.webp) +![CLIP 架構](../../../../../translated_images/zh-HK/clip-arch.b3dbf20b4e8ed8be.webp) > *圖片來源：[這篇博客文章](https://openai.com/blog/clip/)* @@ -29,7 +29,7 @@ CLIP 模型/庫可從 [OpenAI GitHub](https://github.com/openai/CLIP) 獲得。 假設我們需要將圖像分類為例如貓、狗和人。在這種情況下，我們可以向模型提供一張圖像，以及一系列文本提示：“*一張貓的照片*”、“*一張狗的照片*”、“*一張人的照片*”。在結果的三個概率向量中，我們只需選擇值最高的索引。 -![CLIP 用於圖像分類](../../../../../translated_images/hk/clip-class.3af42ef0b2b19369.webp) +![CLIP 用於圖像分類](../../../../../translated_images/zh-HK/clip-class.3af42ef0b2b19369.webp) > *圖片來源：[這篇博客文章](https://openai.com/blog/clip/)* @@ -53,13 +53,13 @@ VQGAN 與普通 [GAN](../../4-ComputerVision/10-GANs/README.md) 的主要區別 VQGAN 與傳統 GAN 的一個重要區別是，後者可以從任何輸入向量生成一張像樣的圖像，而 VQGAN 則可能生成不連貫的圖像。因此，我們需要進一步引導圖像創建過程，這可以通過 CLIP 來完成。 -![VQGAN+CLIP 架構](../../../../../translated_images/hk/vqgan.5027fe05051dfa31.webp) +![VQGAN+CLIP 架構](../../../../../translated_images/zh-HK/vqgan.5027fe05051dfa31.webp) 為了生成與文本提示相對應的圖像，我們首先使用一些隨機編碼向量，通過 VQGAN 生成一張圖像。然後使用 CLIP 生成損失函數，顯示圖像與文本提示的匹配程度。目標是最小化該損失，通過反向傳播調整輸入向量的參數。 一個實現 VQGAN+CLIP 的優秀庫是 [Pixray](http://github.com/pixray/pixray)。 -![由 Pixray 生成的圖片](../../../../../translated_images/hk/a_closeup_watercolor_portrait_of_young_male_teacher_of_literature_with_a_book.2384968e9db8a0d0.webp) | ![由 Pixray 生成的圖片](../../../../../translated_images/hk/a_closeup_oil_portrait_of_young_female_teacher_of_computer_science_with_a_computer.e0b6495f210a4390.webp) | ![由 Pixray 生成的圖片](../../../../../translated_images/hk/a_closeup_oil_portrait_of_old_male_teacher_of_math.5362e67aa7fc2683.webp) +![由 Pixray 生成的圖片](../../../../../translated_images/zh-HK/a_closeup_watercolor_portrait_of_young_male_teacher_of_literature_with_a_book.2384968e9db8a0d0.webp) | ![由 Pixray 生成的圖片](../../../../../translated_images/zh-HK/a_closeup_oil_portrait_of_young_female_teacher_of_computer_science_with_a_computer.e0b6495f210a4390.webp) | ![由 Pixray 生成的圖片](../../../../../translated_images/zh-HK/a_closeup_oil_portrait_of_old_male_teacher_of_math.5362e67aa7fc2683.webp) ----|----|---- 根據提示 *一張年輕男性文學教師拿著書的水彩特寫肖像* 生成的圖片 | 根據提示 *一張年輕女性計算機科學教師拿著電腦的油畫特寫肖像* 生成的圖片 | 根據提示 *一張老年男性數學教師站在黑板前的油畫特寫肖像* 生成的圖片 @@ -75,7 +75,7 @@ DALL-E 是 GPT-3 的一個版本，訓練用於根據提示生成圖像。它擁 DALL-E 1 和 DALL-E 2 的主要區別在於，後者能生成更逼真的圖像和藝術作品。 以下是使用 DALL-E 生成的圖像示例： -![由 DALL-E 生成的圖片](../../../../../translated_images/hk/DALL·E%202023-06-20%2015.56.56%20-%20a%20closeup%20watercolor%20portrait%20of%20young%20male%20teacher%20of%20literature%20with%20a%20book.6c235e8271d9ed10ce985d86aeb241a58518958647973af136912116b9518fce.png) | ![由 DALL-E 生成的圖片](../../../../../translated_images/hk/DALL·E%202023-06-20%2015.57.43%20-%20a%20closeup%20oil%20portrait%20of%20young%20female%20teacher%20of%20computer%20science%20with%20a%20computer.f21dc4166340b6c8b4d1cb57efd1e22127407f9b28c9ac7afe11344065369e64.png) | ![由 DALL-E 生成的圖片](../../../../../translated_images/hk/DALL·E%202023-06-20%2015.58.42%20-%20%20a%20closeup%20oil%20portrait%20of%20old%20male%20teacher%20of%20mathematics%20in%20front%20of%20blackboard.d331c2dfbdc3f7c46aa65c0809066f5e7ed4b49609cd259852e760df21051e4a.png) +![由 DALL-E 生成的圖片](../../../../../translated_images/zh-HK/DALL·E%202023-06-20%2015.56.56%20-%20a%20closeup%20watercolor%20portrait%20of%20young%20male%20teacher%20of%20literature%20with%20a%20book.6c235e8271d9ed10ce985d86aeb241a58518958647973af136912116b9518fce.png) | ![由 DALL-E 生成的圖片](../../../../../translated_images/zh-HK/DALL·E%202023-06-20%2015.57.43%20-%20a%20closeup%20oil%20portrait%20of%20young%20female%20teacher%20of%20computer%20science%20with%20a%20computer.f21dc4166340b6c8b4d1cb57efd1e22127407f9b28c9ac7afe11344065369e64.png) | ![由 DALL-E 生成的圖片](../../../../../translated_images/zh-HK/DALL·E%202023-06-20%2015.58.42%20-%20%20a%20closeup%20oil%20portrait%20of%20old%20male%20teacher%20of%20mathematics%20in%20front%20of%20blackboard.d331c2dfbdc3f7c46aa65c0809066f5e7ed4b49609cd259852e760df21051e4a.png) ----|----|---- 根據提示 *一張年輕男性文學教師拿著書的水彩特寫肖像* 生成的圖片 | 根據提示 *一張年輕女性計算機科學教師拿著電腦的油畫特寫肖像* 生成的圖片 | 根據提示 *一張老年男性數學教師站在黑板前的油畫特寫肖像* 生成的圖片 diff --git a/translations/mo/README.md b/translations/mo/README.md index 5fadddb4..3572c659 100644 --- a/translations/mo/README.md +++ b/translations/mo/README.md @@ -23,7 +23,7 @@ CO_OP_TRANSLATOR_METADATA: # 初學者人工智能課程大綱 -|![Sketchnote by @girlie_mac https://twitter.com/girlie_mac](../../../../translated_images/mo/ai-overview.0857791951d19500.webp)| +|![Sketchnote by @girlie_mac https://twitter.com/girlie_mac](../../../../translated_images/zh-MO/ai-overview.0857791951d19500.webp)| |:---:| | 初學者人工智能 - _手繪筆記來自 [@girlie_mac](https://twitter.com/girlie_mac)_ | diff --git a/translations/mo/lessons/1-Intro/README.md b/translations/mo/lessons/1-Intro/README.md index e2d8e984..8735a801 100644 --- a/translations/mo/lessons/1-Intro/README.md +++ b/translations/mo/lessons/1-Intro/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 人工智能簡介 -![人工智能簡介內容的手繪圖](../../../../translated_images/mo/ai-intro.bf28d1ac4235881c.webp) +![人工智能簡介內容的手繪圖](../../../../translated_images/zh-MO/ai-intro.bf28d1ac4235881c.webp) > 手繪筆記由 [Tomomi Imura](https://twitter.com/girlie_mac) 提供 @@ -19,7 +19,7 @@ CO_OP_TRANSLATOR_METADATA: 最初，電腦是由 [查爾斯·巴貝奇](https://en.wikipedia.org/wiki/Charles_Babbage) 發明的，用於按照明確定義的程序（算法）處理數字。現代電腦雖然比19世紀提出的原始模型先進得多，但仍然遵循受控計算的理念。因此，如果我們知道實現目標所需的精確步驟序列，就可以編程讓電腦完成某些事情。 -![一個人的照片](../../../../translated_images/mo/dsh_age.d212a30d4e54fb5f.webp) +![一個人的照片](../../../../translated_images/zh-MO/dsh_age.d212a30d4e54fb5f.webp) > 照片由 [Vickie Soshnikova](http://twitter.com/vickievalerie) 提供 @@ -46,7 +46,7 @@ CO_OP_TRANSLATOR_METADATA: 處理 **[智能](https://en.wikipedia.org/wiki/Intelligence)** 這個術語時的一個問題是，對於這個術語並沒有明確的定義。有人認為智能與 **抽象思維** 或 **自我意識** 有關，但我們無法準確定義它。 -![一隻貓的照片](../../../../translated_images/mo/photo-cat.8c8e8fb760ffe457.webp) +![一隻貓的照片](../../../../translated_images/zh-MO/photo-cat.8c8e8fb760ffe457.webp) > [照片](https://unsplash.com/photos/75715CVEJhI) 由 [Amber Kipp](https://unsplash.com/@sadmax) 提供，來自 Unsplash @@ -98,13 +98,13 @@ CO_OP_TRANSLATOR_METADATA: > | 那麼機器學習呢？ | | > |--------------|-----------| -> | 基於計算機學習如何根據某些數據解決問題的人工智能部分稱為 **機器學習**。我們不會在本課程中考慮經典的機器學習——我們建議參考單獨的 [機器學習初學者課程](http://aka.ms/ml-beginners)。 | ![機器學習初學者課程](../../../../translated_images/mo/ml-for-beginners.9e4fed176fd5817d.webp) | +> | 基於計算機學習如何根據某些數據解決問題的人工智能部分稱為 **機器學習**。我們不會在本課程中考慮經典的機器學習——我們建議參考單獨的 [機器學習初學者課程](http://aka.ms/ml-beginners)。 | ![機器學習初學者課程](../../../../translated_images/zh-MO/ml-for-beginners.9e4fed176fd5817d.webp) | ## 人工智能的簡史 人工智能作為一個領域始於20世紀中葉。最初，符號推理是一種流行的方法，並且它帶來了一些重要的成功，例如專家系統——能夠在某些有限問題領域中充當專家的計算機程序。然而，很快就發現這種方法並不適用於大規模應用。從專家那裡提取知識、將其表示在計算機中並保持知識庫的準確性，事實證明這是一項非常複雜且在許多情況下成本過高的任務。這導致了20世紀70年代所謂的 [人工智能寒冬](https://en.wikipedia.org/wiki/AI_winter)。 -人工智能簡史 +人工智能簡史 > 圖片由 [Dmitry Soshnikov](http://soshnikov.com) 提供 @@ -124,7 +124,7 @@ CO_OP_TRANSLATOR_METADATA: * 現代助手（如 Cortana、Siri 或 Google Assistant）都是混合系統，使用神經網絡將語音轉換為文本並識別我們的意圖，然後使用一些推理或明確的算法執行所需的操作。 * 未來，我們可能會期待一個完全基於神經網絡的模型能夠自行處理對話。最近的 GPT 和 [Turing-NLG](https://www.microsoft.com/research/blog/turing-nlg-a-17-billion-parameter-language-model-by-microsoft) 神經網絡家族在這方面表現出色。 -圖靈測試的演變 +圖靈測試的演變 > 圖片由 Dmitry Soshnikov 提供，[照片](https://unsplash.com/photos/r8LmVbUKgns) 由 [Marina Abrosimova](https://unsplash.com/@abrosimova_marina_foto) 提供，Unsplash ## 最近的人工智能研究 diff --git a/translations/mo/lessons/2-Symbolic/Animals.ipynb b/translations/mo/lessons/2-Symbolic/Animals.ipynb index 77607738..7277916e 100644 --- a/translations/mo/lessons/2-Symbolic/Animals.ipynb +++ b/translations/mo/lessons/2-Symbolic/Animals.ipynb @@ -12,7 +12,7 @@ "\n", "在本範例中，我們將實作一個簡單的知識庫系統，用來根據一些物理特徵判定動物。系統可以用以下的 AND-OR 樹來表示（這是整棵樹的一部分，我們可以輕易加入更多規則）：\n", "\n", - "![](../../../../../../translated_images/mo/AND-OR-Tree.5592d2c70187f283.webp)\n" + "![](../../../../../../translated_images/zh-MO/AND-OR-Tree.5592d2c70187f283.webp)\n" ] }, { diff --git a/translations/mo/lessons/2-Symbolic/README.md b/translations/mo/lessons/2-Symbolic/README.md index ceadbba3..544dced4 100644 --- a/translations/mo/lessons/2-Symbolic/README.md +++ b/translations/mo/lessons/2-Symbolic/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 知識表示與專家系統 -![符號AI內容摘要](../../../../../../translated_images/mo/ai-symbolic.715a30cb610411a6.webp) +![符號AI內容摘要](../../../../../../translated_images/zh-MO/ai-symbolic.715a30cb610411a6.webp) > 手繪筆記由 [Tomomi Imura](https://twitter.com/girlie_mac) 製作 @@ -35,13 +35,13 @@ CO_OP_TRANSLATOR_METADATA: * **知識**是被整合到我們的世界模型中的資訊。例如，當我們知道電腦是什麼，我們開始對它如何運作、價格多少、用途等有想法。這些相互關聯的概念網絡形成了我們的知識。 * **智慧**是我們對世界理解的更高一層，代表*元知識*，例如關於何時及如何使用知識的一些觀念。 - + *圖片來自 [維基百科](https://commons.wikimedia.org/w/index.php?curid=37705247)，作者 Longlivetheux - 自作，CC BY-SA 4.0* 因此，**知識表示**問題是找到有效的方式，將知識以資料形式表示在電腦內，讓其可自動使用。這可看作是一個光譜： -![知識表示光譜](../../../../../../translated_images/mo/knowledge-spectrum.b60df631852c0217.webp) +![知識表示光譜](../../../../../../translated_images/zh-MO/knowledge-spectrum.b60df631852c0217.webp) > 圖片由 [Dmitry Soshnikov](http://soshnikov.com) 提供 @@ -94,7 +94,7 @@ Python | 區塊語法 | 縮排 符號AI的早期成功之一是所謂的**專家系統**——設計用於充當有限問題領域專家的計算機系統。它基於從一個或多個人類專家提取的**知識庫**，並包含執行推理的**推理引擎**。 -![人類架構](../../../../../../translated_images/mo/arch-human.5d4d35f1bba3ab1c.webp) | ![基於知識系統](../../../../../../translated_images/mo/arch-kbs.3ec5c150b09fa8da.webp) +![人類架構](../../../../../../translated_images/zh-MO/arch-human.5d4d35f1bba3ab1c.webp) | ![基於知識系統](../../../../../../translated_images/zh-MO/arch-kbs.3ec5c150b09fa8da.webp) ---------------------------------------------|------------------------------------------------ 簡化的人類神經系統結構 | 基於知識系統的架構 @@ -106,7 +106,7 @@ Python | 區塊語法 | 縮排 舉例來說，考慮以下透過動物的物理特徵判定動物的專家系統： -![AND-OR樹](../../../../../../translated_images/mo/AND-OR-Tree.5592d2c70187f283.webp) +![AND-OR樹](../../../../../../translated_images/zh-MO/AND-OR-Tree.5592d2c70187f283.webp) > 圖片由 [Dmitry Soshnikov](http://soshnikov.com) 提供 @@ -175,7 +175,7 @@ THEN the animal is a carnivore 在語意網中，所有表示都是基於三元組。每個物件和每個關係都由 URI 唯一標識。例如，如果我們想表示這個 AI 課程是由 Dmitry Soshnikov 於 2022 年 1 月 1 日開發的——這裡是我們可以使用的三元組： - + ``` http://github.com/microsoft/ai-for-beginners http://www.example.com/terms/creation-date “Jan 1, 2022” @@ -186,7 +186,7 @@ http://github.com/microsoft/ai-for-beginners http://purl.org/dc/elements/1.1/cre 在更複雜的情況下，如果我們想定義一個創建者列表，可以使用 RDF 定義的一些資料結構。 - + > 上方圖示由 [Dmitry Soshnikov](http://soshnikov.com) 製作 @@ -210,7 +210,7 @@ GROUP BY ?eyeColorLabel > ✅ 如果你想嘗試自己建立本體，或打開現有的本體，有一個很棒的視覺本體編輯器叫做 [Protégé](https://protege.stanford.edu/)。你可以下載使用，也可以線上使用。 - + *Web Protégé 編輯器開啟了羅曼諾夫家族本體。截圖由 Dmitry Soshnikov 提供* diff --git a/translations/mo/lessons/3-NeuralNetworks/04-OwnFramework/OwnFramework.ipynb b/translations/mo/lessons/3-NeuralNetworks/04-OwnFramework/OwnFramework.ipynb index 6b929e28..1c1b160c 100644 --- a/translations/mo/lessons/3-NeuralNetworks/04-OwnFramework/OwnFramework.ipynb +++ b/translations/mo/lessons/3-NeuralNetworks/04-OwnFramework/OwnFramework.ipynb @@ -487,7 +487,7 @@ "\n", "如果我們有超過兩個類別，softmax會在所有類別之間對概率進行正規化。以下是用於 MNIST 數字分類的網絡架構圖：\n", "\n", - "![MNIST 分類器](../../../../../translated_images/mo/Cross-Entropy-Loss.7acff482d48cc41b2d3ef85ff0b263fbab299da128a462fa9b0640c5b7bc12ef.png)\n" + "![MNIST 分類器](../../../../../translated_images/zh-MO/Cross-Entropy-Loss.7acff482d48cc41b2d3ef85ff0b263fbab299da128a462fa9b0640c5b7bc12ef.png)\n" ] }, { @@ -1259,7 +1259,7 @@ "* 訓練損失低——模型能很好地接近訓練數據，因為它具有足夠的表達能力。\n", "* 驗證損失可能比訓練損失高得多，並且在訓練過程中可能開始增加——這是因為模型“記住”了訓練數據點，卻失去了對“整體情況”的把握。\n", "\n", - "![過擬合](../../../../../translated_images/mo/overfit.a0bd57f717c15769.webp)\n", + "![過擬合](../../../../../translated_images/zh-MO/overfit.a0bd57f717c15769.webp)\n", "\n", "> 在這張圖中，`x` 代表訓練數據，`o` 代表驗證數據。左邊是線性模型（單層），它很好地接近了數據的本質。右邊是過擬合模型，該模型完美地接近了訓練數據，但對其他數據（驗證數據）的表現卻變得毫無意義（驗證誤差非常高）。\n" ] diff --git a/translations/mo/lessons/3-NeuralNetworks/05-Frameworks/README.md b/translations/mo/lessons/3-NeuralNetworks/05-Frameworks/README.md index 6b1b7e27..f3de56bd 100644 --- a/translations/mo/lessons/3-NeuralNetworks/05-Frameworks/README.md +++ b/translations/mo/lessons/3-NeuralNetworks/05-Frameworks/README.md @@ -58,7 +58,7 @@ CO_OP_TRANSLATOR_METADATA: 考慮以下近似 5 個點（圖中的 `x`）的問題： -![線性模型](../../../../../translated_images/mo/overfit1.f24b71c6f652e59e.webp) | ![過擬合模型](../../../../../translated_images/mo/overfit2.131f5800ae10ca5e.webp) +![線性模型](../../../../../translated_images/zh-MO/overfit1.f24b71c6f652e59e.webp) | ![過擬合模型](../../../../../translated_images/zh-MO/overfit2.131f5800ae10ca5e.webp) -------------------------|-------------------------- **線性模型，2 個參數** | **非線性模型，7 個參數** 訓練誤差 = 5.3 | 訓練誤差 = 0 @@ -79,7 +79,7 @@ CO_OP_TRANSLATOR_METADATA: 如上圖所示，過擬合可以通過非常低的訓練誤差和非常高的驗證誤差來檢測。通常在訓練過程中，我們會看到訓練誤差和驗證誤差都開始下降，然後在某個時候驗證誤差可能停止下降並開始上升。這將是過擬合的跡象，也是我們應該停止訓練的指示（或者至少保存模型的快照）。 -![過擬合](../../../../../translated_images/mo/Overfitting.408ad91cd90b4371.webp) +![過擬合](../../../../../translated_images/zh-MO/Overfitting.408ad91cd90b4371.webp) ## 如何防止過擬合 diff --git a/translations/mo/lessons/3-NeuralNetworks/README.md b/translations/mo/lessons/3-NeuralNetworks/README.md index a72f1ec0..75d887aa 100644 --- a/translations/mo/lessons/3-NeuralNetworks/README.md +++ b/translations/mo/lessons/3-NeuralNetworks/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 神經網絡簡介 -![神經網絡簡介內容摘要的手繪圖](../../../../translated_images/mo/ai-neuralnetworks.1c687ae40bc86e83.webp) +![神經網絡簡介內容摘要的手繪圖](../../../../translated_images/zh-MO/ai-neuralnetworks.1c687ae40bc86e83.webp) 如我們在介紹中所討論的，實現智能的一種方法是訓練一個**計算機模型**或**人工大腦**。自20世紀中期以來，研究人員嘗試了不同的數學模型，直到最近這一方向取得了巨大成功。這些模仿大腦的數學模型被稱為**神經網絡**。 @@ -36,13 +36,13 @@ CO_OP_TRANSLATOR_METADATA: 從生物學中，我們知道大腦由神經細胞（神經元）組成，每個神經元都有多個“輸入”（樹突）和一個“輸出”（軸突）。樹突和軸突都可以傳導電信號，而它們之間的連接——稱為突觸——可以表現出不同程度的導電性，這些導電性由神經遞質調節。 -![神經元模型](../../../../translated_images/mo/synapse-wikipedia.ed20a9e4726ea1c6.webp) | ![神經元模型](../../../../translated_images/mo/artneuron.1a5daa88d20ebe6f.webp) +![神經元模型](../../../../translated_images/zh-MO/synapse-wikipedia.ed20a9e4726ea1c6.webp) | ![神經元模型](../../../../translated_images/zh-MO/artneuron.1a5daa88d20ebe6f.webp) ----|---- 真實神經元 *（[圖片](https://en.wikipedia.org/wiki/Synapse#/media/File:SynapseSchematic_lines.svg) 來自維基百科）* | 人工神經元 *（作者提供圖片）* 因此，神經元的最簡單數學模型包含幾個輸入 X1, ..., XN 和一個輸出 Y，以及一系列權重 W1, ..., WN。輸出計算公式為： -Y = f\left(\sum_{i=1}^N X_iW_i\right) +Y = f\left(\sum_{i=1}^N X_iW_i\right) 其中 f 是某種非線性的**激活函數**。 diff --git a/translations/mo/lessons/4-ComputerVision/06-IntroCV/README.md b/translations/mo/lessons/4-ComputerVision/06-IntroCV/README.md index c2c46dbf..c170e3fa 100644 --- a/translations/mo/lessons/4-ComputerVision/06-IntroCV/README.md +++ b/translations/mo/lessons/4-ComputerVision/06-IntroCV/README.md @@ -73,14 +73,14 @@ im = cv2.cvtColor(im,cv2.COLOR_BGR2RGB) * **預處理盲文書的照片**。我們專注於如何使用閾值處理、特徵檢測、透視變換和 NumPy 操作來分離單個盲文符號，以便進一步由神經網路進行分類。 -![盲文影像](../../../../../translated_images/mo/braille.341962ff76b1bd70.webp) | ![盲文影像預處理結果](../../../../../translated_images/mo/braille-result.46530fea020b03c7.webp) | ![盲文符號](../../../../../translated_images/mo/braille-symbols.0159185ab69d5339.webp) +![盲文影像](../../../../../translated_images/zh-MO/braille.341962ff76b1bd70.webp) | ![盲文影像預處理結果](../../../../../translated_images/zh-MO/braille-result.46530fea020b03c7.webp) | ![盲文符號](../../../../../translated_images/zh-MO/braille-symbols.0159185ab69d5339.webp) ----|-----|----- > 圖片來自 [OpenCV.ipynb](OpenCV.ipynb) * **使用幀差檢測影片中的運動**。如果相機固定，則來自相機的幀應該彼此非常相似。由於幀表示為陣列，只需對兩個連續幀的陣列進行相減，我們就能得到像素差異，靜態幀的差異應該很低，而當影像中有顯著運動時，差異會變高。 -![影片幀和幀差的影像](../../../../../translated_images/mo/frame-difference.706f805491a0883c.webp) +![影片幀和幀差的影像](../../../../../translated_images/zh-MO/frame-difference.706f805491a0883c.webp) > 圖片來自 [OpenCV.ipynb](OpenCV.ipynb) @@ -88,7 +88,7 @@ im = cv2.cvtColor(im,cv2.COLOR_BGR2RGB) - **密集光流** 計算顯示每個像素移動方向的向量場 - **稀疏光流** 基於提取影像中的一些顯著特徵（例如邊緣），並從幀到幀構建它們的軌跡。 -![光流影像](../../../../../translated_images/mo/optical.1f4a94464579a83a.webp) +![光流影像](../../../../../translated_images/zh-MO/optical.1f4a94464579a83a.webp) > 圖片來自 [OpenCV.ipynb](OpenCV.ipynb) diff --git a/translations/mo/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md b/translations/mo/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md index 91c9d360..00952ea0 100644 --- a/translations/mo/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md +++ b/translations/mo/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md @@ -13,11 +13,11 @@ CO_OP_TRANSLATOR_METADATA: VGG-16 是一個在 2014 年 ImageNet top-5 分類中達到 92.7% 準確率的網路。它的層結構如下： -![ImageNet Layers](../../../../../translated_images/mo/vgg-16-arch1.d901a5583b3a51ba.webp) +![ImageNet Layers](../../../../../translated_images/zh-MO/vgg-16-arch1.d901a5583b3a51ba.webp) 如圖所示，VGG 採用傳統的金字塔架構，即一系列的卷積-池化層。 -![ImageNet Pyramid](../../../../../translated_images/mo/vgg-16-arch.64ff2137f50dd49f.webp) +![ImageNet Pyramid](../../../../../translated_images/zh-MO/vgg-16-arch.64ff2137f50dd49f.webp) > 圖片來源：[Researchgate](https://www.researchgate.net/figure/Vgg16-model-structure-To-get-the-VGG-NIN-model-we-replace-the-2-nd-4-th-6-th-7-th_fig2_335194493) diff --git a/translations/mo/lessons/4-ComputerVision/07-ConvNets/ConvNetsPyTorch.ipynb b/translations/mo/lessons/4-ComputerVision/07-ConvNets/ConvNetsPyTorch.ipynb index 43bf8832..d1075ebc 100644 --- a/translations/mo/lessons/4-ComputerVision/07-ConvNets/ConvNetsPyTorch.ipynb +++ b/translations/mo/lessons/4-ComputerVision/07-ConvNets/ConvNetsPyTorch.ipynb @@ -260,7 +260,7 @@ "\n", "因此，在典型的 CNN 中，會有多個卷積層，並在它們之間加入池化層以減少影像的維度。我們還會增加濾波器的數量，因為隨著模式變得更複雜，我們需要尋找的可能組合也會更多。\n", "\n", - "![一張展示多個卷積層和池化層的圖片。](../../../../../translated_images/mo/cnn-pyramid.85915455759ef0ce.webp)\n", + "![一張展示多個卷積層和池化層的圖片。](../../../../../translated_images/zh-MO/cnn-pyramid.85915455759ef0ce.webp)\n", "\n", "由於空間維度減少以及特徵/濾波器維度增加，這種架構也被稱為 **金字塔架構**。\n" ] diff --git a/translations/mo/lessons/4-ComputerVision/07-ConvNets/ConvNetsTF.ipynb b/translations/mo/lessons/4-ComputerVision/07-ConvNets/ConvNetsTF.ipynb index dc73ecae..e3a502f0 100644 --- a/translations/mo/lessons/4-ComputerVision/07-ConvNets/ConvNetsTF.ipynb +++ b/translations/mo/lessons/4-ComputerVision/07-ConvNets/ConvNetsTF.ipynb @@ -359,7 +359,7 @@ "\n", "因此，在典型的 CNN 中，會有多個卷積層，並在它們之間插入池化層以減少圖片的尺寸。同時，我們還會增加濾波器的數量，因為隨著模式變得更加複雜，我們需要尋找的可能組合也會更多。\n", "\n", - "![一張展示多個卷積層與池化層的圖片。](../../../../../translated_images/mo/cnn-pyramid.85915455759ef0ce.webp)\n", + "![一張展示多個卷積層與池化層的圖片。](../../../../../translated_images/zh-MO/cnn-pyramid.85915455759ef0ce.webp)\n", "\n", "由於空間尺寸逐漸減小，而特徵/濾波器的維度逐漸增加，這種架構也被稱為 **金字塔架構**。\n" ] diff --git a/translations/mo/lessons/4-ComputerVision/07-ConvNets/README.md b/translations/mo/lessons/4-ComputerVision/07-ConvNets/README.md index 6bcbb9af..bfa3d7c1 100644 --- a/translations/mo/lessons/4-ComputerVision/07-ConvNets/README.md +++ b/translations/mo/lessons/4-ComputerVision/07-ConvNets/README.md @@ -17,7 +17,7 @@ CO_OP_TRANSLATOR_METADATA: 為了提取模式，我們將使用**卷積濾波器**的概念。正如你所知，圖像可以用二維矩陣或帶有色彩深度的三維張量來表示。應用濾波器意味著我們取一個相對較小的**濾波核**矩陣，並對原始圖像中的每個像素與其鄰近點進行加權平均。我們可以將其視為一個小窗口在整個圖像上滑動，並根據濾波核矩陣中的權重對所有像素進行平均。 -![垂直邊緣濾波器](../../../../../translated_images/mo/filter-vert.b7148390ca0bc356.webp) | ![水平邊緣濾波器](../../../../../translated_images/mo/filter-horiz.59b80ed4feb946ef.webp) +![垂直邊緣濾波器](../../../../../translated_images/zh-MO/filter-vert.b7148390ca0bc356.webp) | ![水平邊緣濾波器](../../../../../translated_images/zh-MO/filter-horiz.59b80ed4feb946ef.webp) ----|---- > 圖片來源：Dmitry Soshnikov @@ -38,7 +38,7 @@ CNN 的工作方式基於以下重要思想： * 我們可以設計網絡，使濾波器能夠自動訓練 * 我們可以使用相同的方法來在高層次特徵中找到模式，而不僅僅是在原始圖像中。因此，CNN 的特徵提取在特徵層次上工作，從低層次的像素組合開始，到更高層次的圖片部分組合。 -![層次特徵提取](../../../../../translated_images/mo/FeatureExtractionCNN.d9b456cbdae7cb64.webp) +![層次特徵提取](../../../../../translated_images/zh-MO/FeatureExtractionCNN.d9b456cbdae7cb64.webp) > 圖片來源：[Hislop-Lynch 的論文](https://www.semanticscholar.org/paper/Computer-vision-based-pedestrian-trajectory-Hislop-Lynch/26e6f74853fc9bbb7487b06dc2cf095d36c9021d)，基於[他們的研究](https://dl.acm.org/doi/abs/10.1145/1553374.1553453) @@ -55,9 +55,9 @@ CNN 的工作方式基於以下重要思想： 例如，讓我們看看 VGG-16 的架構，這是一個在 2014 年 ImageNet 的 top-5 分類中達到 92.7% 準確率的網絡： -![ImageNet 層](../../../../../translated_images/mo/vgg-16-arch1.d901a5583b3a51ba.webp) +![ImageNet 層](../../../../../translated_images/zh-MO/vgg-16-arch1.d901a5583b3a51ba.webp) -![ImageNet 金字塔](../../../../../translated_images/mo/vgg-16-arch.64ff2137f50dd49f.webp) +![ImageNet 金字塔](../../../../../translated_images/zh-MO/vgg-16-arch.64ff2137f50dd49f.webp) > 圖片來源：[Researchgate](https://www.researchgate.net/figure/Vgg16-model-structure-To-get-the-VGG-NIN-model-we-replace-the-2-nd-4-th-6-th-7-th_fig2_335194493) diff --git a/translations/mo/lessons/4-ComputerVision/07-ConvNets/lab/README.md b/translations/mo/lessons/4-ComputerVision/07-ConvNets/lab/README.md index bcfcb330..108ee161 100644 --- a/translations/mo/lessons/4-ComputerVision/07-ConvNets/lab/README.md +++ b/translations/mo/lessons/4-ComputerVision/07-ConvNets/lab/README.md @@ -21,7 +21,7 @@ CO_OP_TRANSLATOR_METADATA: 我們將使用 [Oxford-IIIT Pet Dataset](https://www.robots.ox.ac.uk/~vgg/data/pets/)，該數據集包含 37 種不同品種的狗和貓的圖片。 -![我們將處理的數據集](../../../../../../translated_images/mo/data.50b2a9d5484bdbf0.webp) +![我們將處理的數據集](../../../../../../translated_images/zh-MO/data.50b2a9d5484bdbf0.webp) 要下載數據集，請使用以下程式碼片段： diff --git a/translations/mo/lessons/4-ComputerVision/08-TransferLearning/AdversarialCat_TF.ipynb b/translations/mo/lessons/4-ComputerVision/08-TransferLearning/AdversarialCat_TF.ipynb index d40faa20..3c2d3bdb 100644 --- a/translations/mo/lessons/4-ComputerVision/08-TransferLearning/AdversarialCat_TF.ipynb +++ b/translations/mo/lessons/4-ComputerVision/08-TransferLearning/AdversarialCat_TF.ipynb @@ -50,7 +50,7 @@ "\n", "為了呈現理想的貓，我們將從一張隨機噪點圖片開始，並嘗試使用梯度下降優化技術來調整圖片，使網絡能夠識別出貓。\n", "\n", - "![優化循環](../../../../../translated_images/mo/ideal-cat-loop.999fbb8ff306e044.webp)\n", + "![優化循環](../../../../../translated_images/zh-MO/ideal-cat-loop.999fbb8ff306e044.webp)\n", "\n", "以下是我們的起始圖片：\n" ] diff --git a/translations/mo/lessons/4-ComputerVision/08-TransferLearning/README.md b/translations/mo/lessons/4-ComputerVision/08-TransferLearning/README.md index 51acb903..43e493e4 100644 --- a/translations/mo/lessons/4-ComputerVision/08-TransferLearning/README.md +++ b/translations/mo/lessons/4-ComputerVision/08-TransferLearning/README.md @@ -29,7 +29,7 @@ Keras 和 PyTorch 都包含函數，可以輕鬆加載一些常見架構的預 以下是 VGG-16 網絡從一張貓的圖片中提取的特徵示例： -![VGG-16 提取的特徵](../../../../../translated_images/mo/features.6291f9c7ba3a0b95.webp) +![VGG-16 提取的特徵](../../../../../translated_images/zh-MO/features.6291f9c7ba3a0b95.webp) ## 貓與狗數據集 @@ -48,19 +48,19 @@ Keras 和 PyTorch 都包含函數，可以輕鬆加載一些常見架構的預 我們可以採取的一種方法是從一張隨機圖像開始，然後嘗試使用**梯度下降優化**技術調整該圖像，使網絡開始認為它是一隻貓。 -![圖像優化循環](../../../../../translated_images/mo/ideal-cat-loop.999fbb8ff306e044.webp) +![圖像優化循環](../../../../../translated_images/zh-MO/ideal-cat-loop.999fbb8ff306e044.webp) 然而，如果我們這樣做，結果會非常接近隨機噪聲。這是因為*有很多方法可以讓網絡認為輸入圖像是一隻貓*，包括一些在視覺上沒有意義的方式。雖然這些圖像包含了許多典型的貓的模式，但並沒有任何約束使它們在視覺上更具辨識性。 為了改善結果，我們可以在損失函數中添加另一個項，稱為**變異損失**。它是一種度量，顯示圖像中相鄰像素的相似程度。最小化變異損失可以使圖像更平滑，並消除噪聲——從而揭示更具視覺吸引力的模式。以下是一些“理想”圖像的示例，它們被高概率分類為貓和斑馬： -![理想貓](../../../../../translated_images/mo/ideal-cat.203dd4597643d6b0.webp) | ![理想斑馬](../../../../../translated_images/mo/ideal-zebra.7f70e8b54ee15a7a.webp) +![理想貓](../../../../../translated_images/zh-MO/ideal-cat.203dd4597643d6b0.webp) | ![理想斑馬](../../../../../translated_images/zh-MO/ideal-zebra.7f70e8b54ee15a7a.webp) -----|----- *理想貓* | *理想斑馬* 類似的方法可以用於對神經網絡進行所謂的**對抗性攻擊**。假設我們想要欺騙神經網絡，使一隻狗看起來像一隻貓。如果我們拿一張狗的圖片，該圖片被網絡識別為狗，然後稍微調整它，使用梯度下降優化，直到網絡開始將其分類為貓： -![狗的圖片](../../../../../translated_images/mo/original-dog.8f68a67d2fe0911f.webp) | ![被分類為貓的狗圖片](../../../../../translated_images/mo/adversarial-dog.d9fc7773b0142b89.webp) +![狗的圖片](../../../../../translated_images/zh-MO/original-dog.8f68a67d2fe0911f.webp) | ![被分類為貓的狗圖片](../../../../../translated_images/zh-MO/adversarial-dog.d9fc7773b0142b89.webp) -----|----- *狗的原始圖片* | *被分類為貓的狗圖片* diff --git a/translations/mo/lessons/4-ComputerVision/09-Autoencoders/AutoEncodersPyTorch.ipynb b/translations/mo/lessons/4-ComputerVision/09-Autoencoders/AutoEncodersPyTorch.ipynb index 8ab0953b..5494fd31 100644 --- a/translations/mo/lessons/4-ComputerVision/09-Autoencoders/AutoEncodersPyTorch.ipynb +++ b/translations/mo/lessons/4-ComputerVision/09-Autoencoders/AutoEncodersPyTorch.ipynb @@ -21,7 +21,7 @@ "\n", "由於我們訓練自編碼器的目的是儘可能捕捉原始圖像中的信息以進行準確的重建，網絡會嘗試找到輸入圖像的最佳**嵌入（embedding）**來捕捉其含義。\n", "\n", - "![自編碼器示意圖](../../../../../translated_images/mo/autoencoder_schema.5e6fc9ad98a5eb61.webp)\n", + "![自編碼器示意圖](../../../../../translated_images/zh-MO/autoencoder_schema.5e6fc9ad98a5eb61.webp)\n", "\n", "> 圖片來源：[Keras 博客](https://blog.keras.io/building-autoencoders-in-keras.html)\n", "\n", diff --git a/translations/mo/lessons/4-ComputerVision/09-Autoencoders/AutoencodersTF.ipynb b/translations/mo/lessons/4-ComputerVision/09-Autoencoders/AutoencodersTF.ipynb index b6d58f9e..68a9e6e6 100644 --- a/translations/mo/lessons/4-ComputerVision/09-Autoencoders/AutoencodersTF.ipynb +++ b/translations/mo/lessons/4-ComputerVision/09-Autoencoders/AutoencodersTF.ipynb @@ -19,7 +19,7 @@ "\n", "由於我們訓練自編碼器的目的是捕捉原始圖像中的盡可能多的信息以進行準確的重建，網絡會嘗試找到輸入圖像的最佳**嵌入**來捕捉其含義。\n", "\n", - "![自編碼器示意圖](../../../../../translated_images/mo/autoencoder_schema.5e6fc9ad98a5eb61.webp)\n", + "![自編碼器示意圖](../../../../../translated_images/zh-MO/autoencoder_schema.5e6fc9ad98a5eb61.webp)\n", "\n", "*圖片來源：[Keras 部落格](https://blog.keras.io/building-autoencoders-in-keras.html)*\n", "\n", diff --git a/translations/mo/lessons/4-ComputerVision/09-Autoencoders/README.md b/translations/mo/lessons/4-ComputerVision/09-Autoencoders/README.md index 9f5d01b4..144f4b97 100644 --- a/translations/mo/lessons/4-ComputerVision/09-Autoencoders/README.md +++ b/translations/mo/lessons/4-ComputerVision/09-Autoencoders/README.md @@ -19,7 +19,7 @@ CO_OP_TRANSLATOR_METADATA: 由於我們訓練自動編碼器的目的是捕捉原始圖像中的盡可能多的信息以進行準確的重建，網絡會嘗試找到最佳的**嵌入**方式來捕捉輸入圖像的含義。 -![自動編碼器示意圖](../../../../../translated_images/mo/autoencoder_schema.5e6fc9ad98a5eb61.webp) +![自動編碼器示意圖](../../../../../translated_images/zh-MO/autoencoder_schema.5e6fc9ad98a5eb61.webp) > 圖片來源：[Keras 博客](https://blog.keras.io/building-autoencoders-in-keras.html) diff --git a/translations/mo/lessons/4-ComputerVision/11-ObjectDetection/README.md b/translations/mo/lessons/4-ComputerVision/11-ObjectDetection/README.md index cd75d16c..1f5a00f3 100644 --- a/translations/mo/lessons/4-ComputerVision/11-ObjectDetection/README.md +++ b/translations/mo/lessons/4-ComputerVision/11-ObjectDetection/README.md @@ -13,7 +13,7 @@ CO_OP_TRANSLATOR_METADATA: ## [課前測驗](https://ff-quizzes.netlify.app/en/ai/quiz/21) -![物件偵測](../../../../../translated_images/mo/Screen_Shot_2016-11-17_at_11.14.54_AM.b4bb3769353287be.webp) +![物件偵測](../../../../../translated_images/zh-MO/Screen_Shot_2016-11-17_at_11.14.54_AM.b4bb3769353287be.webp) > 圖片來源：[YOLO v2 網站](https://pjreddie.com/darknet/yolov2/) @@ -25,7 +25,7 @@ CO_OP_TRANSLATOR_METADATA: 2. 對每個區塊進行影像分類。 3. 將分類結果中激活值足夠高的區塊視為包含目標物件的區域。 -![簡單的物件偵測](../../../../../translated_images/mo/naive-detection.e7f1ba220ccd08c6.webp) +![簡單的物件偵測](../../../../../translated_images/zh-MO/naive-detection.e7f1ba220ccd08c6.webp) > *圖片來源：[練習筆記本](ObjectDetection-TF.ipynb)* @@ -42,7 +42,7 @@ CO_OP_TRANSLATOR_METADATA: * [PASCAL VOC](http://host.robots.ox.ac.uk/pascal/VOC/) - 包含 20 個類別 * [COCO](http://cocodataset.org/#home) - 常見物件的上下文。包含 80 個類別、邊界框和分割遮罩 -![COCO](../../../../../translated_images/mo/coco-examples.71bc60380fa6cceb.webp) +![COCO](../../../../../translated_images/zh-MO/coco-examples.71bc60380fa6cceb.webp) ## 物件偵測的評估指標 @@ -50,7 +50,7 @@ CO_OP_TRANSLATOR_METADATA: 對於影像分類來說，衡量演算法的表現相對簡單；但對於物件偵測，我們需要同時衡量類別的正確性以及推測邊界框位置的精確性。後者使用所謂的**交集比聯集**（IoU）來衡量，這是一種用來評估兩個框（或任意兩個區域）重疊程度的方法。 -![IoU](../../../../../translated_images/mo/iou_equation.9a4751d40fff4e11.webp) +![IoU](../../../../../translated_images/zh-MO/iou_equation.9a4751d40fff4e11.webp) > *圖片來源：[這篇優秀的 IoU 部落格文章](https://pyimagesearch.com/2016/11/07/intersection-over-union-iou-for-object-detection/)* @@ -98,11 +98,11 @@ $$ [R-CNN](http://islab.ulsan.ac.kr/files/announcement/513/rcnn_pami.pdf) 使用 [Selective Search](http://www.huppelen.nl/publications/selectiveSearchDraft.pdf) 生成層次結構的 ROI 區域，然後通過 CNN 特徵提取器和 SVM 分類器來確定物件類別，並通過線性迴歸確定*邊界框*座標。[官方論文](https://arxiv.org/pdf/1506.01497v1.pdf) -![RCNN](../../../../../translated_images/mo/rcnn1.cae407020dfb1d1f.webp) +![RCNN](../../../../../translated_images/zh-MO/rcnn1.cae407020dfb1d1f.webp) > *圖片來源：van de Sande et al. ICCV’11* -![RCNN-1](../../../../../translated_images/mo/rcnn2.2d9530bb83516484.webp) +![RCNN-1](../../../../../translated_images/zh-MO/rcnn2.2d9530bb83516484.webp) > *圖片來源：[這篇部落格](https://towardsdatascience.com/r-cnn-fast-r-cnn-faster-r-cnn-yolo-object-detection-algorithms-36d53571365e)* @@ -110,7 +110,7 @@ $$ 這種方法與 R-CNN 類似，但區域是在卷積層應用之後定義的。 -![FRCNN](../../../../../translated_images/mo/f-rcnn.3cda6d9bb4188875.webp) +![FRCNN](../../../../../translated_images/zh-MO/f-rcnn.3cda6d9bb4188875.webp) > 圖片來源：[官方論文](https://www.cv-foundation.org/openaccess/content_iccv_2015/papers/Girshick_Fast_R-CNN_ICCV_2015_paper.pdf)，[arXiv](https://arxiv.org/pdf/1504.08083.pdf)，2015 @@ -118,7 +118,7 @@ $$ 這種方法的主要思想是使用神經網路來預測 ROI，即所謂的*區域提議網路*。[論文](https://arxiv.org/pdf/1506.01497.pdf)，2016 -![FasterRCNN](../../../../../translated_images/mo/faster-rcnn.8d46c099b87ef30a.webp) +![FasterRCNN](../../../../../translated_images/zh-MO/faster-rcnn.8d46c099b87ef30a.webp) > 圖片來源：[官方論文](https://arxiv.org/pdf/1506.01497.pdf) @@ -130,7 +130,7 @@ $$ 2. 特徵經過**位置敏感分數圖**處理。每個來自 $C$ 類別的物件被劃分為 $k\times k$ 區域，並訓練網路預測物件的部分。 3. 對於 $k\times k$ 區域中的每個部分，所有網路對物件類別進行投票，選擇得票最多的物件類別。 -![r-fcn 圖片](../../../../../translated_images/mo/r-fcn.13eb88158b99a3da.webp) +![r-fcn 圖片](../../../../../translated_images/zh-MO/r-fcn.13eb88158b99a3da.webp) > 圖片來源：[官方論文](https://arxiv.org/abs/1605.06409) @@ -141,7 +141,7 @@ YOLO 是一種實時的單次通過演算法。主要思想如下： * 將圖片劃分為 $S\times S$ 區域。 * 對於每個區域，**CNN** 預測 $n$ 個可能的物件、*邊界框*座標以及*置信度*=*概率* * IoU。 - ![YOLO](../../../../../translated_images/mo/yolo.a2648ec82ee8bb4e.webp) + ![YOLO](../../../../../translated_images/zh-MO/yolo.a2648ec82ee8bb4e.webp) > 圖片來源：[官方論文](https://arxiv.org/abs/1506.02640) diff --git a/translations/mo/lessons/4-ComputerVision/README.md b/translations/mo/lessons/4-ComputerVision/README.md index 0a1a408e..12e34b29 100644 --- a/translations/mo/lessons/4-ComputerVision/README.md +++ b/translations/mo/lessons/4-ComputerVision/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 電腦視覺 -![電腦視覺內容摘要的手繪圖](../../../../translated_images/mo/ai-computervision.6506ebebac3fbf76.webp) +![電腦視覺內容摘要的手繪圖](../../../../translated_images/zh-MO/ai-computervision.6506ebebac3fbf76.webp) 在本章節中，我們將學習以下內容： diff --git a/translations/mo/lessons/5-NLP/13-TextRep/TextRepresentationPyTorch.ipynb b/translations/mo/lessons/5-NLP/13-TextRep/TextRepresentationPyTorch.ipynb index d9d2e1f6..93ea766b 100644 --- a/translations/mo/lessons/5-NLP/13-TextRep/TextRepresentationPyTorch.ipynb +++ b/translations/mo/lessons/5-NLP/13-TextRep/TextRepresentationPyTorch.ipynb @@ -199,7 +199,7 @@ "\n", "**詞袋** (BoW) 向量表示法是最常用的傳統向量表示法。每個詞語都與一個向量索引相關聯，向量元素包含某個詞語在特定文檔中出現的次數。\n", "\n", - "![顯示詞袋向量表示法在記憶體中如何表示的圖片。](../../../../../translated_images/mo/bag-of-words-example.606fc1738f1d7ba9.webp) \n", + "![顯示詞袋向量表示法在記憶體中如何表示的圖片。](../../../../../translated_images/zh-MO/bag-of-words-example.606fc1738f1d7ba9.webp) \n", "\n", "> **Note**: 你也可以將 BoW 理解為文本中每個詞語的單熱編碼向量的總和。\n", "\n", diff --git a/translations/mo/lessons/5-NLP/13-TextRep/TextRepresentationTF.ipynb b/translations/mo/lessons/5-NLP/13-TextRep/TextRepresentationTF.ipynb index 9be7de31..11f7eea2 100644 --- a/translations/mo/lessons/5-NLP/13-TextRep/TextRepresentationTF.ipynb +++ b/translations/mo/lessons/5-NLP/13-TextRep/TextRepresentationTF.ipynb @@ -188,7 +188,7 @@ "\n", "**詞袋**（Bag-of-words, BoW）向量表示法是最簡單易懂的傳統向量表示法。每個詞語都對應到向量中的一個索引，而向量中的元素則表示該詞語在特定文檔中出現的次數。\n", "\n", - "![顯示詞袋向量表示法在記憶體中如何表示的圖片。](../../../../../translated_images/mo/bag-of-words-example.606fc1738f1d7ba9.webp) \n", + "![顯示詞袋向量表示法在記憶體中如何表示的圖片。](../../../../../translated_images/zh-MO/bag-of-words-example.606fc1738f1d7ba9.webp) \n", "\n", "> **注意**：你也可以將 BoW 理解為文本中每個詞語的單熱編碼（one-hot-encoded）向量的總和。\n", "\n", diff --git a/translations/mo/lessons/5-NLP/14-Embeddings/EmbeddingsPyTorch.ipynb b/translations/mo/lessons/5-NLP/14-Embeddings/EmbeddingsPyTorch.ipynb index 884cac4c..3a9b8ff7 100644 --- a/translations/mo/lessons/5-NLP/14-Embeddings/EmbeddingsPyTorch.ipynb +++ b/translations/mo/lessons/5-NLP/14-Embeddings/EmbeddingsPyTorch.ipynb @@ -62,7 +62,7 @@ "\n", "通過將嵌入層作為我們網絡的第一層，我們可以從詞袋模型（bag-of-words）切換到 **嵌入袋模型**（embedding bag model）。在這種模型中，我們首先將文本中的每個單詞轉換為對應的嵌入向量，然後對所有這些嵌入向量執行某種聚合函數，例如 `sum`、`average` 或 `max`。\n", "\n", - "![展示一個針對五個序列單詞的嵌入分類器的圖片。](../../../../../translated_images/mo/embedding-classifier-example.b77f021a7ee67eee.webp)\n", + "![展示一個針對五個序列單詞的嵌入分類器的圖片。](../../../../../translated_images/zh-MO/embedding-classifier-example.b77f021a7ee67eee.webp)\n", "\n", "我們的分類器神經網絡將以嵌入層開始，接著是聚合層，最後在其上添加一個線性分類器：\n" ] @@ -176,7 +176,7 @@ "\n", "在之前的架構中，我們需要將所有序列填充（pad）到相同的長度，才能將它們放入一個小批次中。這並不是表示變長序列最有效率的方法——另一種方法是使用 **offset** 向量，該向量會保存所有序列在一個大型向量中的偏移量。\n", "\n", - "![顯示偏移序列表示法的圖片](../../../../../translated_images/mo/offset-sequence-representation.eb73fcefb29b46ee.webp)\n", + "![顯示偏移序列表示法的圖片](../../../../../translated_images/zh-MO/offset-sequence-representation.eb73fcefb29b46ee.webp)\n", "\n", "> **注意**：在上圖中，我們展示的是一個字符序列，但在我們的例子中，我們處理的是單詞序列。然而，使用偏移向量表示序列的基本原則是相同的。\n", "\n", @@ -311,7 +311,7 @@ "\n", "CBoW 的訓練速度較快，而 Skip-Gram 雖然較慢，但在表示不常見單詞方面表現更好。\n", "\n", - "![顯示 CBoW 和 Skip-Gram 算法如何將單詞轉換為向量的圖片。](../../../../../translated_images/mo/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp)\n", + "![顯示 CBoW 和 Skip-Gram 算法如何將單詞轉換為向量的圖片。](../../../../../translated_images/zh-MO/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp)\n", "\n", "為了試驗在 Google News 數據集上預訓練的 Word2Vec 嵌入，我們可以使用 **gensim** 庫。以下是找到與 'neural' 最相似的單詞的示例：\n", "\n", diff --git a/translations/mo/lessons/5-NLP/14-Embeddings/EmbeddingsTF.ipynb b/translations/mo/lessons/5-NLP/14-Embeddings/EmbeddingsTF.ipynb index 383f40e2..db93199b 100644 --- a/translations/mo/lessons/5-NLP/14-Embeddings/EmbeddingsTF.ipynb +++ b/translations/mo/lessons/5-NLP/14-Embeddings/EmbeddingsTF.ipynb @@ -37,7 +37,7 @@ "\n", "通過將嵌入層作為我們網絡的第一層，我們可以從詞袋模型（bag-of-words）切換到 **嵌入袋模型**（embedding bag），在這裡我們首先將文本中的每個單詞轉換為對應的嵌入，然後對所有這些嵌入計算某種聚合函數，例如 `sum`、`average` 或 `max`。\n", "\n", - "![顯示五個序列單詞的嵌入分類器的圖片。](../../../../../translated_images/mo/embedding-classifier-example.b77f021a7ee67eee.webp)\n", + "![顯示五個序列單詞的嵌入分類器的圖片。](../../../../../translated_images/zh-MO/embedding-classifier-example.b77f021a7ee67eee.webp)\n", "\n", "我們的分類器神經網絡由以下幾層組成：\n", "\n", @@ -283,7 +283,7 @@ "\n", "CBoW 的速度較快，而 Skip-Gram 雖然較慢，但在表示不常見詞方面表現更好。\n", "\n", - "![展示 CBoW 和 Skip-Gram 算法如何將詞轉換為向量的圖片。](../../../../../translated_images/mo/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp)\n", + "![展示 CBoW 和 Skip-Gram 算法如何將詞轉換為向量的圖片。](../../../../../translated_images/zh-MO/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp)\n", "\n", "為了試驗基於 Google News 數據集預訓練的 Word2Vec 嵌入，我們可以使用 **gensim** 庫。以下是我們找到與「neural」最相似的詞。\n", "\n", diff --git a/translations/mo/lessons/5-NLP/14-Embeddings/README.md b/translations/mo/lessons/5-NLP/14-Embeddings/README.md index b558137b..d54c1d1a 100644 --- a/translations/mo/lessons/5-NLP/14-Embeddings/README.md +++ b/translations/mo/lessons/5-NLP/14-Embeddings/README.md @@ -19,7 +19,7 @@ CO_OP_TRANSLATOR_METADATA: 通過在分類器網絡中使用嵌入層作為第一層，我們可以從詞袋模型切換到 **嵌入袋** 模型。在嵌入袋模型中，我們首先將文本中的每個詞轉換為相應的嵌入，然後對所有嵌入計算某種聚合函數，例如 `sum`、`average` 或 `max`。 -![嵌入分類器處理五個序列詞的示例圖。](../../../../../translated_images/mo/embedding-classifier-example.b77f021a7ee67eee.webp) +![嵌入分類器處理五個序列詞的示例圖。](../../../../../translated_images/zh-MO/embedding-classifier-example.b77f021a7ee67eee.webp) > 圖片由作者提供 @@ -40,7 +40,7 @@ CO_OP_TRANSLATOR_METADATA: CBoW 的速度更快，而 Skip-Gram 雖然較慢，但在表示不常見詞方面效果更好。 -![展示 CBoW 和 Skip-Gram 將詞轉換為向量的算法示例圖。](../../../../../translated_images/mo/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) +![展示 CBoW 和 Skip-Gram 將詞轉換為向量的算法示例圖。](../../../../../translated_images/zh-MO/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) > 圖片來源：[這篇論文](https://arxiv.org/pdf/1301.3781.pdf) diff --git a/translations/mo/lessons/5-NLP/15-LanguageModeling/README.md b/translations/mo/lessons/5-NLP/15-LanguageModeling/README.md index 30c4a56a..1338bddb 100644 --- a/translations/mo/lessons/5-NLP/15-LanguageModeling/README.md +++ b/translations/mo/lessons/5-NLP/15-LanguageModeling/README.md @@ -23,7 +23,7 @@ CO_OP_TRANSLATOR_METADATA: * **連續詞袋模型**（CBoW）：在標記序列 $W_{-N}$, ..., $W_N$ 中預測中間的標記 $W_0$。 * **Skip-gram**：從中間標記 $W_0$ 預測一組相鄰的標記 {$W_{-N},\dots, W_{-1}, W_1,\dots, W_N$}。 -![來自將單詞轉換為向量的論文的圖片](../../../../../translated_images/mo/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) +![來自將單詞轉換為向量的論文的圖片](../../../../../translated_images/zh-MO/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) > 圖片來源：[這篇論文](https://arxiv.org/pdf/1301.3781.pdf) diff --git a/translations/mo/lessons/5-NLP/16-RNN/README.md b/translations/mo/lessons/5-NLP/16-RNN/README.md index 4460466c..0b5a1e76 100644 --- a/translations/mo/lessons/5-NLP/16-RNN/README.md +++ b/translations/mo/lessons/5-NLP/16-RNN/README.md @@ -15,7 +15,7 @@ CO_OP_TRANSLATOR_METADATA: 為了捕捉文本序列的意義，我們需要使用另一種神經網絡架構，稱為**循環神經網絡**（Recurrent Neural Network，簡稱 RNN）。在 RNN 中，我們將句子逐個符號地傳遞給網絡，網絡會生成某種**狀態**，然後將該狀態與下一個符號一起再次傳遞給網絡。 -![RNN](../../../../../translated_images/mo/rnn.27f5c29c53d727b5.webp) +![RNN](../../../../../translated_images/zh-MO/rnn.27f5c29c53d727b5.webp) > 圖片由作者提供 @@ -61,7 +61,7 @@ LSTM 網絡的組織方式與 RNN 類似，但有兩個狀態從層到層傳遞 循環網絡，無論是單向還是雙向，都能捕捉序列中的某些模式，並將它們存儲到狀態向量中或傳遞到輸出中。與卷積網絡類似，我們可以在第一層之上構建另一個循環層，以捕捉更高層次的模式，並基於第一層提取的低層次模式進行構建。這引出了**多層 RNN** 的概念，它由兩個或更多循環網絡組成，其中前一層的輸出作為輸入傳遞到下一層。 -![顯示多層長短期記憶 RNN 的圖片](../../../../../translated_images/mo/multi-layer-lstm.dd975e29bb2a59fe.webp) +![顯示多層長短期記憶 RNN 的圖片](../../../../../translated_images/zh-MO/multi-layer-lstm.dd975e29bb2a59fe.webp) *圖片來自 Fernando López 的[這篇精彩文章](https://towardsdatascience.com/from-a-lstm-cell-to-a-multilayer-lstm-network-with-pytorch-2899eb5696f3)* diff --git a/translations/mo/lessons/5-NLP/16-RNN/RNNPyTorch.ipynb b/translations/mo/lessons/5-NLP/16-RNN/RNNPyTorch.ipynb index 07ce9a47..6243cf62 100644 --- a/translations/mo/lessons/5-NLP/16-RNN/RNNPyTorch.ipynb +++ b/translations/mo/lessons/5-NLP/16-RNN/RNNPyTorch.ipynb @@ -422,7 +422,7 @@ "\n", "無論是單向還是雙向的循環網路，都能捕捉序列中的某些模式，並將其存儲到狀態向量中或傳遞到輸出中。與卷積網路類似，我們可以在第一層之上構建另一個循環層，以捕捉更高層次的模式，這些模式是由第一層提取的低層次模式構成的。這引出了 **多層 RNN** 的概念，它由兩層或更多的循環網路組成，前一層的輸出作為下一層的輸入。\n", "\n", - "![顯示多層長短期記憶 RNN 的圖片](../../../../../translated_images/mo/multi-layer-lstm.dd975e29bb2a59fe.webp)\n", + "![顯示多層長短期記憶 RNN 的圖片](../../../../../translated_images/zh-MO/multi-layer-lstm.dd975e29bb2a59fe.webp)\n", "\n", "*圖片來源：[這篇精彩的文章](https://towardsdatascience.com/from-a-lstm-cell-to-a-multilayer-lstm-network-with-pytorch-2899eb5696f3) 作者 Fernando López*\n", "\n", diff --git a/translations/mo/lessons/5-NLP/16-RNN/RNNTF.ipynb b/translations/mo/lessons/5-NLP/16-RNN/RNNTF.ipynb index 77d37d64..b7da1bbb 100644 --- a/translations/mo/lessons/5-NLP/16-RNN/RNNTF.ipynb +++ b/translations/mo/lessons/5-NLP/16-RNN/RNNTF.ipynb @@ -10,7 +10,7 @@ "\n", "為了捕捉文本序列的意義，我們將使用一種稱為**循環神經網路**（Recurrent Neural Network，簡稱 RNN）的神經網路架構。在使用 RNN 時，我們會將句子逐個標記（token）傳遞給網路，網路會生成某種**狀態**，然後我們將該狀態與下一個標記一起再次傳遞給網路。\n", "\n", - "![顯示循環神經網路生成示例的圖片。](../../../../../translated_images/mo/rnn.27f5c29c53d727b5.webp)\n", + "![顯示循環神經網路生成示例的圖片。](../../../../../translated_images/zh-MO/rnn.27f5c29c53d727b5.webp)\n", "\n", "給定標記輸入序列 $X_0,\\dots,X_n$，RNN 會創建一個神經網路區塊的序列，並通過反向傳播對該序列進行端到端訓練。每個網路區塊接受一對 $(X_i,S_i)$ 作為輸入，並生成 $S_{i+1}$ 作為結果。最終狀態 $S_n$ 或輸出 $Y_n$ 會進入線性分類器以生成結果。所有網路區塊共享相同的權重，並通過一次反向傳播訓練完成端到端學習。\n", "\n", @@ -369,7 +369,7 @@ "\n", "無論是單向還是雙向的循環網絡，都能捕捉序列中的模式，並將其存儲到狀態向量中或作為輸出返回。與卷積網絡類似，我們可以在第一層循環層之後再構建另一層循環層，以捕捉更高層次的模式，這些模式是由第一層提取的低層次模式構建而成的。這引出了 **多層 RNN** 的概念，它由兩層或更多層循環網絡組成，其中前一層的輸出作為下一層的輸入。\n", "\n", - "![顯示多層長短期記憶 RNN 的圖片](../../../../../translated_images/mo/multi-layer-lstm.dd975e29bb2a59fe.webp)\n", + "![顯示多層長短期記憶 RNN 的圖片](../../../../../translated_images/zh-MO/multi-layer-lstm.dd975e29bb2a59fe.webp)\n", "\n", "*圖片來源：[這篇精彩的文章](https://towardsdatascience.com/from-a-lstm-cell-to-a-multilayer-lstm-network-with-pytorch-2899eb5696f3)，作者 Fernando López。*\n", "\n", diff --git a/translations/mo/lessons/5-NLP/17-GenerativeNetworks/GenerativePyTorch.ipynb b/translations/mo/lessons/5-NLP/17-GenerativeNetworks/GenerativePyTorch.ipynb index ece3ecaa..119769ed 100644 --- a/translations/mo/lessons/5-NLP/17-GenerativeNetworks/GenerativePyTorch.ipynb +++ b/translations/mo/lessons/5-NLP/17-GenerativeNetworks/GenerativePyTorch.ipynb @@ -119,7 +119,7 @@ "\n", "我們將以以下方式訓練 RNN 來生成文本。在每一步中，我們會取一段長度為 `nchars` 的字元序列，並讓網路為每個輸入字元生成下一個輸出字元：\n", "\n", - "![顯示 RNN 生成單詞 'HELLO' 的範例圖像。](../../../../../translated_images/mo/rnn-generate.56c54afb52f9781d.webp)\n", + "![顯示 RNN 生成單詞 'HELLO' 的範例圖像。](../../../../../translated_images/zh-MO/rnn-generate.56c54afb52f9781d.webp)\n", "\n", "根據實際情況，我們可能還需要加入一些特殊字元，例如 *序列結束符* ``。在我們的例子中，我們只希望訓練網路進行無限文本生成，因此我們會將每個序列的大小固定為 `nchars` 個標記。因此，每個訓練樣本將包含 `nchars` 個輸入和 `nchars` 個輸出（輸出是將輸入序列向左移動一個符號後的結果）。一個小批次（minibatch）將由多個這樣的序列組成。\n", "\n", diff --git a/translations/mo/lessons/5-NLP/17-GenerativeNetworks/GenerativeTF.ipynb b/translations/mo/lessons/5-NLP/17-GenerativeNetworks/GenerativeTF.ipynb index 797f7153..20aaecd2 100644 --- a/translations/mo/lessons/5-NLP/17-GenerativeNetworks/GenerativeTF.ipynb +++ b/translations/mo/lessons/5-NLP/17-GenerativeNetworks/GenerativeTF.ipynb @@ -111,7 +111,7 @@ "\n", "我們將以以下方式訓練 RNN 來生成新聞標題。在每一步中，我們會取一個標題，將其輸入到 RNN 中，並對於每個輸入的字元，要求網路生成下一個輸出的字元：\n", "\n", - "![顯示 RNN 生成單詞 'HELLO' 的範例圖片。](../../../../../translated_images/mo/rnn-generate.56c54afb52f9781d.webp)\n", + "![顯示 RNN 生成單詞 'HELLO' 的範例圖片。](../../../../../translated_images/zh-MO/rnn-generate.56c54afb52f9781d.webp)\n", "\n", "對於序列中的最後一個字元，我們會要求網路生成 `` 標記。\n", "\n", diff --git a/translations/mo/lessons/5-NLP/17-GenerativeNetworks/README.md b/translations/mo/lessons/5-NLP/17-GenerativeNetworks/README.md index 12654c3c..f49d8bc3 100644 --- a/translations/mo/lessons/5-NLP/17-GenerativeNetworks/README.md +++ b/translations/mo/lessons/5-NLP/17-GenerativeNetworks/README.md @@ -19,7 +19,7 @@ CO_OP_TRANSLATOR_METADATA: 這使得不同的神經網絡架構成為可能，如下圖所示： -![展示常見循環神經網絡模式的圖片。](../../../../../translated_images/mo/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) +![展示常見循環神經網絡模式的圖片。](../../../../../translated_images/zh-MO/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) > 圖片來源：[Andrej Karpaty](http://karpathy.github.io/) 的博客文章 [Unreasonable Effectiveness of Recurrent Neural Networks](http://karpathy.github.io/2015/05/21/rnn-effectiveness/) @@ -32,7 +32,7 @@ CO_OP_TRANSLATOR_METADATA: 我們將訓練這個 RNN 逐步生成文本。在每一步中，我們將取一個長度為 `nchars` 的字符序列，並要求網絡為每個輸入字符生成下一個輸出字符： -![展示 RNN 生成單詞 'HELLO' 的示例圖片。](../../../../../translated_images/mo/rnn-generate.56c54afb52f9781d.webp) +![展示 RNN 生成單詞 'HELLO' 的示例圖片。](../../../../../translated_images/zh-MO/rnn-generate.56c54afb52f9781d.webp) 在生成文本（推理過程）時，我們從某個**提示**開始，將其通過 RNN 單元生成中間狀態，然後從該狀態開始生成。我們一次生成一個字符，並將狀態和生成的字符傳遞給另一個 RNN 單元以生成下一個字符，直到生成足夠的字符。 diff --git a/translations/mo/lessons/5-NLP/18-Transformers/README.md b/translations/mo/lessons/5-NLP/18-Transformers/README.md index 40351150..2fd251bc 100644 --- a/translations/mo/lessons/5-NLP/18-Transformers/README.md +++ b/translations/mo/lessons/5-NLP/18-Transformers/README.md @@ -20,13 +20,13 @@ CO_OP_TRANSLATOR_METADATA: **注意力機制**提供了一種方法，能夠對每個輸入向量對 RNN 每個輸出預測的上下文影響進行加權。其實現方式是通過在輸入 RNN 的中間狀態與輸出 RNN 之間創建捷徑。這樣，在生成輸出符號 yt 時，我們會考慮所有輸入隱藏狀態 hi，並賦予不同的權重係數 αt,i。 -![顯示具有加性注意力層的編碼器/解碼器模型的圖片](../../../../../translated_images/mo/encoder-decoder-attention.7a726296894fb567.webp) +![顯示具有加性注意力層的編碼器/解碼器模型的圖片](../../../../../translated_images/zh-MO/encoder-decoder-attention.7a726296894fb567.webp) > [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf) 中的加性注意力機制編碼器-解碼器模型，圖片來源於[這篇博客文章](https://lilianweng.github.io/lil-log/2018/06/24/attention-attention.html) 注意力矩陣 {αi,j} 表示某些輸入詞在生成輸出序列中特定詞時所起的作用程度。以下是一個這樣的矩陣示例： -![顯示 RNNsearch-50 發現的樣本對齊的圖片，取自 Bahdanau - arviz.org](../../../../../translated_images/mo/bahdanau-fig3.09ba2d37f202a6af.webp) +![顯示 RNNsearch-50 發現的樣本對齊的圖片，取自 Bahdanau - arviz.org](../../../../../translated_images/zh-MO/bahdanau-fig3.09ba2d37f202a6af.webp) > 圖片來自 [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf) (圖3) @@ -66,7 +66,7 @@ Transformer 的核心思想之一是避免 RNN 的序列性質，並創建一個 接下來，我們需要捕捉序列中的一些模式。為此，Transformer 使用了**自注意力**機制，這本質上是將注意力應用於相同的輸入和輸出序列。應用自注意力使我們能夠考慮句子中的**上下文**，並查看哪些詞是相互關聯的。例如，它可以幫助我們理解哪些詞是由代詞（如 *it*）指代的，並考慮上下文： -![](../../../../../translated_images/mo/CoreferenceResolution.861924d6d384a7d6.webp) +![](../../../../../translated_images/zh-MO/CoreferenceResolution.861924d6d384a7d6.webp) > 圖片來自 [Google 博客](https://research.googleblog.com/2017/08/transformer-novel-neural-network.html) @@ -91,7 +91,7 @@ Transformer 的核心思想之一是避免 RNN 的序列性質，並創建一個 **BERT**（Bidirectional Encoder Representations from Transformers）是一個非常大的多層 Transformer 網路，*BERT-base* 有 12 層，*BERT-large* 有 24 層。該模型首先在大規模文本語料庫（維基百科 + 書籍）上進行無監督訓練（預測句子中的遮蔽詞）。在預訓練過程中，模型吸收了大量的語言理解能力，這些能力可以通過微調其他數據集來利用。這個過程稱為**遷移學習**。 -![圖片來自 http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/mo/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp) +![圖片來自 http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/zh-MO/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp) > 圖片 [來源](http://jalammar.github.io/illustrated-bert/) diff --git a/translations/mo/lessons/5-NLP/18-Transformers/TransformersPyTorch.ipynb b/translations/mo/lessons/5-NLP/18-Transformers/TransformersPyTorch.ipynb index b8b67d2d..70281407 100644 --- a/translations/mo/lessons/5-NLP/18-Transformers/TransformersPyTorch.ipynb +++ b/translations/mo/lessons/5-NLP/18-Transformers/TransformersPyTorch.ipynb @@ -12,12 +12,12 @@ "\n", "**注意力機制**提供了一種方法，能夠對每個輸入向量對RNN每個輸出預測的上下文影響進行加權。其實現方式是通過在輸入RNN的中間狀態和輸出RNN之間創建捷徑。這樣，在生成輸出符號 $y_t$ 時，我們會考慮所有輸入隱藏狀態 $h_i$，並賦予不同的權重係數 $\\alpha_{t,i}$。\n", "\n", - "![顯示具有加性注意力層的編碼器/解碼器模型的圖片](../../../../../translated_images/mo/encoder-decoder-attention.7a726296894fb567.webp)\n", + "![顯示具有加性注意力層的編碼器/解碼器模型的圖片](../../../../../translated_images/zh-MO/encoder-decoder-attention.7a726296894fb567.webp)\n", "*來自 [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf) 的加性注意力機制編碼器-解碼器模型，圖片引用自[這篇部落格文章](https://lilianweng.github.io/lil-log/2018/06/24/attention-attention.html)*\n", "\n", "注意力矩陣 $\\{\\alpha_{i,j}\\}$ 表示某些輸入詞在生成輸出序列中特定詞時所起的作用程度。以下是這樣一個矩陣的示例：\n", "\n", - "![顯示由 RNNsearch-50 找到的樣本對齊的圖片，取自 Bahdanau - arviz.org](../../../../../translated_images/mo/bahdanau-fig3.09ba2d37f202a6af.webp)\n", + "![顯示由 RNNsearch-50 找到的樣本對齊的圖片，取自 Bahdanau - arviz.org](../../../../../translated_images/zh-MO/bahdanau-fig3.09ba2d37f202a6af.webp)\n", "\n", "*圖片取自 [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf)（圖3）*\n", "\n", @@ -35,7 +35,7 @@ "\n", "**BERT**（Bidirectional Encoder Representations from Transformers，雙向編碼器表示）是一個非常大的多層Transformer網路，*BERT-base*有12層，*BERT-large*有24層。該模型首先在大規模文本數據（維基百科+書籍）上進行無監督訓練（預測句子中的被遮蔽詞）。在預訓練過程中，模型吸收了大量的語言理解能力，這些能力可以通過微調其他數據集來加以利用。這個過程被稱為**遷移學習**。\n", "\n", - "![圖片來自 http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/mo/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp)\n", + "![圖片來自 http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/zh-MO/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp)\n", "\n", "Transformer架構有許多變體，包括BERT、DistilBERT、BigBird、OpenGPT3等，這些模型都可以進行微調。[HuggingFace套件](https://github.com/huggingface/) 提供了用PyTorch訓練這些架構的資源庫。\n", "\n", diff --git a/translations/mo/lessons/5-NLP/18-Transformers/TransformersTF.ipynb b/translations/mo/lessons/5-NLP/18-Transformers/TransformersTF.ipynb index 47060bea..9d28ad7f 100644 --- a/translations/mo/lessons/5-NLP/18-Transformers/TransformersTF.ipynb +++ b/translations/mo/lessons/5-NLP/18-Transformers/TransformersTF.ipynb @@ -12,12 +12,12 @@ "\n", "**注意力機制**提供了一種方法，能夠對每個輸入向量在RNN的每個輸出預測中的上下文影響進行加權。其實現方式是通過在輸入RNN的中間狀態和輸出RNN之間創建捷徑。在生成輸出符號$y_t$時，我們會考慮所有輸入隱藏狀態$h_i$，並使用不同的權重係數$\\alpha_{t,i}$。\n", "\n", - "![顯示具有加性注意力層的編碼器/解碼器模型的圖片](../../../../../translated_images/mo/encoder-decoder-attention.7a726296894fb567.webp)\n", + "![顯示具有加性注意力層的編碼器/解碼器模型的圖片](../../../../../translated_images/zh-MO/encoder-decoder-attention.7a726296894fb567.webp)\n", "*[Bahdanau等人，2015](https://arxiv.org/pdf/1409.0473.pdf)中的加性注意力機制編碼器-解碼器模型，圖片引用自[這篇博客文章](https://lilianweng.github.io/lil-log/2018/06/24/attention-attention.html)*\n", "\n", "注意力矩陣$\\{\\alpha_{i,j}\\}$表示某些輸入詞語在生成輸出序列中的某個詞語時所起的作用程度。以下是這樣一個矩陣的示例：\n", "\n", - "![顯示RNNsearch-50找到的示例對齊的圖片，取自Bahdanau - arviz.org](../../../../../translated_images/mo/bahdanau-fig3.09ba2d37f202a6af.webp)\n", + "![顯示RNNsearch-50找到的示例對齊的圖片，取自Bahdanau - arviz.org](../../../../../translated_images/zh-MO/bahdanau-fig3.09ba2d37f202a6af.webp)\n", "\n", "*圖片取自[Bahdanau等人，2015](https://arxiv.org/pdf/1409.0473.pdf)（圖3）*\n", "\n", @@ -225,7 +225,7 @@ "\n", "**BERT**（雙向編碼器表示來自 Transformer）是一個非常大型的多層 Transformer 網絡，*BERT-base* 有 12 層，*BERT-large* 則有 24 層。該模型首先在大量文本數據（維基百科 + 書籍）上進行無監督訓練（預測句子中的被遮蔽詞）。在預訓練過程中，模型吸收了大量的語言理解能力，之後可以通過微調與其他數據集結合使用。這個過程被稱為 **遷移學習**。\n", "\n", - "![圖片來源：http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/mo/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp)\n", + "![圖片來源：http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/zh-MO/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp)\n", "\n", "Transformer 架構有許多變體，包括 BERT、DistilBERT、BigBird、OpenGPT3 等，它們都可以進行微調。\n", "\n", diff --git a/translations/mo/lessons/5-NLP/19-NER/README.md b/translations/mo/lessons/5-NLP/19-NER/README.md index 2d98a652..9df2b96f 100644 --- a/translations/mo/lessons/5-NLP/19-NER/README.md +++ b/translations/mo/lessons/5-NLP/19-NER/README.md @@ -57,7 +57,7 @@ NER 模型本質上是 **標記分類模型**，因為對於每個輸入標記 由於我們需要在標記和類別之間建立一對一的對應關係，我們可以從這張圖中訓練一個右側的 **多對多** 神經網絡模型： -![顯示常見循環神經網絡模式的圖片。](../../../../../translated_images/mo/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) +![顯示常見循環神經網絡模式的圖片。](../../../../../translated_images/zh-MO/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) > *圖片來自 [這篇部落格文章](http://karpathy.github.io/2015/05/21/rnn-effectiveness/) 作者 [Andrej Karpathy](http://karpathy.github.io/)。NER 標記分類模型對應於此圖片中的最右側網絡架構。* diff --git a/translations/mo/lessons/5-NLP/README.md b/translations/mo/lessons/5-NLP/README.md index 5d1e15ed..82dbb09b 100644 --- a/translations/mo/lessons/5-NLP/README.md +++ b/translations/mo/lessons/5-NLP/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 自然語言處理 -![NLP 任務的手繪摘要](../../../../translated_images/mo/ai-nlp.b22dcb8ca4707cea.webp) +![NLP 任務的手繪摘要](../../../../translated_images/zh-MO/ai-nlp.b22dcb8ca4707cea.webp) 在本節中，我們將專注於使用神經網絡來處理與**自然語言處理 (NLP)** 相關的任務。我們希望計算機能夠解決許多 NLP 問題： diff --git a/translations/mo/lessons/6-Other/23-MultiagentSystems/README.md b/translations/mo/lessons/6-Other/23-MultiagentSystems/README.md index d28528bb..c7a9118d 100644 --- a/translations/mo/lessons/6-Other/23-MultiagentSystems/README.md +++ b/translations/mo/lessons/6-Other/23-MultiagentSystems/README.md @@ -70,7 +70,7 @@ NetLogo的一大優勢是它包含一個可供試用的工作模型庫。進入* 打開模型後，你會進入NetLogo的主界面。以下是一個描述狼和羊在有限資源（草地）條件下的種群模型示例。 -![NetLogo主界面](../../../../../translated_images/mo/NetLogo-Main.32653711ec1a01b3.webp) +![NetLogo主界面](../../../../../translated_images/zh-MO/NetLogo-Main.32653711ec1a01b3.webp) > Dmitry Soshnikov提供的截圖 diff --git a/translations/mo/lessons/README.md b/translations/mo/lessons/README.md index 6a641a9d..2921c30b 100644 --- a/translations/mo/lessons/README.md +++ b/translations/mo/lessons/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 概述 -![概述的手繪圖](../../../translated_images/mo/ai-overview.0857791951d19500.webp) +![概述的手繪圖](../../../translated_images/zh-MO/ai-overview.0857791951d19500.webp) > 手繪筆記由 [Tomomi Imura](https://twitter.com/girlie_mac) 提供 diff --git a/translations/mo/lessons/X-Extras/X1-MultiModal/README.md b/translations/mo/lessons/X-Extras/X1-MultiModal/README.md index eb704a04..243cf65a 100644 --- a/translations/mo/lessons/X-Extras/X1-MultiModal/README.md +++ b/translations/mo/lessons/X-Extras/X1-MultiModal/README.md @@ -15,7 +15,7 @@ CO_OP_TRANSLATOR_METADATA: CLIP 的核心思想是能夠比較文本提示與圖像，並判斷圖像與提示的匹配程度。 -![CLIP 架構](../../../../../translated_images/mo/clip-arch.b3dbf20b4e8ed8be.webp) +![CLIP 架構](../../../../../translated_images/zh-MO/clip-arch.b3dbf20b4e8ed8be.webp) > *圖片來源於[這篇博客](https://openai.com/blog/clip/)* @@ -29,7 +29,7 @@ CLIP 模型/庫可以從 [OpenAI GitHub](https://github.com/openai/CLIP) 獲取 假設我們需要在貓、狗和人之間對圖像進行分類。在這種情況下，我們可以將圖像和一系列文本提示輸入模型，例如：“*一張貓的照片*”、“*一張狗的照片*”、“*一張人的照片*”。在結果的 3 個概率向量中，我們只需選擇值最大的索引。 -![CLIP 用於圖像分類](../../../../../translated_images/mo/clip-class.3af42ef0b2b19369.webp) +![CLIP 用於圖像分類](../../../../../translated_images/zh-MO/clip-class.3af42ef0b2b19369.webp) > *圖片來源於[這篇博客](https://openai.com/blog/clip/)* @@ -53,13 +53,13 @@ VQGAN 與普通 [GAN](../../4-ComputerVision/10-GANs/README.md) 的主要區別 VQGAN 與傳統 GAN 的一個重要區別是，後者可以從任何輸入向量生成一個體面的圖像，而 VQGAN 可能生成不連貫的圖像。因此，我們需要進一步引導圖像創建過程，這可以通過 CLIP 完成。 -![VQGAN+CLIP 架構](../../../../../translated_images/mo/vqgan.5027fe05051dfa31.webp) +![VQGAN+CLIP 架構](../../../../../translated_images/zh-MO/vqgan.5027fe05051dfa31.webp) 為了生成與文本提示相符的圖像，我們從一些隨機編碼向量開始，將其通過 VQGAN 生成圖像。然後使用 CLIP 生成一個損失函數，該函數顯示圖像與文本提示的匹配程度。接下來的目標是最小化這個損失，通過反向傳播調整輸入向量參數。 一個實現 VQGAN+CLIP 的優秀庫是 [Pixray](http://github.com/pixray/pixray)。 -![Pixray 生成的圖片](../../../../../translated_images/mo/a_closeup_watercolor_portrait_of_young_male_teacher_of_literature_with_a_book.2384968e9db8a0d0.webp) | ![Pixray 生成的圖片](../../../../../translated_images/mo/a_closeup_oil_portrait_of_young_female_teacher_of_computer_science_with_a_computer.e0b6495f210a4390.webp) | ![Pixray 生成的圖片](../../../../../translated_images/mo/a_closeup_oil_portrait_of_old_male_teacher_of_math.5362e67aa7fc2683.webp) +![Pixray 生成的圖片](../../../../../translated_images/zh-MO/a_closeup_watercolor_portrait_of_young_male_teacher_of_literature_with_a_book.2384968e9db8a0d0.webp) | ![Pixray 生成的圖片](../../../../../translated_images/zh-MO/a_closeup_oil_portrait_of_young_female_teacher_of_computer_science_with_a_computer.e0b6495f210a4390.webp) | ![Pixray 生成的圖片](../../../../../translated_images/zh-MO/a_closeup_oil_portrait_of_old_male_teacher_of_math.5362e67aa7fc2683.webp) ----|----|---- 根據提示 *一幅年輕男性文學教師手持書本的水彩特寫肖像* 生成的圖片 | 根據提示 *一幅年輕女性計算機科學教師手持電腦的油畫特寫肖像* 生成的圖片 | 根據提示 *一幅年長男性數學教師站在黑板前的油畫特寫肖像* 生成的圖片 @@ -75,7 +75,7 @@ DALL-E 是一個基於 GPT-3 的模型，專門用於從文本提示生成圖像 DALL-E 1 和 DALL-E 2 的主要區別在於，後者能生成更真實的圖像和藝術作品。 以下是使用 DALL-E 生成圖像的示例： -![DALL-E 生成的圖片](../../../../../translated_images/mo/DALL·E%202023-06-20%2015.56.56%20-%20a%20closeup%20watercolor%20portrait%20of%20young%20male%20teacher%20of%20literature%20with%20a%20book.6c235e8271d9ed10ce985d86aeb241a58518958647973af136912116b9518fce.png) | ![DALL-E 生成的圖片](../../../../../translated_images/mo/DALL·E%202023-06-20%2015.57.43%20-%20a%20closeup%20oil%20portrait%20of%20young%20female%20teacher%20of%20computer%20science%20with%20a%20computer.f21dc4166340b6c8b4d1cb57efd1e22127407f9b28c9ac7afe11344065369e64.png) | ![DALL-E 生成的圖片](../../../../../translated_images/mo/DALL·E%202023-06-20%2015.58.42%20-%20%20a%20closeup%20oil%20portrait%20of%20old%20male%20teacher%20of%20mathematics%20in%20front%20of%20blackboard.d331c2dfbdc3f7c46aa65c0809066f5e7ed4b49609cd259852e760df21051e4a.png) +![DALL-E 生成的圖片](../../../../../translated_images/zh-MO/DALL·E%202023-06-20%2015.56.56%20-%20a%20closeup%20watercolor%20portrait%20of%20young%20male%20teacher%20of%20literature%20with%20a%20book.6c235e8271d9ed10ce985d86aeb241a58518958647973af136912116b9518fce.png) | ![DALL-E 生成的圖片](../../../../../translated_images/zh-MO/DALL·E%202023-06-20%2015.57.43%20-%20a%20closeup%20oil%20portrait%20of%20young%20female%20teacher%20of%20computer%20science%20with%20a%20computer.f21dc4166340b6c8b4d1cb57efd1e22127407f9b28c9ac7afe11344065369e64.png) | ![DALL-E 生成的圖片](../../../../../translated_images/zh-MO/DALL·E%202023-06-20%2015.58.42%20-%20%20a%20closeup%20oil%20portrait%20of%20old%20male%20teacher%20of%20mathematics%20in%20front%20of%20blackboard.d331c2dfbdc3f7c46aa65c0809066f5e7ed4b49609cd259852e760df21051e4a.png) ----|----|---- 根據提示 *一幅年輕男性文學教師手持書本的水彩特寫肖像* 生成的圖片 | 根據提示 *一幅年輕女性計算機科學教師手持電腦的油畫特寫肖像* 生成的圖片 | 根據提示 *一幅年長男性數學教師站在黑板前的油畫特寫肖像* 生成的圖片 diff --git a/translations/pt/README.md b/translations/pt/README.md index 6d83275b..c3d8c67e 100644 --- a/translations/pt/README.md +++ b/translations/pt/README.md @@ -23,7 +23,7 @@ CO_OP_TRANSLATOR_METADATA: # Inteligência Artificial para Iniciantes - Um Currículo -|![Sketchnote by @girlie_mac https://twitter.com/girlie_mac](../../../../translated_images/pt/ai-overview.0857791951d19500.webp)| +|![Sketchnote by @girlie_mac https://twitter.com/girlie_mac](../../../../translated_images/pt-PT/ai-overview.0857791951d19500.webp)| |:---:| | IA Para Iniciantes - _Sketchnote por [@girlie_mac](https://twitter.com/girlie_mac)_ | diff --git a/translations/pt/lessons/1-Intro/README.md b/translations/pt/lessons/1-Intro/README.md index add14a01..3e17c840 100644 --- a/translations/pt/lessons/1-Intro/README.md +++ b/translations/pt/lessons/1-Intro/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # Introdução à IA -![Resumo do conteúdo de Introdução à IA em um desenho](../../../../translated_images/pt/ai-intro.bf28d1ac4235881c.webp) +![Resumo do conteúdo de Introdução à IA em um desenho](../../../../translated_images/pt-PT/ai-intro.bf28d1ac4235881c.webp) > Sketchnote por [Tomomi Imura](https://twitter.com/girlie_mac) @@ -19,7 +19,7 @@ CO_OP_TRANSLATOR_METADATA: Originalmente, os computadores foram inventados por [Charles Babbage](https://en.wikipedia.org/wiki/Charles_Babbage) para operar com números seguindo um procedimento bem definido - um algoritmo. Os computadores modernos, embora significativamente mais avançados do que o modelo original proposto no século XIX, ainda seguem a mesma ideia de cálculos controlados. Assim, é possível programar um computador para fazer algo se soubermos a sequência exata de passos necessários para alcançar o objetivo. -![Foto de uma pessoa](../../../../translated_images/pt/dsh_age.d212a30d4e54fb5f.webp) +![Foto de uma pessoa](../../../../translated_images/pt-PT/dsh_age.d212a30d4e54fb5f.webp) > Foto por [Vickie Soshnikova](http://twitter.com/vickievalerie) @@ -46,7 +46,7 @@ Para mais informações, consulte **[Inteligência Artificial Geral](https://en. Um dos problemas ao lidar com o termo **[Inteligência](https://en.wikipedia.org/wiki/Intelligence)** é que não há uma definição clara para este termo. Pode-se argumentar que inteligência está conectada ao **pensamento abstrato** ou à **autoconsciência**, mas não conseguimos defini-la adequadamente. -![Foto de um gato](../../../../translated_images/pt/photo-cat.8c8e8fb760ffe457.webp) +![Foto de um gato](../../../../translated_images/pt-PT/photo-cat.8c8e8fb760ffe457.webp) > [Foto](https://unsplash.com/photos/75715CVEJhI) por [Amber Kipp](https://unsplash.com/@sadmax) do Unsplash @@ -98,13 +98,13 @@ Alternativamente, podemos tentar modelar os elementos mais simples dentro do nos > | E o ML? | | > |--------------|-----------| -> | Parte da Inteligência Artificial que se baseia no computador aprendendo a resolver um problema com base em alguns dados é chamada de **Machine Learning**. Não consideraremos o aprendizado de máquina clássico neste curso - recomendamos o currículo separado [Machine Learning para Iniciantes](http://aka.ms/ml-beginners). | ![ML para Iniciantes](../../../../translated_images/pt/ml-for-beginners.9e4fed176fd5817d.webp) | +> | Parte da Inteligência Artificial que se baseia no computador aprendendo a resolver um problema com base em alguns dados é chamada de **Machine Learning**. Não consideraremos o aprendizado de máquina clássico neste curso - recomendamos o currículo separado [Machine Learning para Iniciantes](http://aka.ms/ml-beginners). | ![ML para Iniciantes](../../../../translated_images/pt-PT/ml-for-beginners.9e4fed176fd5817d.webp) | ## Um Breve Histórico da IA A Inteligência Artificial começou como um campo no meio do século XX. Inicialmente, o raciocínio simbólico era a abordagem predominante, e isso levou a uma série de sucessos importantes, como sistemas especialistas – programas de computador que eram capazes de agir como especialistas em alguns domínios de problemas limitados. No entanto, logo ficou claro que essa abordagem não escala bem. Extrair o conhecimento de um especialista, representá-lo em um computador e manter essa base de conhecimento precisa acaba sendo uma tarefa muito complexa e cara demais para ser prática em muitos casos. Isso levou ao chamado [Inverno da IA](https://en.wikipedia.org/wiki/AI_winter) na década de 1970. -Breve Histórico da IA +Breve Histórico da IA > Imagem por [Dmitry Soshnikov](http://soshnikov.com) @@ -124,7 +124,7 @@ Da mesma forma, podemos ver como a abordagem para criar “programas falantes” * Assistentes modernos, como Cortana, Siri ou Google Assistant, são todos sistemas híbridos que usam redes neurais para converter fala em texto e reconhecer nossa intenção, e depois empregam algum raciocínio ou algoritmos explícitos para realizar as ações necessárias. * No futuro, podemos esperar um modelo completamente baseado em redes neurais para lidar com diálogos por conta própria. As recentes redes neurais da família GPT e [Turing-NLG](https://www.microsoft.com/research/blog/turing-nlg-a-17-billion-parameter-language-model-by-microsoft) mostram grande sucesso nisso. -a evolução do Teste de Turing +a evolução do Teste de Turing > Imagem de Dmitry Soshnikov, [foto](https://unsplash.com/photos/r8LmVbUKgns) de [Marina Abrosimova](https://unsplash.com/@abrosimova_marina_foto), Unsplash ## Investigação Recente em IA diff --git a/translations/pt/lessons/2-Symbolic/Animals.ipynb b/translations/pt/lessons/2-Symbolic/Animals.ipynb index 50f4014c..d4e98853 100644 --- a/translations/pt/lessons/2-Symbolic/Animals.ipynb +++ b/translations/pt/lessons/2-Symbolic/Animals.ipynb @@ -12,7 +12,7 @@ "\n", "Neste exemplo, vamos implementar um sistema baseado em conhecimento simples para determinar um animal com base em algumas características físicas. O sistema pode ser representado pela seguinte árvore AND-OR (esta é uma parte da árvore completa, podemos facilmente adicionar mais regras):\n", "\n", - "![](../../../../../../translated_images/pt/AND-OR-Tree.5592d2c70187f283.webp)\n" + "![](../../../../../../translated_images/pt-PT/AND-OR-Tree.5592d2c70187f283.webp)\n" ] }, { diff --git a/translations/pt/lessons/2-Symbolic/README.md b/translations/pt/lessons/2-Symbolic/README.md index 4b4b6e6e..048ef694 100644 --- a/translations/pt/lessons/2-Symbolic/README.md +++ b/translations/pt/lessons/2-Symbolic/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # Representação do Conhecimento e Sistemas Especialistas -![Resumo do conteúdo de IA Simbólica](../../../../../../translated_images/pt/ai-symbolic.715a30cb610411a6.webp) +![Resumo do conteúdo de IA Simbólica](../../../../../../translated_images/pt-PT/ai-symbolic.715a30cb610411a6.webp) > Sketchnote por [Tomomi Imura](https://twitter.com/girlie_mac) @@ -35,13 +35,13 @@ Na maior parte das vezes, não definimos conhecimento de forma rigorosa, mas ali * **Conhecimento** é a informação integrada no nosso modelo do mundo. Por exemplo, uma vez que aprendemos o que é um computador, começamos a ter algumas ideias sobre como funciona, quanto custa e para que pode ser usado. Esta rede de conceitos inter-relacionados forma o nosso conhecimento. * **Sabedoria** é mais um nível da nossa compreensão do mundo, representando *meta-conhecimento*, por exemplo, alguma noção de como e quando o conhecimento deve ser utilizado. - + *Imagem [da Wikipedia](https://commons.wikimedia.org/w/index.php?curid=37705247), por Longlivetheux - Trabalho próprio, CC BY-SA 4.0* Assim, o problema da **representação do conhecimento** é encontrar uma forma eficaz de representar conhecimento dentro de um computador sob a forma de dados, para que possa ser utilizado automaticamente. Isto pode ser visto como um espectro: -![Espectro da representação do conhecimento](../../../../../../translated_images/pt/knowledge-spectrum.b60df631852c0217.webp) +![Espectro da representação do conhecimento](../../../../../../translated_images/pt-PT/knowledge-spectrum.b60df631852c0217.webp) > Imagem por [Dmitry Soshnikov](http://soshnikov.com) @@ -94,7 +94,7 @@ Sintaxe de Bloco | Indentação | | | Um dos primeiros sucessos da IA simbólica foram os chamados **sistemas especialistas** – sistemas computacionais desenhados para agir como um especialista num domínio de problema limitado. Eram baseados numa **base de conhecimento** extraída de um ou mais especialistas humanos, e continham um **mecanismo de inferência** que realizava algum raciocínio sobre ela. -![Arquitetura Humana](../../../../../../translated_images/pt/arch-human.5d4d35f1bba3ab1c.webp) | ![Sistema Baseado em Conhecimento](../../../../../../translated_images/pt/arch-kbs.3ec5c150b09fa8da.webp) +![Arquitetura Humana](../../../../../../translated_images/pt-PT/arch-human.5d4d35f1bba3ab1c.webp) | ![Sistema Baseado em Conhecimento](../../../../../../translated_images/pt-PT/arch-kbs.3ec5c150b09fa8da.webp) ---------------------------------------------|----------------------------------------------- Estrutura simplificada do sistema neural humano | Arquitetura de um sistema baseado em conhecimento @@ -106,7 +106,7 @@ Os sistemas especialistas são construídos como o sistema de raciocínio humano Como exemplo, vamos considerar o seguinte sistema especialista para determinar um animal baseado nas suas características físicas: -![Árvore AND-OR](../../../../../../translated_images/pt/AND-OR-Tree.5592d2c70187f283.webp) +![Árvore AND-OR](../../../../../../translated_images/pt-PT/AND-OR-Tree.5592d2c70187f283.webp) > Imagem por [Dmitry Soshnikov](http://soshnikov.com) @@ -175,7 +175,7 @@ Um conceito central na Web Semântica é o conceito de **Ontologia**. Refere-se Na web semântica, todas as representações são baseadas em triplets. Cada objeto e cada relação são identificados unicamente pelo URI. Por exemplo, se quisermos declarar o facto de que este Currículo de IA foi desenvolvido por Dmitry Soshnikov em 1 de Janeiro de 2022 - aqui estão os triplets que podemos usar: - + ``` http://github.com/microsoft/ai-for-beginners http://www.example.com/terms/creation-date “Jan 1, 2022” @@ -186,7 +186,7 @@ http://github.com/microsoft/ai-for-beginners http://purl.org/dc/elements/1.1/cre Num caso mais complexo, se quisermos definir uma lista de criadores, podemos usar algumas estruturas de dados definidas em RDF. - + > Diagramas acima por [Dmitry Soshnikov](http://soshnikov.com) @@ -210,7 +210,7 @@ GROUP BY ?eyeColorLabel > ✅ Se quiser experimentar criar as suas próprias ontologias, ou abrir ontologias existentes, existe um excelente editor visual de ontologias chamado [Protégé](https://protege.stanford.edu/). Faça o download ou use online. - + *Editor Web Protégé aberto com a ontologia da Família Romanov. Captura de ecrã por Dmitry Soshnikov* diff --git a/translations/pt/lessons/3-NeuralNetworks/05-Frameworks/README.md b/translations/pt/lessons/3-NeuralNetworks/05-Frameworks/README.md index 13fb393f..0a07f3d2 100644 --- a/translations/pt/lessons/3-NeuralNetworks/05-Frameworks/README.md +++ b/translations/pt/lessons/3-NeuralNetworks/05-Frameworks/README.md @@ -58,7 +58,7 @@ Overfitting é um conceito extremamente importante em aprendizagem automática, Considere o seguinte problema de aproximar 5 pontos (representados por `x` nos gráficos abaixo): -![linear](../../../../../translated_images/pt/overfit1.f24b71c6f652e59e.webp) | ![overfit](../../../../../translated_images/pt/overfit2.131f5800ae10ca5e.webp) +![linear](../../../../../translated_images/pt-PT/overfit1.f24b71c6f652e59e.webp) | ![overfit](../../../../../translated_images/pt-PT/overfit2.131f5800ae10ca5e.webp) -------------------------|-------------------------- **Modelo linear, 2 parâmetros** | **Modelo não-linear, 7 parâmetros** Erro de treino = 5.3 | Erro de treino = 0 @@ -79,7 +79,7 @@ Erro de validação = 5.1 | Erro de validação = 20 Como pode ver no gráfico acima, o overfitting pode ser detetado por um erro de treino muito baixo e um erro de validação elevado. Normalmente, durante o treino, vemos tanto os erros de treino quanto de validação começarem a diminuir, e então, em algum momento, o erro de validação pode parar de diminuir e começar a aumentar. Este será um sinal de overfitting e um indicador de que provavelmente devemos parar o treino nesse ponto (ou pelo menos fazer um snapshot do modelo). -![overfitting](../../../../../translated_images/pt/Overfitting.408ad91cd90b4371.webp) +![overfitting](../../../../../translated_images/pt-PT/Overfitting.408ad91cd90b4371.webp) ## Como prevenir o overfitting diff --git a/translations/pt/lessons/3-NeuralNetworks/README.md b/translations/pt/lessons/3-NeuralNetworks/README.md index decf2d7b..47919dd0 100644 --- a/translations/pt/lessons/3-NeuralNetworks/README.md +++ b/translations/pt/lessons/3-NeuralNetworks/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # Introdução às Redes Neuronais -![Resumo do conteúdo de Introdução às Redes Neuronais em um desenho](../../../../translated_images/pt/ai-neuralnetworks.1c687ae40bc86e83.webp) +![Resumo do conteúdo de Introdução às Redes Neuronais em um desenho](../../../../translated_images/pt-PT/ai-neuralnetworks.1c687ae40bc86e83.webp) Como discutimos na introdução, uma das formas de alcançar inteligência é treinar um **modelo computacional** ou um **cérebro artificial**. Desde meados do século XX, os investigadores experimentaram diferentes modelos matemáticos, até que, nos últimos anos, esta abordagem provou ser extremamente bem-sucedida. Esses modelos matemáticos do cérebro são chamados de **redes neuronais**. @@ -36,13 +36,13 @@ Neste currículo, focar-nos-emos apenas em modelos de redes neuronais. Na biologia, sabemos que o nosso cérebro é composto por células neuronais (neurónios), cada uma delas com múltiplas "entradas" (dendritos) e uma única "saída" (axónio). Tanto os dendritos como os axónios podem conduzir sinais elétricos, e as conexões entre eles — conhecidas como sinapses — podem apresentar diferentes graus de condutividade, que são regulados por neurotransmissores. -![Modelo de um Neurónio](../../../../translated_images/pt/synapse-wikipedia.ed20a9e4726ea1c6.webp) | ![Modelo de um Neurónio](../../../../translated_images/pt/artneuron.1a5daa88d20ebe6f.webp) +![Modelo de um Neurónio](../../../../translated_images/pt-PT/synapse-wikipedia.ed20a9e4726ea1c6.webp) | ![Modelo de um Neurónio](../../../../translated_images/pt-PT/artneuron.1a5daa88d20ebe6f.webp) ----|---- Neurónio Real *([Imagem](https://en.wikipedia.org/wiki/Synapse#/media/File:SynapseSchematic_lines.svg) da Wikipédia)* | Neurónio Artificial *(Imagem do Autor)* Assim, o modelo matemático mais simples de um neurónio contém várias entradas X1, ..., XN e uma saída Y, e uma série de pesos W1, ..., WN. A saída é calculada como: -Y = f\left(\sum_{i=1}^N X_iW_i\right) +Y = f\left(\sum_{i=1}^N X_iW_i\right) onde **f** é uma **função de ativação** não linear. diff --git a/translations/pt/lessons/4-ComputerVision/06-IntroCV/README.md b/translations/pt/lessons/4-ComputerVision/06-IntroCV/README.md index 7494536f..64b9ee4a 100644 --- a/translations/pt/lessons/4-ComputerVision/06-IntroCV/README.md +++ b/translations/pt/lessons/4-ComputerVision/06-IntroCV/README.md @@ -73,14 +73,14 @@ No nosso [OpenCV Notebook](OpenCV.ipynb), damos alguns exemplos de quando a vis * **Pré-processamento de uma fotografia de um livro em Braille**. Focamos em como podemos usar thresholding, deteção de características, transformação de perspetiva e manipulações NumPy para separar símbolos individuais em Braille para posterior classificação por uma rede neuronal. -![Imagem Braille](../../../../../translated_images/pt/braille.341962ff76b1bd70.webp) | ![Imagem Braille Pré-processada](../../../../../translated_images/pt/braille-result.46530fea020b03c7.webp) | ![Símbolos Braille](../../../../../translated_images/pt/braille-symbols.0159185ab69d5339.webp) +![Imagem Braille](../../../../../translated_images/pt-PT/braille.341962ff76b1bd70.webp) | ![Imagem Braille Pré-processada](../../../../../translated_images/pt-PT/braille-result.46530fea020b03c7.webp) | ![Símbolos Braille](../../../../../translated_images/pt-PT/braille-symbols.0159185ab69d5339.webp) ----|-----|----- > Imagem de [OpenCV.ipynb](OpenCV.ipynb) * **Deteção de movimento em vídeo usando diferença de frames**. Se a câmara estiver fixa, os frames do feed da câmara devem ser bastante semelhantes entre si. Como os frames são representados como arrays, apenas subtraindo esses arrays de dois frames subsequentes obteremos a diferença de pixels, que deve ser baixa para frames estáticos e tornar-se maior quando houver movimento substancial na imagem. -![Imagem de frames de vídeo e diferenças de frames](../../../../../translated_images/pt/frame-difference.706f805491a0883c.webp) +![Imagem de frames de vídeo e diferenças de frames](../../../../../translated_images/pt-PT/frame-difference.706f805491a0883c.webp) > Imagem de [OpenCV.ipynb](OpenCV.ipynb) @@ -89,7 +89,7 @@ No nosso [OpenCV Notebook](OpenCV.ipynb), damos alguns exemplos de quando a vis - **Fluxo Ótico Denso** calcula o campo vetorial que mostra para cada pixel onde ele está a mover-se. - **Fluxo Ótico Esparso** baseia-se em tomar algumas características distintivas na imagem (por exemplo, bordas) e construir a sua trajetória de frame para frame. -![Imagem de Fluxo Ótico](../../../../../translated_images/pt/optical.1f4a94464579a83a.webp) +![Imagem de Fluxo Ótico](../../../../../translated_images/pt-PT/optical.1f4a94464579a83a.webp) > Imagem de [OpenCV.ipynb](OpenCV.ipynb) diff --git a/translations/pt/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md b/translations/pt/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md index 65eeee96..3dab957f 100644 --- a/translations/pt/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md +++ b/translations/pt/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md @@ -13,11 +13,11 @@ CO_OP_TRANSLATOR_METADATA: VGG-16 é uma rede que alcançou 92,7% de precisão na classificação top-5 do ImageNet em 2014. Tem a seguinte estrutura de camadas: -![Camadas do ImageNet](../../../../../translated_images/pt/vgg-16-arch1.d901a5583b3a51ba.webp) +![Camadas do ImageNet](../../../../../translated_images/pt-PT/vgg-16-arch1.d901a5583b3a51ba.webp) Como pode ver, a VGG segue uma arquitetura tradicional em pirâmide, que consiste numa sequência de camadas de convolução e pooling. -![Pirâmide do ImageNet](../../../../../translated_images/pt/vgg-16-arch.64ff2137f50dd49f.webp) +![Pirâmide do ImageNet](../../../../../translated_images/pt-PT/vgg-16-arch.64ff2137f50dd49f.webp) > Imagem de [Researchgate](https://www.researchgate.net/figure/Vgg16-model-structure-To-get-the-VGG-NIN-model-we-replace-the-2-nd-4-th-6-th-7-th_fig2_335194493) diff --git a/translations/pt/lessons/4-ComputerVision/07-ConvNets/README.md b/translations/pt/lessons/4-ComputerVision/07-ConvNets/README.md index 31d90669..20214a65 100644 --- a/translations/pt/lessons/4-ComputerVision/07-ConvNets/README.md +++ b/translations/pt/lessons/4-ComputerVision/07-ConvNets/README.md @@ -17,7 +17,7 @@ Na vida real, queremos ser capazes de reconhecer objetos numa imagem independent Para extrair padrões, utilizaremos o conceito de **filtros convolucionais**. Como sabe, uma imagem é representada por uma matriz 2D ou um tensor 3D com profundidade de cor. Aplicar um filtro significa que utilizamos uma matriz relativamente pequena chamada **kernel do filtro**, e para cada pixel na imagem original calculamos a média ponderada com os pontos vizinhos. Podemos imaginar isto como uma pequena janela que desliza sobre toda a imagem, e que calcula a média de todos os pixels de acordo com os pesos na matriz kernel do filtro. -![Filtro de Borda Vertical](../../../../../translated_images/pt/filter-vert.b7148390ca0bc356.webp) | ![Filtro de Borda Horizontal](../../../../../translated_images/pt/filter-horiz.59b80ed4feb946ef.webp) +![Filtro de Borda Vertical](../../../../../translated_images/pt-PT/filter-vert.b7148390ca0bc356.webp) | ![Filtro de Borda Horizontal](../../../../../translated_images/pt-PT/filter-horiz.59b80ed4feb946ef.webp) ----|---- > Imagem por Dmitry Soshnikov @@ -38,7 +38,7 @@ O funcionamento das CNN baseia-se nas seguintes ideias importantes: * Podemos projetar a rede de forma a que os filtros sejam treinados automaticamente * Podemos usar a mesma abordagem para encontrar padrões em características de alto nível, não apenas na imagem original. Assim, a extração de características pelas CNN funciona numa hierarquia de características, começando por combinações de pixels de baixo nível até combinações de partes da imagem de nível mais alto. -![Extração Hierárquica de Características](../../../../../translated_images/pt/FeatureExtractionCNN.d9b456cbdae7cb64.webp) +![Extração Hierárquica de Características](../../../../../translated_images/pt-PT/FeatureExtractionCNN.d9b456cbdae7cb64.webp) > Imagem de [um artigo de Hislop-Lynch](https://www.semanticscholar.org/paper/Computer-vision-based-pedestrian-trajectory-Hislop-Lynch/26e6f74853fc9bbb7487b06dc2cf095d36c9021d), baseado na [sua pesquisa](https://dl.acm.org/doi/abs/10.1145/1553374.1553453) @@ -55,9 +55,9 @@ A maioria das CNN utilizadas para processamento de imagens segue uma arquitetura Como exemplo, vejamos a arquitetura da VGG-16, uma rede que alcançou 92,7% de precisão na classificação top-5 do ImageNet em 2014: -![Camadas do ImageNet](../../../../../translated_images/pt/vgg-16-arch1.d901a5583b3a51ba.webp) +![Camadas do ImageNet](../../../../../translated_images/pt-PT/vgg-16-arch1.d901a5583b3a51ba.webp) -![Pirâmide do ImageNet](../../../../../translated_images/pt/vgg-16-arch.64ff2137f50dd49f.webp) +![Pirâmide do ImageNet](../../../../../translated_images/pt-PT/vgg-16-arch.64ff2137f50dd49f.webp) > Imagem de [Researchgate](https://www.researchgate.net/figure/Vgg16-model-structure-To-get-the-VGG-NIN-model-we-replace-the-2-nd-4-th-6-th-7-th_fig2_335194493) diff --git a/translations/pt/lessons/4-ComputerVision/07-ConvNets/lab/README.md b/translations/pt/lessons/4-ComputerVision/07-ConvNets/lab/README.md index ad50dceb..8ea1f165 100644 --- a/translations/pt/lessons/4-ComputerVision/07-ConvNets/lab/README.md +++ b/translations/pt/lessons/4-ComputerVision/07-ConvNets/lab/README.md @@ -21,7 +21,7 @@ Precisa treinar uma rede neuronal convolucional para classificar diferentes raç Vamos utilizar o [Oxford-IIIT Pet Dataset](https://www.robots.ox.ac.uk/~vgg/data/pets/), que contém imagens de 37 raças diferentes de cães e gatos. -![Dataset com que iremos trabalhar](../../../../../../translated_images/pt/data.50b2a9d5484bdbf0.webp) +![Dataset com que iremos trabalhar](../../../../../../translated_images/pt-PT/data.50b2a9d5484bdbf0.webp) Para descarregar o dataset, utilize este trecho de código: diff --git a/translations/pt/lessons/4-ComputerVision/08-TransferLearning/README.md b/translations/pt/lessons/4-ComputerVision/08-TransferLearning/README.md index d4ba53cc..791d0d3a 100644 --- a/translations/pt/lessons/4-ComputerVision/08-TransferLearning/README.md +++ b/translations/pt/lessons/4-ComputerVision/08-TransferLearning/README.md @@ -29,7 +29,7 @@ Tanto o Keras como o PyTorch possuem funções para carregar facilmente pesos de Aqui estão características extraídas de uma imagem de um gato pela rede VGG-16: -![Características extraídas pela VGG-16](../../../../../translated_images/pt/features.6291f9c7ba3a0b95.webp) +![Características extraídas pela VGG-16](../../../../../translated_images/pt-PT/features.6291f9c7ba3a0b95.webp) ## Conjunto de Dados de Gatos vs. Cães @@ -48,19 +48,19 @@ Uma rede neural pré-treinada contém diferentes padrões no seu *cérebro*, inc Uma abordagem que podemos adotar é começar com uma imagem aleatória e tentar usar a técnica de **otimização por descida de gradiente** para ajustar essa imagem de forma que a rede comece a pensar que é um gato. -![Ciclo de Otimização de Imagem](../../../../../translated_images/pt/ideal-cat-loop.999fbb8ff306e044.webp) +![Ciclo de Otimização de Imagem](../../../../../translated_images/pt-PT/ideal-cat-loop.999fbb8ff306e044.webp) No entanto, se fizermos isso, obteremos algo muito semelhante a um ruído aleatório. Isso acontece porque *existem muitas maneiras de fazer a rede pensar que a imagem de entrada é um gato*, incluindo algumas que não fazem sentido visualmente. Embora essas imagens contenham muitos padrões típicos de um gato, não há nada que as obrigue a serem visualmente distintas. Para melhorar o resultado, podemos adicionar outro termo à função de perda, chamado **perda de variação**. É uma métrica que mostra quão semelhantes são os pixels vizinhos da imagem. Minimizar a perda de variação torna a imagem mais suave e elimina o ruído, revelando padrões mais visualmente apelativos. Aqui está um exemplo de imagens "ideais", classificadas como gato e zebra com alta probabilidade: -![Gato Ideal](../../../../../translated_images/pt/ideal-cat.203dd4597643d6b0.webp) | ![Zebra Ideal](../../../../../translated_images/pt/ideal-zebra.7f70e8b54ee15a7a.webp) +![Gato Ideal](../../../../../translated_images/pt-PT/ideal-cat.203dd4597643d6b0.webp) | ![Zebra Ideal](../../../../../translated_images/pt-PT/ideal-zebra.7f70e8b54ee15a7a.webp) -----|----- *Gato Ideal* | *Zebra Ideal* Uma abordagem semelhante pode ser usada para realizar os chamados **ataques adversariais** numa rede neural. Suponha que queremos enganar uma rede neural e fazer com que um cão pareça um gato. Se pegarmos na imagem de um cão, que é reconhecida pela rede como um cão, podemos ajustá-la ligeiramente usando otimização por descida de gradiente até que a rede comece a classificá-la como um gato: -![Imagem de um Cão](../../../../../translated_images/pt/original-dog.8f68a67d2fe0911f.webp) | ![Imagem de um cão classificada como gato](../../../../../translated_images/pt/adversarial-dog.d9fc7773b0142b89.webp) +![Imagem de um Cão](../../../../../translated_images/pt-PT/original-dog.8f68a67d2fe0911f.webp) | ![Imagem de um cão classificada como gato](../../../../../translated_images/pt-PT/adversarial-dog.d9fc7773b0142b89.webp) -----|----- *Imagem original de um cão* | *Imagem de um cão classificada como gato* diff --git a/translations/pt/lessons/4-ComputerVision/09-Autoencoders/README.md b/translations/pt/lessons/4-ComputerVision/09-Autoencoders/README.md index 609f0687..431f56d9 100644 --- a/translations/pt/lessons/4-ComputerVision/09-Autoencoders/README.md +++ b/translations/pt/lessons/4-ComputerVision/09-Autoencoders/README.md @@ -19,7 +19,7 @@ No entanto, podemos querer usar dados brutos (não etiquetados) para treinar ext Como estamos a treinar um autoencoder para capturar o máximo de informação possível da imagem original para uma reconstrução precisa, a rede tenta encontrar a melhor **representação** das imagens de entrada para captar o seu significado. -![Diagrama de Autoencoder](../../../../../translated_images/pt/autoencoder_schema.5e6fc9ad98a5eb61.webp) +![Diagrama de Autoencoder](../../../../../translated_images/pt-PT/autoencoder_schema.5e6fc9ad98a5eb61.webp) > Imagem retirada do [blog da Keras](https://blog.keras.io/building-autoencoders-in-keras.html) diff --git a/translations/pt/lessons/4-ComputerVision/11-ObjectDetection/README.md b/translations/pt/lessons/4-ComputerVision/11-ObjectDetection/README.md index cf2a25de..3049c818 100644 --- a/translations/pt/lessons/4-ComputerVision/11-ObjectDetection/README.md +++ b/translations/pt/lessons/4-ComputerVision/11-ObjectDetection/README.md @@ -13,7 +13,7 @@ Os modelos de classificação de imagens que abordámos até agora recebiam uma ## [Questionário pré-aula](https://ff-quizzes.netlify.app/en/ai/quiz/21) -![Detecção de Objetos](../../../../../translated_images/pt/Screen_Shot_2016-11-17_at_11.14.54_AM.b4bb3769353287be.webp) +![Detecção de Objetos](../../../../../translated_images/pt-PT/Screen_Shot_2016-11-17_at_11.14.54_AM.b4bb3769353287be.webp) > Imagem do [site YOLO v2](https://pjreddie.com/darknet/yolov2/) @@ -25,7 +25,7 @@ Supondo que queremos encontrar um gato numa imagem, uma abordagem muito ingénua 2. Executar a classificação de imagem em cada bloco. 3. Os blocos que resultarem numa ativação suficientemente alta podem ser considerados como contendo o objeto em questão. -![Detecção Ingénua de Objetos](../../../../../translated_images/pt/naive-detection.e7f1ba220ccd08c6.webp) +![Detecção Ingénua de Objetos](../../../../../translated_images/pt-PT/naive-detection.e7f1ba220ccd08c6.webp) > *Imagem do [Caderno de Exercícios](ObjectDetection-TF.ipynb)* @@ -42,7 +42,7 @@ Poderá encontrar os seguintes conjuntos de dados para esta tarefa: * [PASCAL VOC](http://host.robots.ox.ac.uk/pascal/VOC/) - 20 classes * [COCO](http://cocodataset.org/#home) - Objetos Comuns em Contexto. 80 classes, caixas delimitadoras e máscaras de segmentação -![COCO](../../../../../translated_images/pt/coco-examples.71bc60380fa6cceb.webp) +![COCO](../../../../../translated_images/pt-PT/coco-examples.71bc60380fa6cceb.webp) ## Métricas de Detecção de Objetos @@ -50,7 +50,7 @@ Poderá encontrar os seguintes conjuntos de dados para esta tarefa: Enquanto na classificação de imagens é fácil medir o desempenho do algoritmo, na detecção de objetos precisamos de medir tanto a correção da classe como a precisão da localização da caixa delimitadora inferida. Para esta última, utilizamos a chamada **Interseção sobre União** (IoU), que mede o quão bem duas caixas (ou duas áreas arbitrárias) se sobrepõem. -![IoU](../../../../../translated_images/pt/iou_equation.9a4751d40fff4e11.webp) +![IoU](../../../../../translated_images/pt-PT/iou_equation.9a4751d40fff4e11.webp) > *Figura 2 de [este excelente artigo sobre IoU](https://pyimagesearch.com/2016/11/07/intersection-over-union-iou-for-object-detection/)* @@ -98,11 +98,11 @@ Existem duas grandes classes de algoritmos de detecção de objetos: [R-CNN](http://islab.ulsan.ac.kr/files/announcement/513/rcnn_pami.pdf) utiliza [Selective Search](http://www.huppelen.nl/publications/selectiveSearchDraft.pdf) para gerar uma estrutura hierárquica de regiões ROI, que são então passadas por extratores de características CNN e classificadores SVM para determinar a classe do objeto, e regressão linear para determinar as coordenadas da *caixa delimitadora*. [Artigo Oficial](https://arxiv.org/pdf/1506.01497v1.pdf) -![RCNN](../../../../../translated_images/pt/rcnn1.cae407020dfb1d1f.webp) +![RCNN](../../../../../translated_images/pt-PT/rcnn1.cae407020dfb1d1f.webp) > *Imagem de van de Sande et al. ICCV’11* -![RCNN-1](../../../../../translated_images/pt/rcnn2.2d9530bb83516484.webp) +![RCNN-1](../../../../../translated_images/pt-PT/rcnn2.2d9530bb83516484.webp) > *Imagens de [este artigo](https://towardsdatascience.com/r-cnn-fast-r-cnn-faster-r-cnn-yolo-object-detection-algorithms-36d53571365e)* @@ -110,7 +110,7 @@ Existem duas grandes classes de algoritmos de detecção de objetos: Esta abordagem é semelhante à R-CNN, mas as regiões são definidas após as camadas de convolução terem sido aplicadas. -![FRCNN](../../../../../translated_images/pt/f-rcnn.3cda6d9bb4188875.webp) +![FRCNN](../../../../../translated_images/pt-PT/f-rcnn.3cda6d9bb4188875.webp) > Imagem do [Artigo Oficial](https://www.cv-foundation.org/openaccess/content_iccv_2015/papers/Girshick_Fast_R-CNN_ICCV_2015_paper.pdf), [arXiv](https://arxiv.org/pdf/1504.08083.pdf), 2015 @@ -118,7 +118,7 @@ Esta abordagem é semelhante à R-CNN, mas as regiões são definidas após as c A ideia principal desta abordagem é usar uma rede neural para prever ROIs - a chamada *Rede de Proposta de Região*. [Artigo](https://arxiv.org/pdf/1506.01497.pdf), 2016 -![FasterRCNN](../../../../../translated_images/pt/faster-rcnn.8d46c099b87ef30a.webp) +![FasterRCNN](../../../../../translated_images/pt-PT/faster-rcnn.8d46c099b87ef30a.webp) > Imagem do [artigo oficial](https://arxiv.org/pdf/1506.01497.pdf) @@ -130,7 +130,7 @@ Este algoritmo é ainda mais rápido que o Faster R-CNN. A ideia principal é a 1. As características são processadas por **Position-Sensitive Score Map**. Cada objeto das classes $C$ é dividido em regiões $k\times k$, e treinamos para prever partes dos objetos. 1. Para cada parte das regiões $k\times k$, todas as redes votam pelas classes de objetos, e a classe de objeto com o voto máximo é selecionada. -![r-fcn image](../../../../../translated_images/pt/r-fcn.13eb88158b99a3da.webp) +![r-fcn image](../../../../../translated_images/pt-PT/r-fcn.13eb88158b99a3da.webp) > Imagem do [artigo oficial](https://arxiv.org/abs/1605.06409) @@ -141,7 +141,7 @@ YOLO é um algoritmo de uma única passagem em tempo real. A ideia principal é * A imagem é dividida em regiões $S\times S$. * Para cada região, **CNN** prevê $n$ objetos possíveis, coordenadas da *caixa delimitadora* e *confiança*=*probabilidade* * IoU. - ![YOLO](../../../../../translated_images/pt/yolo.a2648ec82ee8bb4e.webp) + ![YOLO](../../../../../translated_images/pt-PT/yolo.a2648ec82ee8bb4e.webp) > Imagem do [artigo oficial](https://arxiv.org/abs/1506.02640) diff --git a/translations/pt/lessons/5-NLP/14-Embeddings/README.md b/translations/pt/lessons/5-NLP/14-Embeddings/README.md index ce409ea2..70c760ac 100644 --- a/translations/pt/lessons/5-NLP/14-Embeddings/README.md +++ b/translations/pt/lessons/5-NLP/14-Embeddings/README.md @@ -19,7 +19,7 @@ Assim, a camada de embedding receberia uma palavra como entrada e produziria um Ao usar uma camada de embedding como a primeira camada na nossa rede de classificação, podemos mudar de um modelo de bag-of-words para um modelo de **embedding bag**, onde primeiro convertemos cada palavra no nosso texto no embedding correspondente e, em seguida, calculamos alguma função agregada sobre todos esses embeddings, como `sum`, `average` ou `max`. -![Imagem mostrando um classificador de embedding para cinco palavras de sequência.](../../../../../translated_images/pt/embedding-classifier-example.b77f021a7ee67eee.webp) +![Imagem mostrando um classificador de embedding para cinco palavras de sequência.](../../../../../translated_images/pt-PT/embedding-classifier-example.b77f021a7ee67eee.webp) > Imagem do autor @@ -40,7 +40,7 @@ Para isso, precisamos de pré-treinar o nosso modelo de embedding numa grande co CBoW é mais rápido, enquanto skip-gram é mais lento, mas faz um trabalho melhor ao representar palavras menos frequentes. -![Imagem mostrando os algoritmos CBoW e Skip-Gram para converter palavras em vetores.](../../../../../translated_images/pt/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) +![Imagem mostrando os algoritmos CBoW e Skip-Gram para converter palavras em vetores.](../../../../../translated_images/pt-PT/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) > Imagem retirada [deste artigo](https://arxiv.org/pdf/1301.3781.pdf) diff --git a/translations/pt/lessons/5-NLP/15-LanguageModeling/README.md b/translations/pt/lessons/5-NLP/15-LanguageModeling/README.md index 9113413a..65ff3a18 100644 --- a/translations/pt/lessons/5-NLP/15-LanguageModeling/README.md +++ b/translations/pt/lessons/5-NLP/15-LanguageModeling/README.md @@ -23,7 +23,7 @@ Nos nossos exemplos anteriores, utilizámos embeddings semânticos pré-treinado * **Continuous Bag-of-Words** (CBoW), onde prevemos o token central $W_0$ numa sequência de tokens $W_{-N}$, ..., $W_N$. * **Skip-gram**, onde prevemos um conjunto de tokens vizinhos {$W_{-N},\dots, W_{-1}, W_1,\dots, W_N$} a partir do token central $W_0$. -![imagem do artigo sobre conversão de palavras em vetores](../../../../../translated_images/pt/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) +![imagem do artigo sobre conversão de palavras em vetores](../../../../../translated_images/pt-PT/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) > Imagem retirada [deste artigo](https://arxiv.org/pdf/1301.3781.pdf) diff --git a/translations/pt/lessons/5-NLP/16-RNN/README.md b/translations/pt/lessons/5-NLP/16-RNN/README.md index fc78dd6e..48ca7c52 100644 --- a/translations/pt/lessons/5-NLP/16-RNN/README.md +++ b/translations/pt/lessons/5-NLP/16-RNN/README.md @@ -15,7 +15,7 @@ Nas secções anteriores, utilizámos representações semânticas ricas de text Para capturar o significado de uma sequência de texto, precisamos de usar outra arquitetura de rede neural, chamada **rede neural recorrente**, ou RNN. Numa RNN, passamos a nossa frase pela rede, um símbolo de cada vez, e a rede produz um **estado**, que depois passamos novamente à rede com o próximo símbolo. -![RNN](../../../../../translated_images/pt/rnn.27f5c29c53d727b5.webp) +![RNN](../../../../../translated_images/pt-PT/rnn.27f5c29c53d727b5.webp) > Imagem do autor @@ -61,7 +61,7 @@ Discutimos redes recorrentes que operam numa direção, do início de uma sequê Uma rede recorrente, seja unidirecional ou bidirecional, captura certos padrões dentro de uma sequência e pode armazená-los num vetor de estado ou passá-los para a saída. Tal como nas redes convolucionais, podemos construir outra camada recorrente sobre a primeira para capturar padrões de nível superior e construir a partir dos padrões de baixo nível extraídos pela primeira camada. Isto leva-nos à noção de uma **RNN multicamada**, que consiste em duas ou mais redes recorrentes, onde a saída da camada anterior é passada para a próxima camada como entrada. -![Imagem mostrando uma RNN multicamada com LSTM](../../../../../translated_images/pt/multi-layer-lstm.dd975e29bb2a59fe.webp) +![Imagem mostrando uma RNN multicamada com LSTM](../../../../../translated_images/pt-PT/multi-layer-lstm.dd975e29bb2a59fe.webp) *Imagem retirada [deste excelente artigo](https://towardsdatascience.com/from-a-lstm-cell-to-a-multilayer-lstm-network-with-pytorch-2899eb5696f3) de Fernando López* diff --git a/translations/pt/lessons/5-NLP/17-GenerativeNetworks/README.md b/translations/pt/lessons/5-NLP/17-GenerativeNetworks/README.md index 33f358f8..eccc5d4c 100644 --- a/translations/pt/lessons/5-NLP/17-GenerativeNetworks/README.md +++ b/translations/pt/lessons/5-NLP/17-GenerativeNetworks/README.md @@ -19,7 +19,7 @@ Na arquitetura de RNN discutida na unidade anterior, cada unidade RNN produzia o Isto permite diferentes arquiteturas neuronais, como mostrado na imagem abaixo: -![Imagem mostrando padrões comuns de redes neuronais recorrentes.](../../../../../translated_images/pt/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) +![Imagem mostrando padrões comuns de redes neuronais recorrentes.](../../../../../translated_images/pt-PT/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) > Imagem do artigo [Unreasonable Effectiveness of Recurrent Neural Networks](http://karpathy.github.io/2015/05/21/rnn-effectiveness/) por [Andrej Karpaty](http://karpathy.github.io/) @@ -32,7 +32,7 @@ Nesta unidade, vamos focar-nos em modelos generativos simples que nos ajudam a g Vamos treinar esta RNN para gerar texto passo a passo. Em cada passo, tomaremos uma sequência de caracteres de comprimento `nchars` e pediremos à rede que gere o próximo carácter de saída para cada carácter de entrada: -![Imagem mostrando um exemplo de geração de RNN da palavra 'HELLO'.](../../../../../translated_images/pt/rnn-generate.56c54afb52f9781d.webp) +![Imagem mostrando um exemplo de geração de RNN da palavra 'HELLO'.](../../../../../translated_images/pt-PT/rnn-generate.56c54afb52f9781d.webp) Ao gerar texto (durante a inferência), começamos com um **prompt**, que é passado pelas células RNN para gerar o estado intermédio, e a partir deste estado começa a geração. Geramos um carácter de cada vez e passamos o estado e o carácter gerado para outra célula RNN para gerar o próximo, até gerarmos caracteres suficientes. diff --git a/translations/pt/lessons/5-NLP/18-Transformers/README.md b/translations/pt/lessons/5-NLP/18-Transformers/README.md index 744e3ab2..784c59ad 100644 --- a/translations/pt/lessons/5-NLP/18-Transformers/README.md +++ b/translations/pt/lessons/5-NLP/18-Transformers/README.md @@ -20,13 +20,13 @@ Com RNNs, a tarefa de sequência para sequência é implementada por duas redes Os **Mecanismos de Atenção** fornecem um meio de ponderar o impacto contextual de cada vetor de entrada em cada previsão de saída da RNN. Isto é implementado criando atalhos entre estados intermediários da RNN de entrada e a RNN de saída. Desta forma, ao gerar o símbolo de saída yt, consideramos todos os estados ocultos de entrada hi, com diferentes coeficientes de peso αt,i. -![Imagem mostrando um modelo codificador/descodificador com uma camada de atenção aditiva](../../../../../translated_images/pt/encoder-decoder-attention.7a726296894fb567.webp) +![Imagem mostrando um modelo codificador/descodificador com uma camada de atenção aditiva](../../../../../translated_images/pt-PT/encoder-decoder-attention.7a726296894fb567.webp) > O modelo codificador-descodificador com mecanismo de atenção aditiva em [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf), citado deste [post de blog](https://lilianweng.github.io/lil-log/2018/06/24/attention-attention.html) A matriz de atenção {αi,j} representaria o grau em que certas palavras de entrada influenciam a geração de uma determinada palavra na sequência de saída. Abaixo está um exemplo de tal matriz: -![Imagem mostrando um alinhamento de exemplo encontrado por RNNsearch-50, retirada de Bahdanau - arviz.org](../../../../../translated_images/pt/bahdanau-fig3.09ba2d37f202a6af.webp) +![Imagem mostrando um alinhamento de exemplo encontrado por RNNsearch-50, retirada de Bahdanau - arviz.org](../../../../../translated_images/pt-PT/bahdanau-fig3.09ba2d37f202a6af.webp) > Figura de [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf) (Fig.3) @@ -66,7 +66,7 @@ O resultado que obtemos com o embedding posicional incorpora tanto o token origi A seguir, precisamos capturar alguns padrões dentro da nossa sequência. Para isso, os transformers usam um mecanismo de **auto-atenção**, que é essencialmente atenção aplicada à mesma sequência como entrada e saída. Aplicar auto-atenção permite-nos levar em conta o **contexto** dentro da frase e ver quais palavras estão inter-relacionadas. Por exemplo, permite-nos ver quais palavras são referidas por correferências, como *it*, e também considerar o contexto: -![](../../../../../translated_images/pt/CoreferenceResolution.861924d6d384a7d6.webp) +![](../../../../../translated_images/pt-PT/CoreferenceResolution.861924d6d384a7d6.webp) > Imagem do [Blog do Google](https://research.googleblog.com/2017/08/transformer-novel-neural-network.html) @@ -91,7 +91,7 @@ Como cada posição de entrada é mapeada independentemente para cada posição **BERT** (Bidirectional Encoder Representations from Transformers) é uma rede transformer muito grande com várias camadas: 12 camadas para o *BERT-base* e 24 para o *BERT-large*. O modelo é primeiro pré-treinado num grande corpus de dados de texto (WikiPedia + livros) usando treino não supervisionado (prevendo palavras mascaradas numa frase). Durante o pré-treino, o modelo absorve níveis significativos de compreensão da linguagem, que podem ser aproveitados com outros conjuntos de dados usando ajuste fino. Este processo é chamado de **aprendizagem por transferência**. -![imagem de http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/pt/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp) +![imagem de http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/pt-PT/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp) > Imagem [fonte](http://jalammar.github.io/illustrated-bert/) diff --git a/translations/pt/lessons/5-NLP/18-Transformers/READMEtransformers.md b/translations/pt/lessons/5-NLP/18-Transformers/READMEtransformers.md index 34c2fcad..3d82d830 100644 --- a/translations/pt/lessons/5-NLP/18-Transformers/READMEtransformers.md +++ b/translations/pt/lessons/5-NLP/18-Transformers/READMEtransformers.md @@ -11,13 +11,13 @@ Com RNNs, a sequência-para-sequência é implementada por duas redes recorrente **Mecanismos de Atenção** fornecem um meio de ponderar o impacto contextual de cada vetor de entrada em cada previsão de saída da RNN. A forma como é implementado é criando atalhos entre estados intermediários da RNN de entrada e da RNN de saída. Dessa maneira, ao gerar o símbolo de saída yt, levaremos em conta todos os estados ocultos de entrada hi, com diferentes coeficientes de peso αt,i. -![Imagem mostrando um modelo codificador/decodificador com uma camada de atenção aditiva](../../../../../translated_images/pt/encoder-decoder-attention.7a726296894fb567.webp) +![Imagem mostrando um modelo codificador/decodificador com uma camada de atenção aditiva](../../../../../translated_images/pt-PT/encoder-decoder-attention.7a726296894fb567.webp) > O modelo codificador-decodificador com mecanismo de atenção aditiva em [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf), citado a partir [deste post de blog](https://lilianweng.github.io/lil-log/2018/06/24/attention-attention.html) A matriz de atenção {αi,j} representaria o grau em que certas palavras de entrada desempenham na geração de uma palavra específica na sequência de saída. Abaixo está um exemplo de tal matriz: -![Imagem mostrando um alinhamento de exemplo encontrado por RNNsearch-50, tirada de Bahdanau - arviz.org](../../../../../translated_images/pt/bahdanau-fig3.09ba2d37f202a6af.webp) +![Imagem mostrando um alinhamento de exemplo encontrado por RNNsearch-50, tirada de Bahdanau - arviz.org](../../../../../translated_images/pt-PT/bahdanau-fig3.09ba2d37f202a6af.webp) > Figura de [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf) (Fig.3) @@ -57,7 +57,7 @@ O resultado que obtemos com o embutimento posicional incorpora tanto o token ori Em seguida, precisamos capturar alguns padrões dentro da nossa sequência. Para fazer isso, os transformers usam um mecanismo de **autoatenção**, que é essencialmente atenção aplicada à mesma sequência como entrada e saída. A aplicação de autoatenção nos permite levar em conta o **contexto** dentro da sentença e ver quais palavras estão inter-relacionadas. Por exemplo, isso nos permite ver quais palavras são referidas por co-referências, como *isso*, e também levar o contexto em consideração: -![](../../../../../translated_images/pt/CoreferenceResolution.861924d6d384a7d6.webp) +![](../../../../../translated_images/pt-PT/CoreferenceResolution.861924d6d384a7d6.webp) > Imagem do [Blog do Google](https://research.googleblog.com/2017/08/transformer-novel-neural-network.html) @@ -82,7 +82,7 @@ Como cada posição de entrada é mapeada independentemente para cada posição **BERT** (Representações de Codificador Bidirecional de Transformers) é uma rede transformer multilayer muito grande com 12 camadas para *BERT-base* e 24 para *BERT-large*. O modelo é primeiro pré-treinado em um grande corpus de dados textuais (WikiPedia + livros) usando treinamento não supervisionado (previsão de palavras mascaradas em uma sentença). Durante o pré-treinamento, o modelo absorve níveis significativos de compreensão da linguagem, que podem ser aproveitados com outros conjuntos de dados usando ajuste fino. Este processo é chamado de **aprendizado por transferência**. -![imagem de http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/pt/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp) +![imagem de http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/pt-PT/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp) > Imagem [fonte](http://jalammar.github.io/illustrated-bert/) diff --git a/translations/pt/lessons/5-NLP/19-NER/README.md b/translations/pt/lessons/5-NLP/19-NER/README.md index 1b113d2e..d33ce184 100644 --- a/translations/pt/lessons/5-NLP/19-NER/README.md +++ b/translations/pt/lessons/5-NLP/19-NER/README.md @@ -58,7 +58,7 @@ infant | O Como precisamos construir uma correspondência um-para-um entre tokens e classes, podemos treinar um modelo de rede neural **muitos-para-muitos** da seguinte forma: -![Imagem mostrando padrões comuns de redes neuronais recorrentes.](../../../../../translated_images/pt/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) +![Imagem mostrando padrões comuns de redes neuronais recorrentes.](../../../../../translated_images/pt-PT/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) > *Imagem retirada [deste artigo](http://karpathy.github.io/2015/05/21/rnn-effectiveness/) por [Andrej Karpathy](http://karpathy.github.io/). Os modelos de classificação de tokens NER correspondem à arquitetura de rede mais à direita nesta imagem.* diff --git a/translations/pt/lessons/6-Other/23-MultiagentSystems/README.md b/translations/pt/lessons/6-Other/23-MultiagentSystems/README.md index 59b35b73..4981600c 100644 --- a/translations/pt/lessons/6-Other/23-MultiagentSystems/README.md +++ b/translations/pt/lessons/6-Other/23-MultiagentSystems/README.md @@ -70,7 +70,7 @@ Pode abrir um dos modelos, por exemplo **Biology → Flocking**. Depois de abrir o modelo, será levado ao ecrã principal do NetLogo. Aqui está um modelo de exemplo que descreve a população de lobos e ovelhas, dado recursos finitos (relva). -![NetLogo Main Screen](../../../../../translated_images/pt/NetLogo-Main.32653711ec1a01b3.webp) +![NetLogo Main Screen](../../../../../translated_images/pt-PT/NetLogo-Main.32653711ec1a01b3.webp) > Captura de ecrã por Dmitry Soshnikov diff --git a/translations/tw/README.md b/translations/tw/README.md index 3e6f3280..c548a935 100644 --- a/translations/tw/README.md +++ b/translations/tw/README.md @@ -23,7 +23,7 @@ CO_OP_TRANSLATOR_METADATA: # 人工智慧初學者課程綱要 -|![Sketchnote by @girlie_mac https://twitter.com/girlie_mac](../../../../translated_images/tw/ai-overview.0857791951d19500.webp)| +|![Sketchnote by @girlie_mac https://twitter.com/girlie_mac](../../../../translated_images/zh-TW/ai-overview.0857791951d19500.webp)| |:---:| | AI For Beginners - _筆記圖解由 [@girlie_mac](https://twitter.com/girlie_mac) 製作_ | diff --git a/translations/tw/lessons/1-Intro/README.md b/translations/tw/lessons/1-Intro/README.md index 1f61e8c2..3ce3e51c 100644 --- a/translations/tw/lessons/1-Intro/README.md +++ b/translations/tw/lessons/1-Intro/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 人工智慧簡介 -![人工智慧內容摘要的手繪圖](../../../../translated_images/tw/ai-intro.bf28d1ac4235881c.webp) +![人工智慧內容摘要的手繪圖](../../../../translated_images/zh-TW/ai-intro.bf28d1ac4235881c.webp) > 手繪筆記由 [Tomomi Imura](https://twitter.com/girlie_mac) 提供 @@ -19,7 +19,7 @@ CO_OP_TRANSLATOR_METADATA: 最初，電腦由 [Charles Babbage](https://en.wikipedia.org/wiki/Charles_Babbage) 發明，用來根據明確定義的程序（演算法）操作數字。現代電腦雖然比19世紀提出的原型複雜得多，但仍然遵循受控計算的理念。因此，只要我們知道達成目標所需的精確步驟，就可以編程讓電腦完成某些事情。 -![一個人的照片](../../../../translated_images/tw/dsh_age.d212a30d4e54fb5f.webp) +![一個人的照片](../../../../translated_images/zh-TW/dsh_age.d212a30d4e54fb5f.webp) > 照片由 [Vickie Soshnikova](http://twitter.com/vickievalerie) 提供 @@ -46,7 +46,7 @@ CO_OP_TRANSLATOR_METADATA: 在討論**[智能](https://en.wikipedia.org/wiki/Intelligence)**這個術語時，其中一個問題是我們對此並沒有明確的定義。有人可能認為智能與**抽象思維**或**自我意識**相關，但我們無法準確定義它。 -![貓的照片](../../../../translated_images/tw/photo-cat.8c8e8fb760ffe457.webp) +![貓的照片](../../../../translated_images/zh-TW/photo-cat.8c8e8fb760ffe457.webp) > [照片](https://unsplash.com/photos/75715CVEJhI)由 [Amber Kipp](https://unsplash.com/@sadmax) 提供，來自 Unsplash @@ -98,13 +98,13 @@ CO_OP_TRANSLATOR_METADATA: > | 那機器學習呢？ | | > |--------------|-----------| -> | 基於電腦通過某些數據學習解決問題的人工智慧部分稱為**機器學習**。我們不會在本課程中考慮傳統機器學習——我們推薦您參考單獨的 [機器學習初學者課程](http://aka.ms/ml-beginners)。 | ![機器學習初學者課程](../../../../translated_images/tw/ml-for-beginners.9e4fed176fd5817d.webp) | +> | 基於電腦通過某些數據學習解決問題的人工智慧部分稱為**機器學習**。我們不會在本課程中考慮傳統機器學習——我們推薦您參考單獨的 [機器學習初學者課程](http://aka.ms/ml-beginners)。 | ![機器學習初學者課程](../../../../translated_images/zh-TW/ml-for-beginners.9e4fed176fd5817d.webp) | ## 人工智慧的簡史 人工智慧作為一個領域始於20世紀中期。最初，符號推理是主要方法，並取得了一些重要成功，例如專家系統——能夠在某些有限問題領域中充當專家的電腦程序。然而，很快就發現這種方法的可擴展性不佳。從專家中提取知識、在電腦中表示知識並保持知識庫的準確性，事實證明是一項非常複雜且在許多情況下成本過高的任務。這導致了1970年代的所謂[人工智慧寒冬](https://en.wikipedia.org/wiki/AI_winter)。 -人工智慧簡史 +人工智慧簡史 > 圖片由 [Dmitry Soshnikov](http://soshnikov.com) 提供 @@ -124,7 +124,7 @@ CO_OP_TRANSLATOR_METADATA: * 現代助理，例如 Cortana、Siri 或 Google Assistant，都是混合系統，使用神經網路將語音轉換為文本並識別我們的意圖，然後採用一些推理或明確算法執行所需操作。 * 未來，我們可能期待一個完全基於神經網路的模型能夠自行處理對話。最近的 GPT 和 [Turing-NLG](https://www.microsoft.com/research/blog/turing-nlg-a-17-billion-parameter-language-model-by-microsoft) 神經網路系列顯示了巨大的成功。 -圖靈測試的演變 +圖靈測試的演變 > 圖片由 Dmitry Soshnikov 提供，[照片](https://unsplash.com/photos/r8LmVbUKgns) 由 [Marina Abrosimova](https://unsplash.com/@abrosimova_marina_foto) 提供，Unsplash ## 最近的人工智慧研究 diff --git a/translations/tw/lessons/2-Symbolic/Animals.ipynb b/translations/tw/lessons/2-Symbolic/Animals.ipynb index 32407e00..e277a1a4 100644 --- a/translations/tw/lessons/2-Symbolic/Animals.ipynb +++ b/translations/tw/lessons/2-Symbolic/Animals.ipynb @@ -12,7 +12,7 @@ "\n", "在此範例中，我們將實作一個簡單的知識基礎系統，根據一些物理特徵來判斷動物。系統可以用以下的 AND-OR 樹來表示（這是整個樹的一部分，我們可以輕鬆地添加更多規則）：\n", "\n", - "![](../../../../../../translated_images/tw/AND-OR-Tree.5592d2c70187f283.webp)\n" + "![](../../../../../../translated_images/zh-TW/AND-OR-Tree.5592d2c70187f283.webp)\n" ] }, { diff --git a/translations/tw/lessons/2-Symbolic/README.md b/translations/tw/lessons/2-Symbolic/README.md index 593ef4d4..fa5f9259 100644 --- a/translations/tw/lessons/2-Symbolic/README.md +++ b/translations/tw/lessons/2-Symbolic/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 知識表示與專家系統 -![Symbolic AI 內容摘要](../../../../../../translated_images/tw/ai-symbolic.715a30cb610411a6.webp) +![Symbolic AI 內容摘要](../../../../../../translated_images/zh-TW/ai-symbolic.715a30cb610411a6.webp) > 筆記作者：[Tomomi Imura](https://twitter.com/girlie_mac) @@ -35,13 +35,13 @@ CO_OP_TRANSLATOR_METADATA: * **知識（Knowledge）** 是將資訊整合到我們的世界模型中。例如，一旦學習到電腦是什麼，我們就開始瞭解它如何運作、價格多少、用途為何。這種相互關聯的概念網絡形成了我們的知識。 * **智慧（Wisdom）** 是我們對世界理解的更高層次，代表*元知識*，例如有關知識應用的時機與方式的概念。 - + *圖片來源 [維基百科](https://commons.wikimedia.org/w/index.php?curid=37705247)，作者 Longlivetheux 自作，CC BY-SA 4.0* 因此，**知識表示** 問題是尋找一些有效的方法來以資料形式在電腦內代表知識，使其可被自動使用。這可視為一個光譜： -![知識表示光譜](../../../../../../translated_images/tw/knowledge-spectrum.b60df631852c0217.webp) +![知識表示光譜](../../../../../../translated_images/zh-TW/knowledge-spectrum.b60df631852c0217.webp) > 圖片來源 [Dmitry Soshnikov](http://soshnikov.com) @@ -94,7 +94,7 @@ Python | 區塊語法 | 縮排 符號 AI 的早期成功之一是所謂的**專家系統** — 設計成在有限問題領域中，充當專家的電腦系統。它們基於由一人或多位專家提取的**知識庫**，並含有執行推理的**推理引擎**。 -![人類架構](../../../../../../translated_images/tw/arch-human.5d4d35f1bba3ab1c.webp) | ![知識基礎系統](../../../../../../translated_images/tw/arch-kbs.3ec5c150b09fa8da.webp) +![人類架構](../../../../../../translated_images/zh-TW/arch-human.5d4d35f1bba3ab1c.webp) | ![知識基礎系統](../../../../../../translated_images/zh-TW/arch-kbs.3ec5c150b09fa8da.webp) ---------------------------------------------|------------------------------------------------ 人類神經系統簡化結構 | 知識基礎系統架構 @@ -106,7 +106,7 @@ Python | 區塊語法 | 縮排 以下為根據動物物理特徵判斷動物的專家系統示例： -![AND-OR 樹](../../../../../../translated_images/tw/AND-OR-Tree.5592d2c70187f283.webp) +![AND-OR 樹](../../../../../../translated_images/zh-TW/AND-OR-Tree.5592d2c70187f283.webp) > 圖片來源 [Dmitry Soshnikov](http://soshnikov.com) @@ -175,7 +175,7 @@ THEN the animal is a carnivore 在語義網中，所有表示均基於三元組。每個物件和每個關係均由 URI 唯一標識。例如，如果我們想要陳述這個 AI 課程是由 Dmitry Soshnikov 於 2022 年 1 月 1 日開發的事實——我們可以使用以下三元組： - + ``` http://github.com/microsoft/ai-for-beginners http://www.example.com/terms/creation-date “Jan 1, 2022” @@ -186,7 +186,7 @@ http://github.com/microsoft/ai-for-beginners http://purl.org/dc/elements/1.1/cre 在較複雜的情況下，如果我們想定義一個創建者列表，可以使用 RDF 中定義的一些資料結構。 - + > 上述圖由 [Dmitry Soshnikov](http://soshnikov.com) 製作 @@ -210,7 +210,7 @@ GROUP BY ?eyeColorLabel > ✅ 如果你想嘗試建立自己的本體，或者開啟現有的本體，可以使用一個很棒的視覺化本體編輯器 [Protégé](https://protege.stanford.edu/)。下載它或線上使用。 - + *Web Protégé 編輯器開啟羅曼諾夫家族本體。截圖由 Dmitry Soshnikov 提供* diff --git a/translations/tw/lessons/3-NeuralNetworks/05-Frameworks/README.md b/translations/tw/lessons/3-NeuralNetworks/05-Frameworks/README.md index 3f0efd33..c0494425 100644 --- a/translations/tw/lessons/3-NeuralNetworks/05-Frameworks/README.md +++ b/translations/tw/lessons/3-NeuralNetworks/05-Frameworks/README.md @@ -58,7 +58,7 @@ CO_OP_TRANSLATOR_METADATA: 考慮以下近似 5 個點的問題（圖中的 `x` 表示點）： -![線性模型](../../../../../translated_images/tw/overfit1.f24b71c6f652e59e.webp) | ![過擬合模型](../../../../../translated_images/tw/overfit2.131f5800ae10ca5e.webp) +![線性模型](../../../../../translated_images/zh-TW/overfit1.f24b71c6f652e59e.webp) | ![過擬合模型](../../../../../translated_images/zh-TW/overfit2.131f5800ae10ca5e.webp) -------------------------|-------------------------- **線性模型，2 個參數** | **非線性模型，7 個參數** 訓練誤差 = 5.3 | 訓練誤差 = 0 @@ -79,7 +79,7 @@ CO_OP_TRANSLATOR_METADATA: 從上圖可以看出，過擬合可以通過非常低的訓練誤差和非常高的驗證誤差來檢測。通常在訓練過程中，我們會看到訓練誤差和驗證誤差都開始下降，但在某個時候，驗證誤差可能停止下降並開始上升。這將是過擬合的跡象，表明我們應該停止訓練（或者至少保存模型的快照）。 -![過擬合圖示](../../../../../translated_images/tw/Overfitting.408ad91cd90b4371.webp) +![過擬合圖示](../../../../../translated_images/zh-TW/Overfitting.408ad91cd90b4371.webp) ## 如何防止過擬合 diff --git a/translations/tw/lessons/3-NeuralNetworks/README.md b/translations/tw/lessons/3-NeuralNetworks/README.md index b863925f..fbe9eb67 100644 --- a/translations/tw/lessons/3-NeuralNetworks/README.md +++ b/translations/tw/lessons/3-NeuralNetworks/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 神經網路簡介 -![神經網路內容摘要的手繪圖](../../../../translated_images/tw/ai-neuralnetworks.1c687ae40bc86e83.webp) +![神經網路內容摘要的手繪圖](../../../../translated_images/zh-TW/ai-neuralnetworks.1c687ae40bc86e83.webp) 如我們在介紹中所討論的，實現智能的一種方法是訓練一個**計算機模型**或**人工大腦**。自20世紀中期以來，研究人員嘗試了不同的數學模型，直到最近這一方向證明非常成功。這些模仿大腦的數學模型被稱為**神經網路**。 @@ -36,13 +36,13 @@ CO_OP_TRANSLATOR_METADATA: 從生物學中，我們知道大腦由神經細胞（神經元）組成，每個神經元有多個“輸入”（樹突）和一個“輸出”（軸突）。樹突和軸突都能傳導電信號，而它們之間的連接——稱為突觸——可以表現出不同程度的導電性，這些導電性由神經遞質調節。 -![神經元模型](../../../../translated_images/tw/synapse-wikipedia.ed20a9e4726ea1c6.webp) | ![神經元模型](../../../../translated_images/tw/artneuron.1a5daa88d20ebe6f.webp) +![神經元模型](../../../../translated_images/zh-TW/synapse-wikipedia.ed20a9e4726ea1c6.webp) | ![神經元模型](../../../../translated_images/zh-TW/artneuron.1a5daa88d20ebe6f.webp) ----|---- 真實神經元 *（[圖片](https://en.wikipedia.org/wiki/Synapse#/media/File:SynapseSchematic_lines.svg) 來自維基百科）* | 人工神經元 *（作者提供圖片）* 因此，神經元的最簡單數學模型包含幾個輸入 X1, ..., XN 和一個輸出 Y，以及一系列權重 W1, ..., WN。輸出計算公式為： -Y = f\left(\sum_{i=1}^N X_iW_i\right) +Y = f\left(\sum_{i=1}^N X_iW_i\right) 其中 f 是某種非線性的**激活函數**。 diff --git a/translations/tw/lessons/4-ComputerVision/06-IntroCV/README.md b/translations/tw/lessons/4-ComputerVision/06-IntroCV/README.md index d8bee3d1..39421aea 100644 --- a/translations/tw/lessons/4-ComputerVision/06-IntroCV/README.md +++ b/translations/tw/lessons/4-ComputerVision/06-IntroCV/README.md @@ -73,14 +73,14 @@ im = cv2.cvtColor(im,cv2.COLOR_BGR2RGB) * **預處理盲文書的照片**。我們專注於如何使用閾值處理、特徵檢測、透視變換和 NumPy 操作來分離單個盲文符號，以便進一步由神經網路進行分類。 -![盲文影像](../../../../../translated_images/tw/braille.341962ff76b1bd70.webp) | ![盲文影像預處理結果](../../../../../translated_images/tw/braille-result.46530fea020b03c7.webp) | ![盲文符號](../../../../../translated_images/tw/braille-symbols.0159185ab69d5339.webp) +![盲文影像](../../../../../translated_images/zh-TW/braille.341962ff76b1bd70.webp) | ![盲文影像預處理結果](../../../../../translated_images/zh-TW/braille-result.46530fea020b03c7.webp) | ![盲文符號](../../../../../translated_images/zh-TW/braille-symbols.0159185ab69d5339.webp) ----|-----|----- > 圖片來自 [OpenCV.ipynb](OpenCV.ipynb) * **使用幀差檢測影片中的運動**。如果相機是固定的，那麼相機畫面中的幀應該彼此非常相似。由於幀被表示為陣列，只需對兩個連續幀的陣列進行相減，我們就能得到像素差異，靜態幀的差異應該很低，而當影像中有顯著運動時，差異會變高。 -![影片幀和幀差的影像](../../../../../translated_images/tw/frame-difference.706f805491a0883c.webp) +![影片幀和幀差的影像](../../../../../translated_images/zh-TW/frame-difference.706f805491a0883c.webp) > 圖片來自 [OpenCV.ipynb](OpenCV.ipynb) @@ -89,7 +89,7 @@ im = cv2.cvtColor(im,cv2.COLOR_BGR2RGB) - **密集光流** 計算每個像素的移動向量場 - **稀疏光流** 基於影像中的一些顯著特徵（例如邊緣），並從幀到幀構建它們的軌跡。 -![光流影像](../../../../../translated_images/tw/optical.1f4a94464579a83a.webp) +![光流影像](../../../../../translated_images/zh-TW/optical.1f4a94464579a83a.webp) > 圖片來自 [OpenCV.ipynb](OpenCV.ipynb) diff --git a/translations/tw/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md b/translations/tw/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md index 9f045a74..ff354c27 100644 --- a/translations/tw/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md +++ b/translations/tw/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md @@ -13,11 +13,11 @@ CO_OP_TRANSLATOR_METADATA: VGG-16 是一個在 2014 年 ImageNet top-5 分類中達到 92.7% 準確率的網路。它的層結構如下： -![ImageNet 層結構](../../../../../translated_images/tw/vgg-16-arch1.d901a5583b3a51ba.webp) +![ImageNet 層結構](../../../../../translated_images/zh-TW/vgg-16-arch1.d901a5583b3a51ba.webp) 如圖所示，VGG 採用了傳統的金字塔架構，也就是一系列的卷積-池化層。 -![ImageNet 金字塔](../../../../../translated_images/tw/vgg-16-arch.64ff2137f50dd49f.webp) +![ImageNet 金字塔](../../../../../translated_images/zh-TW/vgg-16-arch.64ff2137f50dd49f.webp) > 圖片來源：[Researchgate](https://www.researchgate.net/figure/Vgg16-model-structure-To-get-the-VGG-NIN-model-we-replace-the-2-nd-4-th-6-th-7-th_fig2_335194493) diff --git a/translations/tw/lessons/4-ComputerVision/07-ConvNets/README.md b/translations/tw/lessons/4-ComputerVision/07-ConvNets/README.md index 6ad4d03e..da62e105 100644 --- a/translations/tw/lessons/4-ComputerVision/07-ConvNets/README.md +++ b/translations/tw/lessons/4-ComputerVision/07-ConvNets/README.md @@ -17,7 +17,7 @@ CO_OP_TRANSLATOR_METADATA: 為了提取模式，我們將使用**卷積濾波器**的概念。正如你所知，圖像是由一個二維矩陣或具有顏色深度的三維張量表示的。應用濾波器意味著我們取一個相對較小的**濾波器核**矩陣，並對原始圖像中的每個像素與其鄰近點進行加權平均。我們可以將其視為一個小窗口滑過整個圖像，並根據濾波器核矩陣中的權重對所有像素進行平均。 -![垂直邊緣濾波器](../../../../../translated_images/tw/filter-vert.b7148390ca0bc356.webp) | ![水平邊緣濾波器](../../../../../translated_images/tw/filter-horiz.59b80ed4feb946ef.webp) +![垂直邊緣濾波器](../../../../../translated_images/zh-TW/filter-vert.b7148390ca0bc356.webp) | ![水平邊緣濾波器](../../../../../translated_images/zh-TW/filter-horiz.59b80ed4feb946ef.webp) ----|---- > 圖片來源：Dmitry Soshnikov @@ -38,7 +38,7 @@ CNN 的工作方式基於以下重要思想： * 我們可以設計網絡，使濾波器能夠自動訓練 * 我們可以使用相同的方法來在高層次特徵中找到模式，而不僅僅是在原始圖像中。因此，CNN 的特徵提取在特徵的層次結構中工作，從低層次的像素組合開始，到更高層次的圖像部分組合。 -![層次特徵提取](../../../../../translated_images/tw/FeatureExtractionCNN.d9b456cbdae7cb64.webp) +![層次特徵提取](../../../../../translated_images/zh-TW/FeatureExtractionCNN.d9b456cbdae7cb64.webp) > 圖片來源：[Hislop-Lynch 的論文](https://www.semanticscholar.org/paper/Computer-vision-based-pedestrian-trajectory-Hislop-Lynch/26e6f74853fc9bbb7487b06dc2cf095d36c9021d)，基於[他們的研究](https://dl.acm.org/doi/abs/10.1145/1553374.1553453) @@ -55,9 +55,9 @@ CNN 的工作方式基於以下重要思想： 例如，讓我們看看 VGG-16 的架構，這是一個在 2014 年 ImageNet 的 top-5 分類中達到 92.7% 準確率的網絡： -![ImageNet 層](../../../../../translated_images/tw/vgg-16-arch1.d901a5583b3a51ba.webp) +![ImageNet 層](../../../../../translated_images/zh-TW/vgg-16-arch1.d901a5583b3a51ba.webp) -![ImageNet 金字塔](../../../../../translated_images/tw/vgg-16-arch.64ff2137f50dd49f.webp) +![ImageNet 金字塔](../../../../../translated_images/zh-TW/vgg-16-arch.64ff2137f50dd49f.webp) > 圖片來源：[Researchgate](https://www.researchgate.net/figure/Vgg16-model-structure-To-get-the-VGG-NIN-model-we-replace-the-2-nd-4-th-6-th-7-th_fig2_335194493) diff --git a/translations/tw/lessons/4-ComputerVision/07-ConvNets/lab/README.md b/translations/tw/lessons/4-ComputerVision/07-ConvNets/lab/README.md index 50b9a447..aaf86196 100644 --- a/translations/tw/lessons/4-ComputerVision/07-ConvNets/lab/README.md +++ b/translations/tw/lessons/4-ComputerVision/07-ConvNets/lab/README.md @@ -21,7 +21,7 @@ CO_OP_TRANSLATOR_METADATA: 我們將使用 [Oxford-IIIT Pet Dataset](https://www.robots.ox.ac.uk/~vgg/data/pets/)，該數據集包含 37 種不同品種的狗和貓的圖片。 -![我們將處理的數據集](../../../../../../translated_images/tw/data.50b2a9d5484bdbf0.webp) +![我們將處理的數據集](../../../../../../translated_images/zh-TW/data.50b2a9d5484bdbf0.webp) 要下載數據集，請使用以下程式碼片段： diff --git a/translations/tw/lessons/4-ComputerVision/08-TransferLearning/README.md b/translations/tw/lessons/4-ComputerVision/08-TransferLearning/README.md index 5711ed83..610076ba 100644 --- a/translations/tw/lessons/4-ComputerVision/08-TransferLearning/README.md +++ b/translations/tw/lessons/4-ComputerVision/08-TransferLearning/README.md @@ -29,7 +29,7 @@ Keras 和 PyTorch 都包含了方便的函數，可以輕鬆加載一些常見 以下是 VGG-16 網路從一張貓的圖片中提取的特徵示例： -![VGG-16 提取的特徵](../../../../../translated_images/tw/features.6291f9c7ba3a0b95.webp) +![VGG-16 提取的特徵](../../../../../translated_images/zh-TW/features.6291f9c7ba3a0b95.webp) ## 貓與狗數據集 @@ -48,19 +48,19 @@ Keras 和 PyTorch 都包含了方便的函數，可以輕鬆加載一些常見 我們可以採取的一種方法是從一張隨機圖像開始，然後嘗試使用**梯度下降優化**技術調整該圖像，使得網路認為它是一隻貓。 -![圖像優化循環](../../../../../translated_images/tw/ideal-cat-loop.999fbb8ff306e044.webp) +![圖像優化循環](../../../../../translated_images/zh-TW/ideal-cat-loop.999fbb8ff306e044.webp) 然而，如果我們這樣做，我們會得到一些非常接近隨機噪聲的東西。這是因為*有很多方法可以讓網路認為輸入圖像是一隻貓*，其中一些方法在視覺上並不合理。雖然這些圖像包含了許多典型於貓的模式，但並沒有任何約束使它們在視覺上具有辨識度。 為了改善結果，我們可以在損失函數中添加另一個項，稱為**變異損失**。這是一種衡量圖像中相鄰像素相似程度的指標。最小化變異損失可以使圖像更平滑，並消除噪聲——從而揭示出更具視覺吸引力的模式。以下是一些這樣的「理想」圖像的例子，它們被高概率地分類為貓和斑馬： -![理想貓](../../../../../translated_images/tw/ideal-cat.203dd4597643d6b0.webp) | ![理想斑馬](../../../../../translated_images/tw/ideal-zebra.7f70e8b54ee15a7a.webp) +![理想貓](../../../../../translated_images/zh-TW/ideal-cat.203dd4597643d6b0.webp) | ![理想斑馬](../../../../../translated_images/zh-TW/ideal-zebra.7f70e8b54ee15a7a.webp) -----|----- *理想貓* | *理想斑馬* 類似的方法可以用來對神經網路進行所謂的**對抗性攻擊**。假設我們想要欺騙神經網路，讓一隻狗看起來像一隻貓。如果我們拿一張狗的圖片，該圖片被網路識別為狗，然後稍微調整它，使用梯度下降優化，直到網路開始將其分類為貓： -![狗的圖片](../../../../../translated_images/tw/original-dog.8f68a67d2fe0911f.webp) | ![被分類為貓的狗圖片](../../../../../translated_images/tw/adversarial-dog.d9fc7773b0142b89.webp) +![狗的圖片](../../../../../translated_images/zh-TW/original-dog.8f68a67d2fe0911f.webp) | ![被分類為貓的狗圖片](../../../../../translated_images/zh-TW/adversarial-dog.d9fc7773b0142b89.webp) -----|----- *原始狗圖片* | *被分類為貓的狗圖片* diff --git a/translations/tw/lessons/4-ComputerVision/09-Autoencoders/README.md b/translations/tw/lessons/4-ComputerVision/09-Autoencoders/README.md index 98fbd94c..9c9fe8f1 100644 --- a/translations/tw/lessons/4-ComputerVision/09-Autoencoders/README.md +++ b/translations/tw/lessons/4-ComputerVision/09-Autoencoders/README.md @@ -19,7 +19,7 @@ CO_OP_TRANSLATOR_METADATA: 由於我們訓練自動編碼器以捕捉原始圖像中的盡可能多的信息以進行準確重建，網絡會嘗試找到最佳的**嵌入**來捕捉輸入圖像的含義。 -![自動編碼器示意圖](../../../../../translated_images/tw/autoencoder_schema.5e6fc9ad98a5eb61.webp) +![自動編碼器示意圖](../../../../../translated_images/zh-TW/autoencoder_schema.5e6fc9ad98a5eb61.webp) > 圖片來源：[Keras 博客](https://blog.keras.io/building-autoencoders-in-keras.html) diff --git a/translations/tw/lessons/4-ComputerVision/11-ObjectDetection/README.md b/translations/tw/lessons/4-ComputerVision/11-ObjectDetection/README.md index fbbf96d0..345eea8a 100644 --- a/translations/tw/lessons/4-ComputerVision/11-ObjectDetection/README.md +++ b/translations/tw/lessons/4-ComputerVision/11-ObjectDetection/README.md @@ -13,7 +13,7 @@ CO_OP_TRANSLATOR_METADATA: ## [課前測驗](https://ff-quizzes.netlify.app/en/ai/quiz/21) -![物件偵測](../../../../../translated_images/tw/Screen_Shot_2016-11-17_at_11.14.54_AM.b4bb3769353287be.webp) +![物件偵測](../../../../../translated_images/zh-TW/Screen_Shot_2016-11-17_at_11.14.54_AM.b4bb3769353287be.webp) > 圖片來源：[YOLO v2 網站](https://pjreddie.com/darknet/yolov2/) @@ -25,7 +25,7 @@ CO_OP_TRANSLATOR_METADATA: 2. 對每個區塊進行影像分類。 3. 對於分類結果有足夠高信心的區塊，可以認為包含目標物件。 -![簡單物件偵測](../../../../../translated_images/tw/naive-detection.e7f1ba220ccd08c6.webp) +![簡單物件偵測](../../../../../translated_images/zh-TW/naive-detection.e7f1ba220ccd08c6.webp) > *圖片來源：[練習筆記本](ObjectDetection-TF.ipynb)* @@ -42,7 +42,7 @@ CO_OP_TRANSLATOR_METADATA: * [PASCAL VOC](http://host.robots.ox.ac.uk/pascal/VOC/) - 包含 20 個類別 * [COCO](http://cocodataset.org/#home) - 常見物件的上下文。包含 80 個類別、邊界框和分割遮罩 -![COCO](../../../../../translated_images/tw/coco-examples.71bc60380fa6cceb.webp) +![COCO](../../../../../translated_images/zh-TW/coco-examples.71bc60380fa6cceb.webp) ## 物件偵測的評估指標 @@ -50,7 +50,7 @@ CO_OP_TRANSLATOR_METADATA: 在影像分類中，衡量算法表現相對簡單；但在物件偵測中，我們需要同時衡量類別的正確性以及推測邊界框位置的精確性。對於後者，我們使用所謂的**交集比聯集** (IoU)，它衡量兩個框（或任意兩個區域）的重疊程度。 -![IoU](../../../../../translated_images/tw/iou_equation.9a4751d40fff4e11.webp) +![IoU](../../../../../translated_images/zh-TW/iou_equation.9a4751d40fff4e11.webp) > *圖片來源：[這篇優秀的 IoU 部落格文章](https://pyimagesearch.com/2016/11/07/intersection-over-union-iou-for-object-detection/)* @@ -98,11 +98,11 @@ $$ [R-CNN](http://islab.ulsan.ac.kr/files/announcement/513/rcnn_pami.pdf) 使用[選擇性搜索](http://www.huppelen.nl/publications/selectiveSearchDraft.pdf)生成 ROI 區域的層次結構，然後通過 CNN 特徵提取器和 SVM 分類器來確定物件類別，並通過線性回歸確定*邊界框*座標。[官方論文](https://arxiv.org/pdf/1506.01497v1.pdf) -![RCNN](../../../../../translated_images/tw/rcnn1.cae407020dfb1d1f.webp) +![RCNN](../../../../../translated_images/zh-TW/rcnn1.cae407020dfb1d1f.webp) > *圖片來源：van de Sande et al. ICCV’11* -![RCNN-1](../../../../../translated_images/tw/rcnn2.2d9530bb83516484.webp) +![RCNN-1](../../../../../translated_images/zh-TW/rcnn2.2d9530bb83516484.webp) > *圖片來源：[這篇部落格](https://towardsdatascience.com/r-cnn-fast-r-cnn-faster-r-cnn-yolo-object-detection-algorithms-36d53571365e)* @@ -110,7 +110,7 @@ $$ 這種方法與 R-CNN 類似，但區域是在卷積層應用後定義的。 -![FRCNN](../../../../../translated_images/tw/f-rcnn.3cda6d9bb4188875.webp) +![FRCNN](../../../../../translated_images/zh-TW/f-rcnn.3cda6d9bb4188875.webp) > 圖片來源：[官方論文](https://www.cv-foundation.org/openaccess/content_iccv_2015/papers/Girshick_Fast_R-CNN_ICCV_2015_paper.pdf)，[arXiv](https://arxiv.org/pdf/1504.08083.pdf)，2015 @@ -118,7 +118,7 @@ $$ 這種方法的主要思想是使用神經網路來預測 ROI，即所謂的*區域提案網路* (Region Proposal Network)。[論文](https://arxiv.org/pdf/1506.01497.pdf)，2016 -![FasterRCNN](../../../../../translated_images/tw/faster-rcnn.8d46c099b87ef30a.webp) +![FasterRCNN](../../../../../translated_images/zh-TW/faster-rcnn.8d46c099b87ef30a.webp) > 圖片來源：[官方論文](https://arxiv.org/pdf/1506.01497.pdf) @@ -130,7 +130,7 @@ $$ 2. 特徵通過**位置敏感分數圖** (Position-Sensitive Score Map) 處理。每個類別 $C$ 的物件被分成 $k\times k$ 區域，我們訓練模型來預測物件的部分。 3. 對於 $k\times k$ 區域中的每個部分，所有網路對物件類別進行投票，選擇投票最多的物件類別。 -![r-fcn 圖片](../../../../../translated_images/tw/r-fcn.13eb88158b99a3da.webp) +![r-fcn 圖片](../../../../../translated_images/zh-TW/r-fcn.13eb88158b99a3da.webp) > 圖片來源：[官方論文](https://arxiv.org/abs/1605.06409) @@ -141,7 +141,7 @@ YOLO 是一種即時的一次通過算法。主要思想如下： * 將圖片分成 $S\times S$ 區域。 * 對每個區域，**CNN** 預測 $n$ 個可能的物件、*邊界框*座標以及*信心值*=*概率* * IoU。 - ![YOLO](../../../../../translated_images/tw/yolo.a2648ec82ee8bb4e.webp) + ![YOLO](../../../../../translated_images/zh-TW/yolo.a2648ec82ee8bb4e.webp) > 圖片來源：[官方論文](https://arxiv.org/abs/1506.02640) diff --git a/translations/tw/lessons/4-ComputerVision/README.md b/translations/tw/lessons/4-ComputerVision/README.md index 0e954f53..732b40ba 100644 --- a/translations/tw/lessons/4-ComputerVision/README.md +++ b/translations/tw/lessons/4-ComputerVision/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 電腦視覺 -![電腦視覺內容摘要手繪圖](../../../../translated_images/tw/ai-computervision.6506ebebac3fbf76.webp) +![電腦視覺內容摘要手繪圖](../../../../translated_images/zh-TW/ai-computervision.6506ebebac3fbf76.webp) 在本節中，我們將學習以下內容： diff --git a/translations/tw/lessons/5-NLP/14-Embeddings/README.md b/translations/tw/lessons/5-NLP/14-Embeddings/README.md index e5bb69db..67be1a2d 100644 --- a/translations/tw/lessons/5-NLP/14-Embeddings/README.md +++ b/translations/tw/lessons/5-NLP/14-Embeddings/README.md @@ -19,7 +19,7 @@ CO_OP_TRANSLATOR_METADATA: 通過在分類器網絡中使用嵌入層作為第一層，我們可以從詞袋模型切換到 **嵌入袋** 模型。在嵌入袋模型中，我們首先將文本中的每個詞轉換為相應的嵌入，然後對所有嵌入計算某種聚合函數，例如 `sum`、`average` 或 `max`。 -![展示五個序列詞的嵌入分類器的圖片。](../../../../../translated_images/tw/embedding-classifier-example.b77f021a7ee67eee.webp) +![展示五個序列詞的嵌入分類器的圖片。](../../../../../translated_images/zh-TW/embedding-classifier-example.b77f021a7ee67eee.webp) > 圖片由作者提供 @@ -40,7 +40,7 @@ CO_OP_TRANSLATOR_METADATA: CBoW 訓練速度更快，而 Skip-Gram 雖然較慢，但在表示不常見詞方面效果更好。 -![展示 CBoW 和 Skip-Gram 算法如何將詞轉換為向量的圖片。](../../../../../translated_images/tw/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) +![展示 CBoW 和 Skip-Gram 算法如何將詞轉換為向量的圖片。](../../../../../translated_images/zh-TW/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) > 圖片來源：[這篇論文](https://arxiv.org/pdf/1301.3781.pdf) diff --git a/translations/tw/lessons/5-NLP/15-LanguageModeling/README.md b/translations/tw/lessons/5-NLP/15-LanguageModeling/README.md index 20d70706..a3e03291 100644 --- a/translations/tw/lessons/5-NLP/15-LanguageModeling/README.md +++ b/translations/tw/lessons/5-NLP/15-LanguageModeling/README.md @@ -23,7 +23,7 @@ CO_OP_TRANSLATOR_METADATA: * **連續詞袋模型** (CBoW)，在詞元序列 $W_{-N}$, ..., $W_N$ 中預測中間的詞元 $W_0$。 * **Skip-gram**，通過中間詞元 $W_0$ 預測一組相鄰的詞元 {$W_{-N},\dots, W_{-1}, W_1,\dots, W_N$}。 -![來自論文的將詞轉換為向量的算法示例](../../../../../translated_images/tw/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) +![來自論文的將詞轉換為向量的算法示例](../../../../../translated_images/zh-TW/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) > 圖片來源：[這篇論文](https://arxiv.org/pdf/1301.3781.pdf) diff --git a/translations/tw/lessons/5-NLP/16-RNN/README.md b/translations/tw/lessons/5-NLP/16-RNN/README.md index 7cf5b6f9..aa0537f3 100644 --- a/translations/tw/lessons/5-NLP/16-RNN/README.md +++ b/translations/tw/lessons/5-NLP/16-RNN/README.md @@ -15,7 +15,7 @@ CO_OP_TRANSLATOR_METADATA: 為了捕捉文本序列的意義，我們需要使用另一種神經網絡架構，稱為**循環神經網絡**（Recurrent Neural Network，簡稱 RNN）。在 RNN 中，我們將句子逐個符號地傳遞給網絡，網絡會生成某種**狀態**，然後將該狀態與下一個符號一起再次傳遞給網絡。 -![RNN](../../../../../translated_images/tw/rnn.27f5c29c53d727b5.webp) +![RNN](../../../../../translated_images/zh-TW/rnn.27f5c29c53d727b5.webp) > 圖片由作者提供 @@ -61,7 +61,7 @@ LSTM 網絡的組織方式與 RNN 類似，但有兩個狀態會從層到層傳 循環網絡（無論是單向還是雙向）能夠捕捉序列中的某些模式，並將它們存儲到狀態向量中或傳遞到輸出中。與卷積網絡類似，我們可以在第一層之上構建另一個循環層，以捕捉更高層次的模式，並基於第一層提取的低層次模式進行構建。這引出了**多層 RNN** 的概念，它由兩個或更多循環網絡組成，其中前一層的輸出作為輸入傳遞到下一層。 -![顯示多層長短期記憶 RNN 的圖片](../../../../../translated_images/tw/multi-layer-lstm.dd975e29bb2a59fe.webp) +![顯示多層長短期記憶 RNN 的圖片](../../../../../translated_images/zh-TW/multi-layer-lstm.dd975e29bb2a59fe.webp) *圖片來自 Fernando López 的[這篇精彩文章](https://towardsdatascience.com/from-a-lstm-cell-to-a-multilayer-lstm-network-with-pytorch-2899eb5696f3)* diff --git a/translations/tw/lessons/5-NLP/17-GenerativeNetworks/README.md b/translations/tw/lessons/5-NLP/17-GenerativeNetworks/README.md index c35412cf..4033398e 100644 --- a/translations/tw/lessons/5-NLP/17-GenerativeNetworks/README.md +++ b/translations/tw/lessons/5-NLP/17-GenerativeNetworks/README.md @@ -19,7 +19,7 @@ CO_OP_TRANSLATOR_METADATA: 這使得不同的神經網絡架構成為可能，如下圖所示： -![展示常見循環神經網絡模式的圖片。](../../../../../translated_images/tw/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) +![展示常見循環神經網絡模式的圖片。](../../../../../translated_images/zh-TW/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) > 圖片來源：[Andrej Karpaty](http://karpathy.github.io/) 的博客文章 [Unreasonable Effectiveness of Recurrent Neural Networks](http://karpathy.github.io/2015/05/21/rnn-effectiveness/) @@ -32,7 +32,7 @@ CO_OP_TRANSLATOR_METADATA: 我們將訓練這個 RNN 逐步生成文本。在每一步中，我們將取一個長度為 `nchars` 的字符序列，並要求網絡為每個輸入字符生成下一個輸出字符： -![展示 RNN 生成單詞 'HELLO' 的示例圖片。](../../../../../translated_images/tw/rnn-generate.56c54afb52f9781d.webp) +![展示 RNN 生成單詞 'HELLO' 的示例圖片。](../../../../../translated_images/zh-TW/rnn-generate.56c54afb52f9781d.webp) 在生成文本（推理過程中），我們從某個**提示**開始，將其通過 RNN 單元生成中間狀態，然後從該狀態開始生成。我們一次生成一個字符，並將狀態和生成的字符傳遞給另一個 RNN 單元以生成下一個字符，直到生成足夠的字符。 diff --git a/translations/tw/lessons/5-NLP/18-Transformers/README.md b/translations/tw/lessons/5-NLP/18-Transformers/README.md index 7356902e..81ce23ae 100644 --- a/translations/tw/lessons/5-NLP/18-Transformers/README.md +++ b/translations/tw/lessons/5-NLP/18-Transformers/README.md @@ -20,13 +20,13 @@ CO_OP_TRANSLATOR_METADATA: **注意機制**提供了一種方法，能夠對每個輸入向量對RNN輸出預測的上下文影響進行加權。其實現方式是通過在輸入RNN的中間狀態和輸出RNN之間建立捷徑。在生成輸出符號yt時，我們會考慮所有輸入隱藏狀態hi，並賦予不同的權重係數αt,i。 -![顯示具有加性注意層的編碼器/解碼器模型的圖片](../../../../../translated_images/tw/encoder-decoder-attention.7a726296894fb567.webp) +![顯示具有加性注意層的編碼器/解碼器模型的圖片](../../../../../translated_images/zh-TW/encoder-decoder-attention.7a726296894fb567.webp) > [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf)中的加性注意機制編碼器-解碼器模型，引用自[這篇博客文章](https://lilianweng.github.io/lil-log/2018/06/24/attention-attention.html) 注意矩陣{αi,j}表示某些輸入詞在生成輸出序列中的某個詞時所起的作用程度。以下是一個這樣的矩陣示例： -![顯示由RNNsearch-50找到的樣本對齊的圖片，取自Bahdanau - arviz.org](../../../../../translated_images/tw/bahdanau-fig3.09ba2d37f202a6af.webp) +![顯示由RNNsearch-50找到的樣本對齊的圖片，取自Bahdanau - arviz.org](../../../../../translated_images/zh-TW/bahdanau-fig3.09ba2d37f202a6af.webp) > 圖片來自[Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf)（圖3） @@ -66,7 +66,7 @@ Transformer的主要理念之一是避免RNN的序列特性，並創建一個在 接下來，我們需要捕捉序列中的一些模式。為此，Transformer使用了**自注意機制**，這本質上是將注意機制應用於相同的輸入和輸出序列。應用自注意機制使我們能夠考慮句子中的**上下文**，並查看哪些詞是相互關聯的。例如，它使我們能夠看到哪些詞是由指代詞（如*它*）指代的，並且能夠考慮上下文： -![](../../../../../translated_images/tw/CoreferenceResolution.861924d6d384a7d6.webp) +![](../../../../../translated_images/zh-TW/CoreferenceResolution.861924d6d384a7d6.webp) > 圖片來自[Google的博客](https://research.googleblog.com/2017/08/transformer-novel-neural-network.html) @@ -91,7 +91,7 @@ Transformer的主要理念之一是避免RNN的序列特性，並創建一個在 **BERT**（Bidirectional Encoder Representations from Transformers）是一個非常大的多層Transformer網絡，*BERT-base*有12層，*BERT-large*有24層。該模型首先在大規模文本數據（維基百科+書籍）上進行無監督訓練（預測句子中的被遮蔽詞）。在預訓練過程中，模型吸收了大量的語言理解能力，這些能力可以通過微調其他數據集來利用。這個過程稱為**遷移學習**。 -![圖片來自http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/tw/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp) +![圖片來自http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/zh-TW/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp) > 圖片[來源](http://jalammar.github.io/illustrated-bert/) diff --git a/translations/tw/lessons/5-NLP/19-NER/README.md b/translations/tw/lessons/5-NLP/19-NER/README.md index 7f02ce0f..4434eeef 100644 --- a/translations/tw/lessons/5-NLP/19-NER/README.md +++ b/translations/tw/lessons/5-NLP/19-NER/README.md @@ -58,7 +58,7 @@ NER 模型本質上是 **標記分類模型**，因為對於每個輸入的標 由於我們需要在標記和類別之間建立一對一的對應關係，我們可以從這張圖中訓練一個右側的 **多對多** 神經網絡模型： -![顯示常見循環神經網絡模式的圖片。](../../../../../translated_images/tw/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) +![顯示常見循環神經網絡模式的圖片。](../../../../../translated_images/zh-TW/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) > *圖片來自 [這篇部落格文章](http://karpathy.github.io/2015/05/21/rnn-effectiveness/)，作者為 [Andrej Karpathy](http://karpathy.github.io/)。NER 標記分類模型對應於圖片中最右側的網絡架構。* diff --git a/translations/tw/lessons/5-NLP/README.md b/translations/tw/lessons/5-NLP/README.md index 725aacc0..6dbde006 100644 --- a/translations/tw/lessons/5-NLP/README.md +++ b/translations/tw/lessons/5-NLP/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 自然語言處理 -![NLP任務摘要手繪圖](../../../../translated_images/tw/ai-nlp.b22dcb8ca4707cea.webp) +![NLP任務摘要手繪圖](../../../../translated_images/zh-TW/ai-nlp.b22dcb8ca4707cea.webp) 在本章節中，我們將專注於使用神經網絡來處理與**自然語言處理 (NLP)**相關的任務。我們希望電腦能夠解決許多NLP問題： diff --git a/translations/tw/lessons/6-Other/23-MultiagentSystems/README.md b/translations/tw/lessons/6-Other/23-MultiagentSystems/README.md index 5d1a41ef..15e818d0 100644 --- a/translations/tw/lessons/6-Other/23-MultiagentSystems/README.md +++ b/translations/tw/lessons/6-Other/23-MultiagentSystems/README.md @@ -70,7 +70,7 @@ NetLogo的一大優點是它包含一個可供試用的工作模型庫。進入* 打開模型後，你會進入NetLogo的主界面。以下是一個描述狼和羊的種群模型，給定有限資源（草地）。 -![NetLogo主界面](../../../../../translated_images/tw/NetLogo-Main.32653711ec1a01b3.webp) +![NetLogo主界面](../../../../../translated_images/zh-TW/NetLogo-Main.32653711ec1a01b3.webp) > Dmitry Soshnikov提供的截圖 diff --git a/translations/tw/lessons/README.md b/translations/tw/lessons/README.md index 7e5e0e55..54e6c30d 100644 --- a/translations/tw/lessons/README.md +++ b/translations/tw/lessons/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 概覽 -![概覽手繪圖](../../../translated_images/tw/ai-overview.0857791951d19500.webp) +![概覽手繪圖](../../../translated_images/zh-TW/ai-overview.0857791951d19500.webp) > 手繪筆記由 [Tomomi Imura](https://twitter.com/girlie_mac) 提供 diff --git a/translations/tw/lessons/X-Extras/X1-MultiModal/README.md b/translations/tw/lessons/X-Extras/X1-MultiModal/README.md index 6ddc6450..d527d72a 100644 --- a/translations/tw/lessons/X-Extras/X1-MultiModal/README.md +++ b/translations/tw/lessons/X-Extras/X1-MultiModal/README.md @@ -15,7 +15,7 @@ CO_OP_TRANSLATOR_METADATA: CLIP 的主要理念是能夠比較文本提示與圖像，並確定圖像與提示的匹配程度。 -![CLIP 架構](../../../../../translated_images/tw/clip-arch.b3dbf20b4e8ed8be.webp) +![CLIP 架構](../../../../../translated_images/zh-TW/clip-arch.b3dbf20b4e8ed8be.webp) > *圖片來自[這篇博客文章](https://openai.com/blog/clip/)* @@ -29,7 +29,7 @@ CLIP 模型/庫可以從 [OpenAI GitHub](https://github.com/openai/CLIP) 獲取 假設我們需要將圖像分類為貓、狗和人類。在這種情況下，我們可以給模型一張圖像，以及一系列文本提示：“*一張貓的圖片*”、“*一張狗的圖片*”、“*一張人類的圖片*”。在結果的 3 個概率向量中，我們只需選擇值最高的索引。 -![CLIP 用於圖像分類](../../../../../translated_images/tw/clip-class.3af42ef0b2b19369.webp) +![CLIP 用於圖像分類](../../../../../translated_images/zh-TW/clip-class.3af42ef0b2b19369.webp) > *圖片來自[這篇博客文章](https://openai.com/blog/clip/)* @@ -53,13 +53,13 @@ VQGAN 與普通 [GAN](../../4-ComputerVision/10-GANs/README.md) 的主要區別 VQGAN 與傳統 GAN 的一個重要區別在於，後者可以從任何輸入向量生成一張像樣的圖像，而 VQGAN 可能生成一張不連貫的圖像。因此，我們需要進一步引導圖像創建過程，而這可以通過 CLIP 完成。 -![VQGAN+CLIP 架構](../../../../../translated_images/tw/vqgan.5027fe05051dfa31.webp) +![VQGAN+CLIP 架構](../../../../../translated_images/zh-TW/vqgan.5027fe05051dfa31.webp) 為了生成與文本提示相對應的圖像，我們從一些隨機編碼向量開始，將其傳遞給 VQGAN 以生成圖像。然後使用 CLIP 生成一個損失函數，該函數顯示圖像與文本提示的匹配程度。目標是最小化這個損失，通過反向傳播調整輸入向量參數。 一個實現 VQGAN+CLIP 的優秀庫是 [Pixray](http://github.com/pixray/pixray) -![由 Pixray 生成的圖片](../../../../../translated_images/tw/a_closeup_watercolor_portrait_of_young_male_teacher_of_literature_with_a_book.2384968e9db8a0d0.webp) | ![由 Pixray 生成的圖片](../../../../../translated_images/tw/a_closeup_oil_portrait_of_young_female_teacher_of_computer_science_with_a_computer.e0b6495f210a4390.webp) | ![由 Pixray 生成的圖片](../../../../../translated_images/tw/a_closeup_oil_portrait_of_old_male_teacher_of_math.5362e67aa7fc2683.webp) +![由 Pixray 生成的圖片](../../../../../translated_images/zh-TW/a_closeup_watercolor_portrait_of_young_male_teacher_of_literature_with_a_book.2384968e9db8a0d0.webp) | ![由 Pixray 生成的圖片](../../../../../translated_images/zh-TW/a_closeup_oil_portrait_of_young_female_teacher_of_computer_science_with_a_computer.e0b6495f210a4390.webp) | ![由 Pixray 生成的圖片](../../../../../translated_images/zh-TW/a_closeup_oil_portrait_of_old_male_teacher_of_math.5362e67aa7fc2683.webp) ----|----|---- 由提示 *一幅年輕男性文學教師手持書本的水彩特寫肖像* 生成的圖片 | 由提示 *一幅年輕女性計算機科學教師手持電腦的油畫特寫肖像* 生成的圖片 | 由提示 *一幅年長男性數學教師站在黑板前的油畫特寫肖像* 生成的圖片 @@ -75,7 +75,7 @@ DALL-E 是一個基於 GPT-3 的版本，專門用於從提示生成圖像。它 DALL-E 1 和 2 的主要區別在於，DALL-E 2 能夠生成更真實的圖像和藝術作品。 使用 DALL-E 生成圖像的範例： -![由 DALL-E 生成的圖片](../../../../../translated_images/tw/DALL·E%202023-06-20%2015.56.56%20-%20a%20closeup%20watercolor%20portrait%20of%20young%20male%20teacher%20of%20literature%20with%20a%20book.6c235e8271d9ed10ce985d86aeb241a58518958647973af136912116b9518fce.png) | ![由 DALL-E 生成的圖片](../../../../../translated_images/tw/DALL·E%202023-06-20%2015.57.43%20-%20a%20closeup%20oil%20portrait%20of%20young%20female%20teacher%20of%20computer%20science%20with%20a%20computer.f21dc4166340b6c8b4d1cb57efd1e22127407f9b28c9ac7afe11344065369e64.png) | ![由 DALL-E 生成的圖片](../../../../../translated_images/tw/DALL·E%202023-06-20%2015.58.42%20-%20%20a%20closeup%20oil%20portrait%20of%20old%20male%20teacher%20of%20mathematics%20in%20front%20of%20blackboard.d331c2dfbdc3f7c46aa65c0809066f5e7ed4b49609cd259852e760df21051e4a.png) +![由 DALL-E 生成的圖片](../../../../../translated_images/zh-TW/DALL·E%202023-06-20%2015.56.56%20-%20a%20closeup%20watercolor%20portrait%20of%20young%20male%20teacher%20of%20literature%20with%20a%20book.6c235e8271d9ed10ce985d86aeb241a58518958647973af136912116b9518fce.png) | ![由 DALL-E 生成的圖片](../../../../../translated_images/zh-TW/DALL·E%202023-06-20%2015.57.43%20-%20a%20closeup%20oil%20portrait%20of%20young%20female%20teacher%20of%20computer%20science%20with%20a%20computer.f21dc4166340b6c8b4d1cb57efd1e22127407f9b28c9ac7afe11344065369e64.png) | ![由 DALL-E 生成的圖片](../../../../../translated_images/zh-TW/DALL·E%202023-06-20%2015.58.42%20-%20%20a%20closeup%20oil%20portrait%20of%20old%20male%20teacher%20of%20mathematics%20in%20front%20of%20blackboard.d331c2dfbdc3f7c46aa65c0809066f5e7ed4b49609cd259852e760df21051e4a.png) ----|----|---- 由提示 *一幅年輕男性文學教師手持書本的水彩特寫肖像* 生成的圖片 | 由提示 *一幅年輕女性計算機科學教師手持電腦的油畫特寫肖像* 生成的圖片 | 由提示 *一幅年長男性數學教師站在黑板前的油畫特寫肖像* 生成的圖片 diff --git a/translations/zh/README.md b/translations/zh/README.md index d5fc7216..e51930a4 100644 --- a/translations/zh/README.md +++ b/translations/zh/README.md @@ -23,7 +23,7 @@ CO_OP_TRANSLATOR_METADATA: # 初学者人工智能课程 -|![Sketchnote by @girlie_mac https://twitter.com/girlie_mac](../../../../translated_images/zh/ai-overview.0857791951d19500.webp)| +|![Sketchnote by @girlie_mac https://twitter.com/girlie_mac](../../../../translated_images/zh-CN/ai-overview.0857791951d19500.webp)| |:---:| | 初学者人工智能 - _由[@girlie_mac](https://twitter.com/girlie_mac) 绘制的手绘笔记_ | diff --git a/translations/zh/lessons/1-Intro/README.md b/translations/zh/lessons/1-Intro/README.md index 241337e1..288f1b29 100644 --- a/translations/zh/lessons/1-Intro/README.md +++ b/translations/zh/lessons/1-Intro/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 人工智能简介 -![人工智能内容简介的涂鸦](../../../../translated_images/zh/ai-intro.bf28d1ac4235881c.webp) +![人工智能内容简介的涂鸦](../../../../translated_images/zh-CN/ai-intro.bf28d1ac4235881c.webp) > 涂鸦作者：[Tomomi Imura](https://twitter.com/girlie_mac) @@ -19,7 +19,7 @@ CO_OP_TRANSLATOR_METADATA: 最初，计算机是由[查尔斯·巴贝奇](https://en.wikipedia.org/wiki/Charles_Babbage)发明的，用于按照明确的程序（算法）对数字进行操作。尽管现代计算机比19世纪提出的原始模型先进得多，但它们仍然遵循受控计算的基本理念。因此，如果我们知道实现目标所需的确切步骤序列，就可以编程让计算机完成某项任务。 -![一个人的照片](../../../../translated_images/zh/dsh_age.d212a30d4e54fb5f.webp) +![一个人的照片](../../../../translated_images/zh-CN/dsh_age.d212a30d4e54fb5f.webp) > 照片由 [Vickie Soshnikova](http://twitter.com/vickievalerie) 提供 @@ -46,7 +46,7 @@ CO_OP_TRANSLATOR_METADATA: 在讨论**[智能](https://en.wikipedia.org/wiki/Intelligence)**这个术语时，一个问题是我们并没有对其明确的定义。有人认为智能与**抽象思维**或**自我意识**相关，但我们无法准确定义它。 -![一只猫的照片](../../../../translated_images/zh/photo-cat.8c8e8fb760ffe457.webp) +![一只猫的照片](../../../../translated_images/zh-CN/photo-cat.8c8e8fb760ffe457.webp) > [照片](https://unsplash.com/photos/75715CVEJhI)由 [Amber Kipp](https://unsplash.com/@sadmax) 提供，来自Unsplash @@ -98,13 +98,13 @@ CO_OP_TRANSLATOR_METADATA: > | 那机器学习呢？ | | > |--------------|-----------| -> | 人工智能的一部分是基于计算机通过一些数据学习解决问题的，这被称为**机器学习**。我们不会在本课程中讨论经典的机器学习——请参考单独的[机器学习初学者课程](http://aka.ms/ml-beginners)。 | ![机器学习初学者](../../../../translated_images/zh/ml-for-beginners.9e4fed176fd5817d.webp) | +> | 人工智能的一部分是基于计算机通过一些数据学习解决问题的，这被称为**机器学习**。我们不会在本课程中讨论经典的机器学习——请参考单独的[机器学习初学者课程](http://aka.ms/ml-beginners)。 | ![机器学习初学者](../../../../translated_images/zh-CN/ml-for-beginners.9e4fed176fd5817d.webp) | ## 人工智能简史 人工智能作为一个领域始于20世纪中期。最初，符号推理是一种流行的方法，并取得了一些重要的成功，例如专家系统——能够在某些有限问题领域中充当专家的计算机程序。然而，很快就发现这种方法并不具有良好的扩展性。从专家那里提取知识、在计算机中表示这些知识并保持知识库的准确性，结果证明是一项非常复杂且在许多情况下成本过高的任务。这导致了20世纪70年代所谓的[人工智能寒冬](https://en.wikipedia.org/wiki/AI_winter)。 -人工智能简史 +人工智能简史 > 图片由 [Dmitry Soshnikov](http://soshnikov.com) 提供 @@ -124,7 +124,7 @@ CO_OP_TRANSLATOR_METADATA: * 现代助手，如Cortana、Siri或Google Assistant，都是混合系统，使用神经网络将语音转换为文本并识别我们的意图，然后采用一些推理或明确的算法来执行所需的操作。 * 未来，我们可能会期待一个完全基于神经网络的模型能够自行处理对话。最近的GPT和[Turing-NLG](https://www.microsoft.com/research/blog/turing-nlg-a-17-billion-parameter-language-model-by-microsoft)系列神经网络在这方面表现出色。 -图灵测试的演变 +图灵测试的演变 > 图片由 Dmitry Soshnikov 提供，[照片](https://unsplash.com/photos/r8LmVbUKgns)由 [Marina Abrosimova](https://unsplash.com/@abrosimova_marina_foto) 提供，来源于 Unsplash ## 最近的人工智能研究 diff --git a/translations/zh/lessons/2-Symbolic/Animals.ipynb b/translations/zh/lessons/2-Symbolic/Animals.ipynb index b2aab4de..cd75d7a4 100644 --- a/translations/zh/lessons/2-Symbolic/Animals.ipynb +++ b/translations/zh/lessons/2-Symbolic/Animals.ipynb @@ -12,7 +12,7 @@ "\n", "在此示例中，我们将实现一个简单的基于知识的系统，根据一些物理特征来确定动物。该系统可以用以下 AND-OR 树表示（这是整个树的一部分，我们可以很容易地添加更多规则）：\n", "\n", - "![](../../../../../../translated_images/zh/AND-OR-Tree.5592d2c70187f283.webp)\n" + "![](../../../../../../translated_images/zh-CN/AND-OR-Tree.5592d2c70187f283.webp)\n" ] }, { diff --git a/translations/zh/lessons/2-Symbolic/README.md b/translations/zh/lessons/2-Symbolic/README.md index c5d769af..5a0e4acf 100644 --- a/translations/zh/lessons/2-Symbolic/README.md +++ b/translations/zh/lessons/2-Symbolic/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 知识表示与专家系统 -![符号人工智能内容摘要](../../../../../../translated_images/zh/ai-symbolic.715a30cb610411a6.webp) +![符号人工智能内容摘要](../../../../../../translated_images/zh-CN/ai-symbolic.715a30cb610411a6.webp) > 草图由 [Tomomi Imura](https://twitter.com/girlie_mac) 绘制 @@ -35,13 +35,13 @@ CO_OP_TRANSLATOR_METADATA: * **知识**是信息被整合进我们的世界模型。例如，一旦我们了解计算机是什么，我们便开始对它的工作原理、价格及用途有一些想法。这些相互关联的概念网络构成了我们的知识。 * **智慧**是我们对世界理解的更高层次，代表*元知识*，例如关于知识何时及如何使用的某种认知。 - + *图片来自[维基百科](https://commons.wikimedia.org/w/index.php?curid=37705247)，作者Longlivetheux，CC BY-SA 4.0* 因此，**知识表示**的问题是找到一种有效方式，将知识以数据形式表示于计算机内部，使其能被自动使用。这可以看作一个光谱： -![知识表示光谱](../../../../../../translated_images/zh/knowledge-spectrum.b60df631852c0217.webp) +![知识表示光谱](../../../../../../translated_images/zh-CN/knowledge-spectrum.b60df631852c0217.webp) > 图片来自 [Dmitry Soshnikov](http://soshnikov.com) @@ -94,7 +94,7 @@ Untyped-Language | doesn't have | type definitions 符号人工智能的早期成功之一是所谓的**专家系统**——设计用来在有限问题领域充当专家的计算机系统。它们基于从一个或多个专家处提取的**知识库**，并包含一个在其上执行推理的**推理引擎**。 -![人类结构](../../../../../../translated_images/zh/arch-human.5d4d35f1bba3ab1c.webp) | ![基于知识的系统](../../../../../../translated_images/zh/arch-kbs.3ec5c150b09fa8da.webp) +![人类结构](../../../../../../translated_images/zh-CN/arch-human.5d4d35f1bba3ab1c.webp) | ![基于知识的系统](../../../../../../translated_images/zh-CN/arch-kbs.3ec5c150b09fa8da.webp) ---------------------------------------------|------------------------------------------------ 人类神经系统简化结构 | 基于知识的系统架构 @@ -106,7 +106,7 @@ Untyped-Language | doesn't have | type definitions 例如，考虑一个基于动物物理特征决定动物种类的专家系统： -![与-或树](../../../../../../translated_images/zh/AND-OR-Tree.5592d2c70187f283.webp) +![与-或树](../../../../../../translated_images/zh-CN/AND-OR-Tree.5592d2c70187f283.webp) > 图片来自 [Dmitry Soshnikov](http://soshnikov.com) @@ -175,7 +175,7 @@ THEN the animal is a carnivore 在语义网中，所有表示均基于三元组。每个对象和每个关系都由 URI 唯一标识。例如，如果我们想陈述这个 AI 课程是由 Dmitry Soshnikov 于 2022 年 1 月 1 日开发的事实——我们可以使用以下三元组： - + ``` http://github.com/microsoft/ai-for-beginners http://www.example.com/terms/creation-date “Jan 1, 2022” @@ -186,7 +186,7 @@ http://github.com/microsoft/ai-for-beginners http://purl.org/dc/elements/1.1/cre 在更复杂的情况下，如果我们想定义一个创建者列表，可以使用 RDF 中定义的一些数据结构。 - + > 上图由 [Dmitry Soshnikov](http://soshnikov.com) 绘制 @@ -210,7 +210,7 @@ GROUP BY ?eyeColorLabel > ✅ 如果你想试验构建自己的本体，或者打开已有本体，有一个很棒的图形化本体编辑器叫做 [Protégé](https://protege.stanford.edu/)。下载它，或者在线使用。 - + *Web Protégé 编辑器打开了罗曼诺夫家族的本体。截屏来自 Dmitry Soshnikov* diff --git a/translations/zh/lessons/3-NeuralNetworks/05-Frameworks/README.md b/translations/zh/lessons/3-NeuralNetworks/05-Frameworks/README.md index be2d2c16..7bfab24b 100644 --- a/translations/zh/lessons/3-NeuralNetworks/05-Frameworks/README.md +++ b/translations/zh/lessons/3-NeuralNetworks/05-Frameworks/README.md @@ -58,7 +58,7 @@ CO_OP_TRANSLATOR_METADATA: 考虑以下拟合 5 个点的问题（图中的 `x` 表示点）： -![线性模型](../../../../../translated_images/zh/overfit1.f24b71c6f652e59e.webp) | ![过拟合模型](../../../../../translated_images/zh/overfit2.131f5800ae10ca5e.webp) +![线性模型](../../../../../translated_images/zh-CN/overfit1.f24b71c6f652e59e.webp) | ![过拟合模型](../../../../../translated_images/zh-CN/overfit2.131f5800ae10ca5e.webp) -------------------------|-------------------------- **线性模型，2 个参数** | **非线性模型，7 个参数** 训练误差 = 5.3 | 训练误差 = 0 @@ -79,7 +79,7 @@ CO_OP_TRANSLATOR_METADATA: 从上图可以看出，过拟合可以通过非常低的训练误差和非常高的验证误差来检测。通常在训练过程中，我们会看到训练误差和验证误差都开始下降，但在某个点之后，验证误差可能停止下降并开始上升。这是过拟合的信号，表明我们可能应该停止训练（或者至少保存模型的快照）。 -![过拟合](../../../../../translated_images/zh/Overfitting.408ad91cd90b4371.webp) +![过拟合](../../../../../translated_images/zh-CN/Overfitting.408ad91cd90b4371.webp) ## 如何防止过拟合 diff --git a/translations/zh/lessons/3-NeuralNetworks/README.md b/translations/zh/lessons/3-NeuralNetworks/README.md index 143f4176..d9e7d2af 100644 --- a/translations/zh/lessons/3-NeuralNetworks/README.md +++ b/translations/zh/lessons/3-NeuralNetworks/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 神经网络简介 -![神经网络内容总结涂鸦](../../../../translated_images/zh/ai-neuralnetworks.1c687ae40bc86e83.webp) +![神经网络内容总结涂鸦](../../../../translated_images/zh-CN/ai-neuralnetworks.1c687ae40bc86e83.webp) 正如我们在介绍中讨论的那样，实现智能的一种方法是训练一个**计算机模型**或一个**人工大脑**。自20世纪中期以来，研究人员尝试了各种数学模型，直到近年来，这一方向取得了巨大的成功。这些大脑的数学模型被称为**神经网络**。 @@ -36,13 +36,13 @@ CO_OP_TRANSLATOR_METADATA: 从生物学中我们知道，大脑由神经细胞（神经元）组成，每个神经元有多个“输入”（树突）和一个“输出”（轴突）。树突和轴突都可以传导电信号，它们之间的连接——称为突触——可以表现出不同程度的导电性，这种导电性由神经递质调节。 -![神经元模型](../../../../translated_images/zh/synapse-wikipedia.ed20a9e4726ea1c6.webp) | ![神经元模型](../../../../translated_images/zh/artneuron.1a5daa88d20ebe6f.webp) +![神经元模型](../../../../translated_images/zh-CN/synapse-wikipedia.ed20a9e4726ea1c6.webp) | ![神经元模型](../../../../translated_images/zh-CN/artneuron.1a5daa88d20ebe6f.webp) ----|---- 真实神经元 *（[图片](https://en.wikipedia.org/wiki/Synapse#/media/File:SynapseSchematic_lines.svg)来自维基百科）* | 人工神经元 *（作者提供图片）* 因此，神经元的最简单数学模型包含若干输入 X1, ..., XN 和一个输出 Y，以及一系列权重 W1, ..., WN。输出的计算公式为： -Y = f\left(\sum_{i=1}^N X_iW_i\right) +Y = f\left(\sum_{i=1}^N X_iW_i\right) 其中 f 是某种非线性的**激活函数**。 diff --git a/translations/zh/lessons/4-ComputerVision/06-IntroCV/README.md b/translations/zh/lessons/4-ComputerVision/06-IntroCV/README.md index aff8320c..a6075514 100644 --- a/translations/zh/lessons/4-ComputerVision/06-IntroCV/README.md +++ b/translations/zh/lessons/4-ComputerVision/06-IntroCV/README.md @@ -73,14 +73,14 @@ im = cv2.cvtColor(im,cv2.COLOR_BGR2RGB) * **对盲文书籍的照片进行预处理**。我们重点介绍如何使用阈值处理、特征检测、透视变换和NumPy操作来分离单个盲文符号，以便神经网络进一步分类。 -![盲文图像](../../../../../translated_images/zh/braille.341962ff76b1bd70.webp) | ![盲文图像预处理结果](../../../../../translated_images/zh/braille-result.46530fea020b03c7.webp) | ![盲文符号](../../../../../translated_images/zh/braille-symbols.0159185ab69d5339.webp) +![盲文图像](../../../../../translated_images/zh-CN/braille.341962ff76b1bd70.webp) | ![盲文图像预处理结果](../../../../../translated_images/zh-CN/braille-result.46530fea020b03c7.webp) | ![盲文符号](../../../../../translated_images/zh-CN/braille-symbols.0159185ab69d5339.webp) ----|-----|----- > 图片来源：[OpenCV.ipynb](OpenCV.ipynb) * **使用帧差检测视频中的运动**。如果摄像机是固定的，那么摄像机画面中的帧应该彼此非常相似。由于帧被表示为数组，只需对两个连续帧的数组进行减法运算，就可以得到像素差异，对于静态帧来说差异应该很小，而当图像中有显著运动时差异会变大。 -![视频帧和帧差异图像](../../../../../translated_images/zh/frame-difference.706f805491a0883c.webp) +![视频帧和帧差异图像](../../../../../translated_images/zh-CN/frame-difference.706f805491a0883c.webp) > 图片来源：[OpenCV.ipynb](OpenCV.ipynb) @@ -89,7 +89,7 @@ im = cv2.cvtColor(im,cv2.COLOR_BGR2RGB) - **密集光流**计算每个像素的运动向量场。 - **稀疏光流**基于图像中的一些显著特征（例如边缘），并从帧到帧构建它们的轨迹。 -![光流图像](../../../../../translated_images/zh/optical.1f4a94464579a83a.webp) +![光流图像](../../../../../translated_images/zh-CN/optical.1f4a94464579a83a.webp) > 图片来源：[OpenCV.ipynb](OpenCV.ipynb) diff --git a/translations/zh/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md b/translations/zh/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md index 5a1dc757..7cd0a8eb 100644 --- a/translations/zh/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md +++ b/translations/zh/lessons/4-ComputerVision/07-ConvNets/CNN_Architectures.md @@ -13,11 +13,11 @@ CO_OP_TRANSLATOR_METADATA: VGG-16是一种网络，在2014年的ImageNet top-5分类中达到了92.7%的准确率。它的层结构如下： -![ImageNet Layers](../../../../../translated_images/zh/vgg-16-arch1.d901a5583b3a51ba.webp) +![ImageNet Layers](../../../../../translated_images/zh-CN/vgg-16-arch1.d901a5583b3a51ba.webp) 如图所示，VGG采用了传统的金字塔架构，即一系列卷积-池化层的组合。 -![ImageNet Pyramid](../../../../../translated_images/zh/vgg-16-arch.64ff2137f50dd49f.webp) +![ImageNet Pyramid](../../../../../translated_images/zh-CN/vgg-16-arch.64ff2137f50dd49f.webp) > 图片来源：[Researchgate](https://www.researchgate.net/figure/Vgg16-model-structure-To-get-the-VGG-NIN-model-we-replace-the-2-nd-4-th-6-th-7-th_fig2_335194493) diff --git a/translations/zh/lessons/4-ComputerVision/07-ConvNets/README.md b/translations/zh/lessons/4-ComputerVision/07-ConvNets/README.md index 1bf375f6..e8d56a03 100644 --- a/translations/zh/lessons/4-ComputerVision/07-ConvNets/README.md +++ b/translations/zh/lessons/4-ComputerVision/07-ConvNets/README.md @@ -17,7 +17,7 @@ CO_OP_TRANSLATOR_METADATA: 为了提取模式，我们将使用**卷积滤波器**的概念。正如你所知，图像可以用二维矩阵或带有颜色深度的三维张量来表示。应用滤波器意味着我们使用一个相对较小的**滤波核**矩阵，并对原始图像中的每个像素与其邻近点进行加权平均计算。我们可以将其视为一个小窗口在整个图像上滑动，并根据滤波核矩阵中的权重对所有像素进行平均。 -![垂直边缘滤波器](../../../../../translated_images/zh/filter-vert.b7148390ca0bc356.webp) | ![水平边缘滤波器](../../../../../translated_images/zh/filter-horiz.59b80ed4feb946ef.webp) +![垂直边缘滤波器](../../../../../translated_images/zh-CN/filter-vert.b7148390ca0bc356.webp) | ![水平边缘滤波器](../../../../../translated_images/zh-CN/filter-horiz.59b80ed4feb946ef.webp) ----|---- > 图片来源：Dmitry Soshnikov @@ -38,7 +38,7 @@ CNN 的工作方式基于以下重要思想： * 我们可以设计网络，使滤波器能够自动训练 * 我们可以使用相同的方法来发现高级特征中的模式，而不仅仅是原始图像中的模式。因此，CNN 的特征提取在特征的层次结构中工作，从低级像素组合开始，到更高级的图像部分组合。 -![层次特征提取](../../../../../translated_images/zh/FeatureExtractionCNN.d9b456cbdae7cb64.webp) +![层次特征提取](../../../../../translated_images/zh-CN/FeatureExtractionCNN.d9b456cbdae7cb64.webp) > 图片来源：[Hislop-Lynch 的论文](https://www.semanticscholar.org/paper/Computer-vision-based-pedestrian-trajectory-Hislop-Lynch/26e6f74853fc9bbb7487b06dc2cf095d36c9021d)，基于[他们的研究](https://dl.acm.org/doi/abs/10.1145/1553374.1553453) @@ -55,9 +55,9 @@ CNN 的工作方式基于以下重要思想： 例如，让我们看看 VGG-16 的架构，这是一种在 2014 年 ImageNet 的 top-5 分类中实现了 92.7% 准确率的网络： -![ImageNet 层次结构](../../../../../translated_images/zh/vgg-16-arch1.d901a5583b3a51ba.webp) +![ImageNet 层次结构](../../../../../translated_images/zh-CN/vgg-16-arch1.d901a5583b3a51ba.webp) -![ImageNet 金字塔](../../../../../translated_images/zh/vgg-16-arch.64ff2137f50dd49f.webp) +![ImageNet 金字塔](../../../../../translated_images/zh-CN/vgg-16-arch.64ff2137f50dd49f.webp) > 图片来源：[Researchgate](https://www.researchgate.net/figure/Vgg16-model-structure-To-get-the-VGG-NIN-model-we-replace-the-2-nd-4-th-6-th-7-th_fig2_335194493) diff --git a/translations/zh/lessons/4-ComputerVision/07-ConvNets/lab/README.md b/translations/zh/lessons/4-ComputerVision/07-ConvNets/lab/README.md index 3e5e8968..f2e098a7 100644 --- a/translations/zh/lessons/4-ComputerVision/07-ConvNets/lab/README.md +++ b/translations/zh/lessons/4-ComputerVision/07-ConvNets/lab/README.md @@ -21,7 +21,7 @@ CO_OP_TRANSLATOR_METADATA: 我们将使用 [Oxford-IIIT Pet Dataset](https://www.robots.ox.ac.uk/~vgg/data/pets/)，该数据集包含37种不同品种的猫和狗的图像。 -![我们将处理的数据集](../../../../../../translated_images/zh/data.50b2a9d5484bdbf0.webp) +![我们将处理的数据集](../../../../../../translated_images/zh-CN/data.50b2a9d5484bdbf0.webp) 要下载数据集，请使用以下代码片段： diff --git a/translations/zh/lessons/4-ComputerVision/08-TransferLearning/README.md b/translations/zh/lessons/4-ComputerVision/08-TransferLearning/README.md index ba35a033..ba5aa4fc 100644 --- a/translations/zh/lessons/4-ComputerVision/08-TransferLearning/README.md +++ b/translations/zh/lessons/4-ComputerVision/08-TransferLearning/README.md @@ -29,7 +29,7 @@ Keras 和 PyTorch 都包含函数，可以轻松加载一些常见架构的预 以下是 VGG-16 网络从一张猫的图片中提取的示例特征： -![VGG-16 提取的特征](../../../../../translated_images/zh/features.6291f9c7ba3a0b95.webp) +![VGG-16 提取的特征](../../../../../translated_images/zh-CN/features.6291f9c7ba3a0b95.webp) ## 猫与狗数据集 @@ -48,19 +48,19 @@ Keras 和 PyTorch 都包含函数，可以轻松加载一些常见架构的预 我们可以采取的一种方法是从一个随机图像开始，然后尝试使用**梯度下降优化**技术调整该图像，使网络认为它是一只猫。 -![图像优化循环](../../../../../translated_images/zh/ideal-cat-loop.999fbb8ff306e044.webp) +![图像优化循环](../../../../../translated_images/zh-CN/ideal-cat-loop.999fbb8ff306e044.webp) 然而，如果我们这样做，我们会得到一些非常类似于随机噪声的东西。这是因为*有很多方法可以让网络认为输入图像是一只猫*，包括一些在视觉上没有意义的方式。虽然这些图像包含了许多典型的猫的模式，但没有任何约束使它们在视觉上具有辨识度。 为了改善结果，我们可以在损失函数中添加另一个项，称为**变化损失**。它是一种度量，显示图像中相邻像素的相似程度。最小化变化损失可以使图像更平滑，并消除噪声，从而揭示更具视觉吸引力的模式。以下是一些这样的“理想”图像示例，它们被高概率分类为猫和斑马： -![理想猫](../../../../../translated_images/zh/ideal-cat.203dd4597643d6b0.webp) | ![理想斑马](../../../../../translated_images/zh/ideal-zebra.7f70e8b54ee15a7a.webp) +![理想猫](../../../../../translated_images/zh-CN/ideal-cat.203dd4597643d6b0.webp) | ![理想斑马](../../../../../translated_images/zh-CN/ideal-zebra.7f70e8b54ee15a7a.webp) -----|----- *理想猫* | *理想斑马* 类似的方法可以用来对神经网络进行所谓的**对抗性攻击**。假设我们想欺骗一个神经网络，让一只狗看起来像一只猫。如果我们拿一张被网络识别为狗的狗的图片，然后稍微调整它，直到网络开始将其分类为猫： -![狗的图片](../../../../../translated_images/zh/original-dog.8f68a67d2fe0911f.webp) | ![被分类为猫的狗的图片](../../../../../translated_images/zh/adversarial-dog.d9fc7773b0142b89.webp) +![狗的图片](../../../../../translated_images/zh-CN/original-dog.8f68a67d2fe0911f.webp) | ![被分类为猫的狗的图片](../../../../../translated_images/zh-CN/adversarial-dog.d9fc7773b0142b89.webp) -----|----- *狗的原始图片* | *被分类为猫的狗的图片* diff --git a/translations/zh/lessons/4-ComputerVision/09-Autoencoders/README.md b/translations/zh/lessons/4-ComputerVision/09-Autoencoders/README.md index 2617041d..96e17015 100644 --- a/translations/zh/lessons/4-ComputerVision/09-Autoencoders/README.md +++ b/translations/zh/lessons/4-ComputerVision/09-Autoencoders/README.md @@ -19,7 +19,7 @@ CO_OP_TRANSLATOR_METADATA: 由于我们训练自动编码器的目标是尽可能捕捉原始图像的信息以实现准确的重建，网络会尝试找到输入图像的最佳**嵌入**以捕捉其意义。 -![自动编码器示意图](../../../../../translated_images/zh/autoencoder_schema.5e6fc9ad98a5eb61.webp) +![自动编码器示意图](../../../../../translated_images/zh-CN/autoencoder_schema.5e6fc9ad98a5eb61.webp) > 图片来源：[Keras 博客](https://blog.keras.io/building-autoencoders-in-keras.html) diff --git a/translations/zh/lessons/4-ComputerVision/11-ObjectDetection/README.md b/translations/zh/lessons/4-ComputerVision/11-ObjectDetection/README.md index 6110dc9c..b153db99 100644 --- a/translations/zh/lessons/4-ComputerVision/11-ObjectDetection/README.md +++ b/translations/zh/lessons/4-ComputerVision/11-ObjectDetection/README.md @@ -13,7 +13,7 @@ CO_OP_TRANSLATOR_METADATA: ## [课前测验](https://ff-quizzes.netlify.app/en/ai/quiz/21) -![目标检测](../../../../../translated_images/zh/Screen_Shot_2016-11-17_at_11.14.54_AM.b4bb3769353287be.webp) +![目标检测](../../../../../translated_images/zh-CN/Screen_Shot_2016-11-17_at_11.14.54_AM.b4bb3769353287be.webp) > 图片来源：[YOLO v2 网站](https://pjreddie.com/darknet/yolov2/) @@ -25,7 +25,7 @@ CO_OP_TRANSLATOR_METADATA: 2. 对每个小块进行图像分类。 3. 对于分类结果激活值足够高的小块，可以认为其中包含目标物体。 -![简单目标检测](../../../../../translated_images/zh/naive-detection.e7f1ba220ccd08c6.webp) +![简单目标检测](../../../../../translated_images/zh-CN/naive-detection.e7f1ba220ccd08c6.webp) > *图片来源：[练习笔记本](ObjectDetection-TF.ipynb)* @@ -42,7 +42,7 @@ CO_OP_TRANSLATOR_METADATA: * [PASCAL VOC](http://host.robots.ox.ac.uk/pascal/VOC/) - 包含20个类别 * [COCO](http://cocodataset.org/#home) - 常见物体上下文数据集。包含80个类别、边界框和分割掩码 -![COCO](../../../../../translated_images/zh/coco-examples.71bc60380fa6cceb.webp) +![COCO](../../../../../translated_images/zh-CN/coco-examples.71bc60380fa6cceb.webp) ## 目标检测评估指标 @@ -50,7 +50,7 @@ CO_OP_TRANSLATOR_METADATA: 对于图像分类来说，评估算法性能相对简单；但对于目标检测，我们需要同时评估类别的正确性以及推断出的边界框位置的精确性。后者通常使用**交并比**（IoU）来衡量两个框（或任意两个区域）的重叠程度。 -![IoU](../../../../../translated_images/zh/iou_equation.9a4751d40fff4e11.webp) +![IoU](../../../../../translated_images/zh-CN/iou_equation.9a4751d40fff4e11.webp) > *图片来源：[这篇关于IoU的优秀博客文章](https://pyimagesearch.com/2016/11/07/intersection-over-union-iou-for-object-detection/)* @@ -98,11 +98,11 @@ $$ [R-CNN](http://islab.ulsan.ac.kr/files/announcement/513/rcnn_pami.pdf)使用[选择性搜索](http://www.huppelen.nl/publications/selectiveSearchDraft.pdf)生成ROI区域的层次结构，然后通过CNN特征提取器和SVM分类器确定物体类别，并通过线性回归确定*边界框*坐标。[官方论文](https://arxiv.org/pdf/1506.01497v1.pdf) -![RCNN](../../../../../translated_images/zh/rcnn1.cae407020dfb1d1f.webp) +![RCNN](../../../../../translated_images/zh-CN/rcnn1.cae407020dfb1d1f.webp) > *图片来源：van de Sande et al. ICCV’11* -![RCNN-1](../../../../../translated_images/zh/rcnn2.2d9530bb83516484.webp) +![RCNN-1](../../../../../translated_images/zh-CN/rcnn2.2d9530bb83516484.webp) > *图片来源：[这篇博客](https://towardsdatascience.com/r-cnn-fast-r-cnn-faster-r-cnn-yolo-object-detection-algorithms-36d53571365e)* @@ -110,7 +110,7 @@ $$ 这种方法与R-CNN类似，但区域是在应用卷积层之后定义的。 -![FRCNN](../../../../../translated_images/zh/f-rcnn.3cda6d9bb4188875.webp) +![FRCNN](../../../../../translated_images/zh-CN/f-rcnn.3cda6d9bb4188875.webp) > 图片来源：[官方论文](https://www.cv-foundation.org/openaccess/content_iccv_2015/papers/Girshick_Fast_R-CNN_ICCV_2015_paper.pdf)，[arXiv](https://arxiv.org/pdf/1504.08083.pdf)，2015 @@ -118,7 +118,7 @@ $$ 这种方法的核心思想是使用神经网络预测ROI——即所谓的*区域提议网络*。[论文](https://arxiv.org/pdf/1506.01497.pdf)，2016 -![FasterRCNN](../../../../../translated_images/zh/faster-rcnn.8d46c099b87ef30a.webp) +![FasterRCNN](../../../../../translated_images/zh-CN/faster-rcnn.8d46c099b87ef30a.webp) > 图片来源：[官方论文](https://arxiv.org/pdf/1506.01497.pdf) @@ -130,7 +130,7 @@ $$ 2. 特征通过**位置敏感得分图**处理。每个类别$C$的物体被划分为$k\times k$区域，并训练预测物体的各部分。 3. 对于$k\times k$区域中的每个部分，所有网络对物体类别进行投票，选择投票最多的类别。 -![r-fcn image](../../../../../translated_images/zh/r-fcn.13eb88158b99a3da.webp) +![r-fcn image](../../../../../translated_images/zh-CN/r-fcn.13eb88158b99a3da.webp) > 图片来源：[官方论文](https://arxiv.org/abs/1605.06409) @@ -141,7 +141,7 @@ YOLO是一种实时单次检测算法。其核心思想如下： * 将图像划分为$S\times S$区域。 * 对每个区域，**CNN**预测$n$个可能的物体、*边界框*坐标和*置信度*=*概率* * IoU。 - ![YOLO](../../../../../translated_images/zh/yolo.a2648ec82ee8bb4e.webp) + ![YOLO](../../../../../translated_images/zh-CN/yolo.a2648ec82ee8bb4e.webp) > 图片来源：[官方论文](https://arxiv.org/abs/1506.02640) diff --git a/translations/zh/lessons/4-ComputerVision/README.md b/translations/zh/lessons/4-ComputerVision/README.md index f559fc76..8c2f7060 100644 --- a/translations/zh/lessons/4-ComputerVision/README.md +++ b/translations/zh/lessons/4-ComputerVision/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 计算机视觉 -![计算机视觉内容总结涂鸦](../../../../translated_images/zh/ai-computervision.6506ebebac3fbf76.webp) +![计算机视觉内容总结涂鸦](../../../../translated_images/zh-CN/ai-computervision.6506ebebac3fbf76.webp) 在本节中，我们将学习以下内容： diff --git a/translations/zh/lessons/5-NLP/14-Embeddings/README.md b/translations/zh/lessons/5-NLP/14-Embeddings/README.md index 666ecd7a..b16b12c6 100644 --- a/translations/zh/lessons/5-NLP/14-Embeddings/README.md +++ b/translations/zh/lessons/5-NLP/14-Embeddings/README.md @@ -19,7 +19,7 @@ CO_OP_TRANSLATOR_METADATA: 通过在分类器网络中使用嵌入层作为第一层，我们可以从词袋模型切换到 **嵌入袋** 模型。在嵌入袋模型中，我们首先将文本中的每个单词转换为对应的嵌入，然后对所有这些嵌入计算某种聚合函数，例如 `sum`、`average` 或 `max`。 -![展示五个序列单词的嵌入分类器的图片。](../../../../../translated_images/zh/embedding-classifier-example.b77f021a7ee67eee.webp) +![展示五个序列单词的嵌入分类器的图片。](../../../../../translated_images/zh-CN/embedding-classifier-example.b77f021a7ee67eee.webp) > 图片由作者提供 @@ -40,7 +40,7 @@ CO_OP_TRANSLATOR_METADATA: CBoW 速度更快，而 Skip-Gram 虽然较慢，但在表示不常见单词方面表现更好。 -![展示 CBoW 和 Skip-Gram 算法将单词转换为向量的图片。](../../../../../translated_images/zh/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) +![展示 CBoW 和 Skip-Gram 算法将单词转换为向量的图片。](../../../../../translated_images/zh-CN/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) > 图片来源于 [这篇论文](https://arxiv.org/pdf/1301.3781.pdf) diff --git a/translations/zh/lessons/5-NLP/15-LanguageModeling/README.md b/translations/zh/lessons/5-NLP/15-LanguageModeling/README.md index aec20b2d..aa9a39ca 100644 --- a/translations/zh/lessons/5-NLP/15-LanguageModeling/README.md +++ b/translations/zh/lessons/5-NLP/15-LanguageModeling/README.md @@ -23,7 +23,7 @@ CO_OP_TRANSLATOR_METADATA: * **连续词袋模型** (CBoW)，通过预测标记序列 $W_{-N}$, ..., $W_N$ 中的中间标记 $W_0$。 * **Skip-gram**，通过中间标记 $W_0$ 来预测一组邻近标记 {$W_{-N},\dots, W_{-1}, W_1,\dots, W_N$}。 -![论文中关于将单词转换为向量的算法示例](../../../../../translated_images/zh/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) +![论文中关于将单词转换为向量的算法示例](../../../../../translated_images/zh-CN/example-algorithms-for-converting-words-to-vectors.fbe9207a726922f6.webp) > 图片来源于[这篇论文](https://arxiv.org/pdf/1301.3781.pdf) diff --git a/translations/zh/lessons/5-NLP/16-RNN/README.md b/translations/zh/lessons/5-NLP/16-RNN/README.md index 903d1eb0..f3ee7ec9 100644 --- a/translations/zh/lessons/5-NLP/16-RNN/README.md +++ b/translations/zh/lessons/5-NLP/16-RNN/README.md @@ -15,7 +15,7 @@ CO_OP_TRANSLATOR_METADATA: 为了捕捉文本序列的意义，我们需要使用另一种神经网络架构，称为**循环神经网络**（Recurrent Neural Network，RNN）。在 RNN 中，我们将句子逐个符号输入网络，网络会生成某种**状态**，然后将该状态与下一个符号一起再次输入网络。 -![RNN](../../../../../translated_images/zh/rnn.27f5c29c53d727b5.webp) +![RNN](../../../../../translated_images/zh-CN/rnn.27f5c29c53d727b5.webp) > 图片由作者提供 @@ -61,7 +61,7 @@ LSTM 网络的组织方式与 RNN 类似，但有两个状态会从层到层传 循环网络，无论是单向还是双向，都能捕捉序列中的某些模式，并将其存储到状态向量中或传递到输出中。与卷积网络类似，我们可以在第一层之上构建另一层循环网络，以捕捉更高级的模式，并从第一层提取的低级模式中构建。这引出了**多层 RNN** 的概念，它由两个或更多循环网络组成，其中前一层的输出作为输入传递到下一层。 -![显示多层长短时记忆 RNN 的图片](../../../../../translated_images/zh/multi-layer-lstm.dd975e29bb2a59fe.webp) +![显示多层长短时记忆 RNN 的图片](../../../../../translated_images/zh-CN/multi-layer-lstm.dd975e29bb2a59fe.webp) *图片来自 Fernando López 的[这篇精彩文章](https://towardsdatascience.com/from-a-lstm-cell-to-a-multilayer-lstm-network-with-pytorch-2899eb5696f3)* diff --git a/translations/zh/lessons/5-NLP/17-GenerativeNetworks/README.md b/translations/zh/lessons/5-NLP/17-GenerativeNetworks/README.md index 6c6e2b2d..425f48c9 100644 --- a/translations/zh/lessons/5-NLP/17-GenerativeNetworks/README.md +++ b/translations/zh/lessons/5-NLP/17-GenerativeNetworks/README.md @@ -19,7 +19,7 @@ CO_OP_TRANSLATOR_METADATA: 这使得可以构建不同的神经网络架构，如下图所示： -![展示常见循环神经网络模式的图片。](../../../../../translated_images/zh/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) +![展示常见循环神经网络模式的图片。](../../../../../translated_images/zh-CN/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) > 图片来源于 [Andrej Karpaty](http://karpathy.github.io/) 的博客文章 [循环神经网络的非凡有效性](http://karpathy.github.io/2015/05/21/rnn-effectiveness/) @@ -32,7 +32,7 @@ CO_OP_TRANSLATOR_METADATA: 我们将训练这个 RNN 逐步生成文本。在每一步中，我们将取长度为 `nchars` 的字符序列，并要求网络为每个输入字符生成下一个输出字符： -![展示 RNN 生成单词 'HELLO' 的示例图片。](../../../../../translated_images/zh/rnn-generate.56c54afb52f9781d.webp) +![展示 RNN 生成单词 'HELLO' 的示例图片。](../../../../../translated_images/zh-CN/rnn-generate.56c54afb52f9781d.webp) 在生成文本（推理阶段）时，我们从某个**提示**开始，将其传递给 RNN 单元以生成中间状态，然后从该状态开始生成。我们一次生成一个字符，并将状态和生成的字符传递给另一个 RNN 单元以生成下一个字符，直到生成足够的字符。 diff --git a/translations/zh/lessons/5-NLP/18-Transformers/README.md b/translations/zh/lessons/5-NLP/18-Transformers/README.md index cfa242f1..fbc1c29c 100644 --- a/translations/zh/lessons/5-NLP/18-Transformers/README.md +++ b/translations/zh/lessons/5-NLP/18-Transformers/README.md @@ -20,13 +20,13 @@ CO_OP_TRANSLATOR_METADATA: **注意机制**提供了一种方法，可以对每个输入向量对RNN输出预测的上下文影响进行加权。其实现方式是创建输入RNN的中间状态与输出RNN之间的快捷路径。这样，在生成输出符号yt时，我们会考虑所有输入隐藏状态hi，并赋予不同的权重系数αt,i。 -![显示带有加性注意层的编码器/解码器模型的图像](../../../../../translated_images/zh/encoder-decoder-attention.7a726296894fb567.webp) +![显示带有加性注意层的编码器/解码器模型的图像](../../../../../translated_images/zh-CN/encoder-decoder-attention.7a726296894fb567.webp) > [Bahdanau等人，2015](https://arxiv.org/pdf/1409.0473.pdf)中的加性注意机制编码器-解码器模型，图片来源于[这篇博客](https://lilianweng.github.io/lil-log/2018/06/24/attention-attention.html) 注意矩阵{αi,j}表示某些输入词在生成输出序列中某个词时的影响程度。下面是一个这样的矩阵示例： -![显示由RNNsearch-50找到的样本对齐的图像，取自Bahdanau - arviz.org](../../../../../translated_images/zh/bahdanau-fig3.09ba2d37f202a6af.webp) +![显示由RNNsearch-50找到的样本对齐的图像，取自Bahdanau - arviz.org](../../../../../translated_images/zh-CN/bahdanau-fig3.09ba2d37f202a6af.webp) > 图片来自[Bahdanau等人，2015](https://arxiv.org/pdf/1409.0473.pdf)（图3） @@ -66,7 +66,7 @@ Transformer的核心思想之一是避免RNN的顺序处理特性，并创建一 接下来，我们需要捕捉序列中的一些模式。为此，Transformer使用了**自注意机制**，即将注意机制同时应用于输入和输出的同一序列。应用自注意机制使我们能够考虑句子中的**上下文**，并查看哪些词是相互关联的。例如，它可以帮助我们识别代词*it*所指代的词，并考虑上下文： -![](../../../../../translated_images/zh/CoreferenceResolution.861924d6d384a7d6.webp) +![](../../../../../translated_images/zh-CN/CoreferenceResolution.861924d6d384a7d6.webp) > 图片来源于[谷歌博客](https://research.googleblog.com/2017/08/transformer-novel-neural-network.html) @@ -91,7 +91,7 @@ Transformer的核心思想之一是避免RNN的顺序处理特性，并创建一 **BERT**（Bidirectional Encoder Representations from Transformers，双向编码器表示）是一个非常大的多层Transformer网络，*BERT-base*有12层，*BERT-large*有24层。模型首先在大规模文本数据（维基百科+书籍）上进行无监督训练（预测句子中的被遮蔽词）。在预训练过程中，模型吸收了大量的语言理解能力，这些能力可以通过微调其他数据集来利用。这一过程称为**迁移学习**。 -![图片来源于http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/zh/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp) +![图片来源于http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/zh-CN/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp) > 图片[来源](http://jalammar.github.io/illustrated-bert/) diff --git a/translations/zh/lessons/5-NLP/18-Transformers/READMEtransformers.md b/translations/zh/lessons/5-NLP/18-Transformers/READMEtransformers.md index dcc9e593..b252d930 100644 --- a/translations/zh/lessons/5-NLP/18-Transformers/READMEtransformers.md +++ b/translations/zh/lessons/5-NLP/18-Transformers/READMEtransformers.md @@ -11,13 +11,13 @@ **注意力机制**提供了一种对每个输入向量在RNN每个输出预测中的上下文影响进行加权的方法。它的实现方式是创建输入RNN和输出RNN之间的中间状态的快捷方式。通过这种方式，在生成输出符号 yt 时，我们将考虑所有输入隐藏状态 hi，并使用不同的权重系数 αt,i。 -![展示带有加性注意力层的编码器/解码器模型的图像](../../../../../translated_images/zh/encoder-decoder-attention.7a726296894fb567.webp) +![展示带有加性注意力层的编码器/解码器模型的图像](../../../../../translated_images/zh-CN/encoder-decoder-attention.7a726296894fb567.webp) > 该编码器-解码器模型与加性注意力机制见于 [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf)，引用自 [这篇博客文章](https://lilianweng.github.io/lil-log/2018/06/24/attention-attention.html) 注意力矩阵 {αi,j} 表示某些输入单词在生成输出序列中给定单词的程度。以下是这样一个矩阵的示例： -![展示由RNNsearch-50找到的示例对齐的图像，取自Bahdanau - arviz.org](../../../../../translated_images/zh/bahdanau-fig3.09ba2d37f202a6af.webp) +![展示由RNNsearch-50找到的示例对齐的图像，取自Bahdanau - arviz.org](../../../../../translated_images/zh-CN/bahdanau-fig3.09ba2d37f202a6af.webp) > 图自 [Bahdanau et al., 2015](https://arxiv.org/pdf/1409.0473.pdf) (图3) @@ -57,7 +57,7 @@ 接下来，我们需要捕捉序列中的一些模式。为此，变换器使用**自注意力**机制，这本质上是对同一序列应用的注意力。应用自注意力使我们能够考虑句子中的**上下文**，并查看哪些单词是相互关联的。例如，它使我们能够看到哪些单词是由指代词（如 *它*）引用的，并且还考虑上下文： -![](../../../../../translated_images/zh/CoreferenceResolution.861924d6d384a7d6.webp) +![](../../../../../translated_images/zh-CN/CoreferenceResolution.861924d6d384a7d6.webp) > 图自 [Google博客](https://research.googleblog.com/2017/08/transformer-novel-neural-network.html) @@ -82,7 +82,7 @@ **BERT**（来自变换器的双向编码器表示）是一个非常大的多层变换器网络，其中 *BERT-base* 有12层，*BERT-large* 有24层。该模型首先在一个大型文本数据集（维基百科 + 书籍）上进行预训练，采用无监督训练（预测句子中被屏蔽的单词）。在预训练期间，模型吸收了显著的语言理解能力，这可以通过微调与其他数据集结合使用。这个过程被称为**迁移学习**。 -![图片来自 http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/zh/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp) +![图片来自 http://jalammar.github.io/illustrated-bert/](../../../../../translated_images/zh-CN/jalammarBERT-language-modeling-masked-lm.34f113ea5fec4362.webp) > 图片 [来源](http://jalammar.github.io/illustrated-bert/) diff --git a/translations/zh/lessons/5-NLP/19-NER/README.md b/translations/zh/lessons/5-NLP/19-NER/README.md index 5dc723d3..57391fd4 100644 --- a/translations/zh/lessons/5-NLP/19-NER/README.md +++ b/translations/zh/lessons/5-NLP/19-NER/README.md @@ -58,7 +58,7 @@ NER模型本质上是**标记分类模型**，因为对于每个输入标记， 由于我们需要在标记和类别之间建立一一对应关系，我们可以从下图中训练一个最右侧的**多对多**神经网络模型： -![展示常见循环神经网络模式的图片。](../../../../../translated_images/zh/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) +![展示常见循环神经网络模式的图片。](../../../../../translated_images/zh-CN/unreasonable-effectiveness-of-rnn.541ead816778f42d.webp) > *图片来自[这篇博客文章](http://karpathy.github.io/2015/05/21/rnn-effectiveness/)作者为[Andrej Karpathy](http://karpathy.github.io/)。NER标记分类模型对应于此图片中最右侧的网络架构。* diff --git a/translations/zh/lessons/5-NLP/README.md b/translations/zh/lessons/5-NLP/README.md index 58f745e0..75381601 100644 --- a/translations/zh/lessons/5-NLP/README.md +++ b/translations/zh/lessons/5-NLP/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 自然语言处理 -![NLP任务总结图](../../../../translated_images/zh/ai-nlp.b22dcb8ca4707cea.webp) +![NLP任务总结图](../../../../translated_images/zh-CN/ai-nlp.b22dcb8ca4707cea.webp) 在本节中，我们将重点使用神经网络来处理与**自然语言处理 (NLP)**相关的任务。我们希望计算机能够解决许多NLP问题： diff --git a/translations/zh/lessons/6-Other/23-MultiagentSystems/README.md b/translations/zh/lessons/6-Other/23-MultiagentSystems/README.md index 771ee192..2b32a56c 100644 --- a/translations/zh/lessons/6-Other/23-MultiagentSystems/README.md +++ b/translations/zh/lessons/6-Other/23-MultiagentSystems/README.md @@ -71,7 +71,7 @@ NetLogo的一个优点是它包含一个可供尝试的工作模型库。进入* 打开模型后，你会进入NetLogo的主界面。以下是一个描述狼和羊种群的示例模型，考虑到有限资源（草地）。 -![NetLogo主界面](../../../../../translated_images/zh/NetLogo-Main.32653711ec1a01b3.webp) +![NetLogo主界面](../../../../../translated_images/zh-CN/NetLogo-Main.32653711ec1a01b3.webp) > Dmitry Soshnikov提供的截图 diff --git a/translations/zh/lessons/README.md b/translations/zh/lessons/README.md index 8dd52219..2fae2c6e 100644 --- a/translations/zh/lessons/README.md +++ b/translations/zh/lessons/README.md @@ -9,7 +9,7 @@ CO_OP_TRANSLATOR_METADATA: --> # 概述 -![概述的手绘图](../../../translated_images/zh/ai-overview.0857791951d19500.webp) +![概述的手绘图](../../../translated_images/zh-CN/ai-overview.0857791951d19500.webp) > 手绘笔记由 [Tomomi Imura](https://twitter.com/girlie_mac) 提供 diff --git a/translations/zh/lessons/X-Extras/X1-MultiModal/README.md b/translations/zh/lessons/X-Extras/X1-MultiModal/README.md index 8567ac3e..2658bb1c 100644 --- a/translations/zh/lessons/X-Extras/X1-MultiModal/README.md +++ b/translations/zh/lessons/X-Extras/X1-MultiModal/README.md @@ -15,7 +15,7 @@ CO_OP_TRANSLATOR_METADATA: CLIP的核心思想是能够比较文本提示与图像，并确定图像与提示的匹配程度。 -![CLIP架构](../../../../../translated_images/zh/clip-arch.b3dbf20b4e8ed8be.webp) +![CLIP架构](../../../../../translated_images/zh-CN/clip-arch.b3dbf20b4e8ed8be.webp) > *图片来源于[这篇博客](https://openai.com/blog/clip/)* @@ -29,7 +29,7 @@ CLIP模型/库可以从[OpenAI GitHub](https://github.com/openai/CLIP)获取。 假设我们需要将图像分类为猫、狗和人类。在这种情况下，我们可以给模型一个图像，以及一系列文本提示：“*一张猫的图片*”、“*一张狗的图片*”、“*一张人类的图片*”。在结果的3个概率向量中，我们只需选择值最高的索引。 -![CLIP用于图像分类](../../../../../translated_images/zh/clip-class.3af42ef0b2b19369.webp) +![CLIP用于图像分类](../../../../../translated_images/zh-CN/clip-class.3af42ef0b2b19369.webp) > *图片来源于[这篇博客](https://openai.com/blog/clip/)* @@ -53,13 +53,13 @@ VQGAN与普通[GAN](../../4-ComputerVision/10-GANs/README.md)的主要区别在 VQGAN与传统GAN的一个重要区别在于，后者可以从任何输入向量生成一个不错的图像，而VQGAN可能生成一个不连贯的图像。因此，我们需要进一步引导图像创建过程，这可以通过CLIP来实现。 -![VQGAN+CLIP架构](../../../../../translated_images/zh/vqgan.5027fe05051dfa31.webp) +![VQGAN+CLIP架构](../../../../../translated_images/zh-CN/vqgan.5027fe05051dfa31.webp) 为了生成与文本提示相对应的图像，我们从一些随机编码向量开始，将其传递给VQGAN以生成图像。然后使用CLIP生成一个损失函数，显示图像与文本提示的匹配程度。目标是通过反向传播调整输入向量参数以最小化该损失。 一个实现VQGAN+CLIP的优秀库是[Pixray](http://github.com/pixray/pixray)。 -![Pixray生成的图片](../../../../../translated_images/zh/a_closeup_watercolor_portrait_of_young_male_teacher_of_literature_with_a_book.2384968e9db8a0d0.webp) | ![Pixray生成的图片](../../../../../translated_images/zh/a_closeup_oil_portrait_of_young_female_teacher_of_computer_science_with_a_computer.e0b6495f210a4390.webp) | ![Pixray生成的图片](../../../../../translated_images/zh/a_closeup_oil_portrait_of_old_male_teacher_of_math.5362e67aa7fc2683.webp) +![Pixray生成的图片](../../../../../translated_images/zh-CN/a_closeup_watercolor_portrait_of_young_male_teacher_of_literature_with_a_book.2384968e9db8a0d0.webp) | ![Pixray生成的图片](../../../../../translated_images/zh-CN/a_closeup_oil_portrait_of_young_female_teacher_of_computer_science_with_a_computer.e0b6495f210a4390.webp) | ![Pixray生成的图片](../../../../../translated_images/zh-CN/a_closeup_oil_portrait_of_old_male_teacher_of_math.5362e67aa7fc2683.webp) ----|----|---- 从提示*一张年轻男性文学教师拿着书的水彩特写肖像*生成的图片 | 从提示*一张年轻女性计算机科学教师拿着电脑的油画特写肖像*生成的图片 | 从提示*一张老年男性数学教师站在黑板前的油画特写肖像*生成的图片 @@ -75,7 +75,7 @@ DALL-E是一个基于GPT-3的版本，能够根据提示生成图像。它使用 DALL-E 1和2的主要区别在于，后者能够生成更真实的图像和艺术作品。 以下是使用DALL-E生成图像的示例： -![DALL-E生成的图片](../../../../../translated_images/zh/DALL·E%202023-06-20%2015.56.56%20-%20a%20closeup%20watercolor%20portrait%20of%20young%20male%20teacher%20of%20literature%20with%20a%20book.6c235e8271d9ed10ce985d86aeb241a58518958647973af136912116b9518fce.png) | ![DALL-E生成的图片](../../../../../translated_images/zh/DALL·E%202023-06-20%2015.57.43%20-%20a%20closeup%20oil%20portrait%20of%20young%20female%20teacher%20of%20computer%20science%20with%20a%20computer.f21dc4166340b6c8b4d1cb57efd1e22127407f9b28c9ac7afe11344065369e64.png) | ![DALL-E生成的图片](../../../../../translated_images/zh/DALL·E%202023-06-20%2015.58.42%20-%20%20a%20closeup%20oil%20portrait%20of%20old%20male%20teacher%20of%20mathematics%20in%20front%20of%20blackboard.d331c2dfbdc3f7c46aa65c0809066f5e7ed4b49609cd259852e760df21051e4a.png) +![DALL-E生成的图片](../../../../../translated_images/zh-CN/DALL·E%202023-06-20%2015.56.56%20-%20a%20closeup%20watercolor%20portrait%20of%20young%20male%20teacher%20of%20literature%20with%20a%20book.6c235e8271d9ed10ce985d86aeb241a58518958647973af136912116b9518fce.png) | ![DALL-E生成的图片](../../../../../translated_images/zh-CN/DALL·E%202023-06-20%2015.57.43%20-%20a%20closeup%20oil%20portrait%20of%20young%20female%20teacher%20of%20computer%20science%20with%20a%20computer.f21dc4166340b6c8b4d1cb57efd1e22127407f9b28c9ac7afe11344065369e64.png) | ![DALL-E生成的图片](../../../../../translated_images/zh-CN/DALL·E%202023-06-20%2015.58.42%20-%20%20a%20closeup%20oil%20portrait%20of%20old%20male%20teacher%20of%20mathematics%20in%20front%20of%20blackboard.d331c2dfbdc3f7c46aa65c0809066f5e7ed4b49609cd259852e760df21051e4a.png) ----|----|---- 从提示*一张年轻男性文学教师拿着书的水彩特写肖像*生成的图片 | 从提示*一张年轻女性计算机科学教师拿着电脑的油画特写肖像*生成的图片 | 从提示*一张老年男性数学教师站在黑板前的油画特写肖像*生成的图片