[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-leriomaggio--deep-learning-keras-tensorflow":3,"tool-leriomaggio--deep-learning-keras-tensorflow":64},[4,17,27,35,43,56],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":16},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,3,"2026-04-05T11:01:52",[13,14,15],"开发框架","图像","Agent","ready",{"id":18,"name":19,"github_repo":20,"description_zh":21,"stars":22,"difficulty_score":23,"last_commit_at":24,"category_tags":25,"status":16},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",138956,2,"2026-04-05T11:33:21",[13,15,26],"语言模型",{"id":28,"name":29,"github_repo":30,"description_zh":31,"stars":32,"difficulty_score":23,"last_commit_at":33,"category_tags":34,"status":16},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",107662,"2026-04-03T11:11:01",[13,14,15],{"id":36,"name":37,"github_repo":38,"description_zh":39,"stars":40,"difficulty_score":23,"last_commit_at":41,"category_tags":42,"status":16},3704,"NextChat","ChatGPTNextWeb\u002FNextChat","NextChat 是一款轻量且极速的 AI 助手，旨在为用户提供流畅、跨平台的大模型交互体验。它完美解决了用户在多设备间切换时难以保持对话连续性，以及面对众多 AI 模型不知如何统一管理的痛点。无论是日常办公、学习辅助还是创意激发，NextChat 都能让用户随时随地通过网页、iOS、Android、Windows、MacOS 或 Linux 端无缝接入智能服务。\n\n这款工具非常适合普通用户、学生、职场人士以及需要私有化部署的企业团队使用。对于开发者而言，它也提供了便捷的自托管方案，支持一键部署到 Vercel 或 Zeabur 等平台。\n\nNextChat 的核心亮点在于其广泛的模型兼容性，原生支持 Claude、DeepSeek、GPT-4 及 Gemini Pro 等主流大模型，让用户在一个界面即可自由切换不同 AI 能力。此外，它还率先支持 MCP（Model Context Protocol）协议，增强了上下文处理能力。针对企业用户，NextChat 提供专业版解决方案，具备品牌定制、细粒度权限控制、内部知识库整合及安全审计等功能，满足公司对数据隐私和个性化管理的高标准要求。",87618,"2026-04-05T07:20:52",[13,26],{"id":44,"name":45,"github_repo":46,"description_zh":47,"stars":48,"difficulty_score":23,"last_commit_at":49,"category_tags":50,"status":16},2268,"ML-For-Beginners","microsoft\u002FML-For-Beginners","ML-For-Beginners 是由微软推出的一套系统化机器学习入门课程，旨在帮助零基础用户轻松掌握经典机器学习知识。这套课程将学习路径规划为 12 周，包含 26 节精炼课程和 52 道配套测验，内容涵盖从基础概念到实际应用的完整流程，有效解决了初学者面对庞大知识体系时无从下手、缺乏结构化指导的痛点。\n\n无论是希望转型的开发者、需要补充算法背景的研究人员，还是对人工智能充满好奇的普通爱好者，都能从中受益。课程不仅提供了清晰的理论讲解，还强调动手实践，让用户在循序渐进中建立扎实的技能基础。其独特的亮点在于强大的多语言支持，通过自动化机制提供了包括简体中文在内的 50 多种语言版本，极大地降低了全球不同背景用户的学习门槛。此外，项目采用开源协作模式，社区活跃且内容持续更新，确保学习者能获取前沿且准确的技术资讯。如果你正寻找一条清晰、友好且专业的机器学习入门之路，ML-For-Beginners 将是理想的起点。",84991,"2026-04-05T10:45:23",[14,51,52,53,15,54,26,13,55],"数据工具","视频","插件","其他","音频",{"id":57,"name":58,"github_repo":59,"description_zh":60,"stars":61,"difficulty_score":10,"last_commit_at":62,"category_tags":63,"status":16},3128,"ragflow","infiniflow\u002Fragflow","RAGFlow 是一款领先的开源检索增强生成（RAG）引擎，旨在为大语言模型构建更精准、可靠的上下文层。它巧妙地将前沿的 RAG 技术与智能体（Agent）能力相结合，不仅支持从各类文档中高效提取知识，还能让模型基于这些知识进行逻辑推理和任务执行。\n\n在大模型应用中，幻觉问题和知识滞后是常见痛点。RAGFlow 通过深度解析复杂文档结构（如表格、图表及混合排版），显著提升了信息检索的准确度，从而有效减少模型“胡编乱造”的现象，确保回答既有据可依又具备时效性。其内置的智能体机制更进一步，使系统不仅能回答问题，还能自主规划步骤解决复杂问题。\n\n这款工具特别适合开发者、企业技术团队以及 AI 研究人员使用。无论是希望快速搭建私有知识库问答系统，还是致力于探索大模型在垂直领域落地的创新者，都能从中受益。RAGFlow 提供了可视化的工作流编排界面和灵活的 API 接口，既降低了非算法背景用户的上手门槛，也满足了专业开发者对系统深度定制的需求。作为基于 Apache 2.0 协议开源的项目，它正成为连接通用大模型与行业专有知识之间的重要桥梁。",77062,"2026-04-04T04:44:48",[15,14,13,26,54],{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":69,"readme_en":70,"readme_zh":71,"quickstart_zh":72,"use_case_zh":73,"hero_image_url":74,"owner_login":75,"owner_name":76,"owner_avatar_url":77,"owner_bio":78,"owner_company":79,"owner_location":80,"owner_email":79,"owner_twitter":75,"owner_website":79,"owner_url":81,"languages":82,"stars":91,"forks":92,"last_commit_at":93,"license":94,"difficulty_score":10,"env_os":95,"env_gpu":96,"env_ram":97,"env_deps":98,"category_tags":112,"github_topics":113,"view_count":23,"oss_zip_url":79,"oss_zip_packed_at":79,"status":16,"created_at":124,"updated_at":125,"faqs":126,"releases":162},2003,"leriomaggio\u002Fdeep-learning-keras-tensorflow","deep-learning-keras-tensorflow","Introduction to Deep Neural Networks with Keras and Tensorflow","deep-learning-keras-tensorflow 是一套面向初学者和中级学习者的深度学习实战教程，通过 Keras 和 TensorFlow 从零讲解神经网络的核心概念与应用。它涵盖从基础的多层感知机（MLP）、卷积神经网络（CNN）、循环神经网络（RNN），到自动编码器、迁移学习和超参数优化等实用技术，并提供大量可运行的代码示例，如在 MNIST 数据集上训练模型、使用 VGG 和 ResNet 进行图像分类等。教程特别注重实践，每个章节都配有清晰的注释和 Jupyter Notebook，帮助用户边学边做，真正理解模型背后的原理。它解决了深度学习入门门槛高、理论与实践脱节的问题，尤其适合希望系统掌握现代 AI 开发流程的开发者、数据科学学生和研究人员。教程不依赖复杂环境，仅需 Python 3 和常见库即可运行，支持 CPU 和 GPU 加速（可选 cuDNN），并融入了 Keras 的高级特性如自定义层与多模态网络，是学习工业级深度学习的优质入门资源。","\u003Cdiv>\r\n    \u003Ch1 style=\"text-align: center;\">Deep Learning with Keras and Tensorflow\u003C\u002Fh1>\r\n    \u003Cimg style=\"text-align: left\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fleriomaggio_deep-learning-keras-tensorflow_readme_2c18e3986a1f.jpg\" width=\"15%\" \u002F>\r\n\u003Cdiv>\r\n\u003Cbr>\r\n\r\n### Author: Valerio Maggio\r\n\r\n#### Contacts:\r\n\r\n\u003Ctable style=\"border: 0px; display: inline-table\">\r\n    \u003Ctbody>\r\n        \u003Ctr style=\"border: 0px;\">\r\n            \u003Ctd style=\"border: 0px;\">\r\n                \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fleriomaggio_deep-learning-keras-tensorflow_readme_da46c48e57f3.png\" style=\"display: inline-block;\" \u002F> \r\n                \u003Ca href=\"http:\u002F\u002Ftwitter.com\u002Fleriomaggio\" target=\"_blank\">@leriomaggio\u003C\u002Fa>\r\n\t    \u003C\u002Ftd>\r\n            \u003Ctd style=\"border: 0px;\">\r\n                \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fleriomaggio_deep-learning-keras-tensorflow_readme_792d827227e3.png\" style=\"display: inline-block;\" \u002F> \r\n                \u003Ca href=\"it.linkedin.com\u002Fin\u002Fvaleriomaggio\" target=\"_blank\">valeriomaggio\u003C\u002Fa>\r\n            \u003C\u002Ftd>\r\n            \u003Ctd style=\"border: 0px;\">\r\n                \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fleriomaggio_deep-learning-keras-tensorflow_readme_be8cc437ecf8.png\" style=\"display: inline-block;\" \u002F> \r\n                valeriomaggio_at_gmail\r\n            \u003C\u002Ftd>\r\n       \u003C\u002Ftr>\r\n  \u003C\u002Ftbody>\r\n\u003C\u002Ftable>\r\n\r\n\r\n```shell\r\n\r\ngit clone https:\u002F\u002Fgithub.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow.git\r\n```\r\n\r\n---\r\n\r\n## Table of Contents\r\n\r\n- **Part I**: **Introduction**\r\n\r\n    - Intro to Artificial Neural Networks\r\n        - Perceptron and MLP    \r\n        - naive pure-Python implementation\r\n        - fast forward, sgd, backprop\r\n        \r\n    - Introduction to Deep Learning Frameworks\r\n        - Intro to Theano\r\n        - Intro to Tensorflow\r\n        - Intro to Keras\r\n            - Overview and main features\r\n            - Overview of the `core` layers\r\n            - Multi-Layer Perceptron and Fully Connected\r\n                - Examples with `keras.models.Sequential` and `Dense`\r\n            - Keras Backend\r\n    \r\n- **Part II**: **Supervised Learning**\r\n    \r\n    - Fully Connected Networks and Embeddings\r\n        - Intro to MNIST Dataset\r\n        - Hidden Leayer Representation and Embeddings\r\n        \r\n    - Convolutional Neural Networks\r\n        - meaning of convolutional filters\r\n            - examples from ImageNet    \r\n        - Visualising ConvNets \r\n\r\n        - Advanced CNN\r\n            - Dropout\r\n            - MaxPooling\r\n            - Batch Normalisation\r\n\r\n        - HandsOn: MNIST Dataset\r\n            - FC and MNIST\r\n            - CNN and MNIST\r\n        \r\n        - Deep Convolutional Neural Networks with Keras (ref: `keras.applications`)\r\n            - VGG16\r\n            - VGG19\r\n            - ResNet50\r\n    - Transfer Learning and FineTuning\r\n    - Hyperparameters Optimisation \r\n        \r\n- **Part III**: **Unsupervised Learning**\r\n\r\n    - AutoEncoders and Embeddings\r\n\t- AutoEncoders and MNIST\r\n    \t- word2vec and doc2vec (gensim) with `keras.datasets`\r\n        - word2vec and CNN\r\n    \r\n- **Part IV**: **Recurrent Neural Networks**\r\n    - Recurrent Neural Network in Keras \r\n        -  `SimpleRNN`, `LSTM`, `GRU`\r\n    - LSTM for Sentence Generation\r\n\t\t\r\n- **PartV**: **Additional Materials**:  \r\n   - Custom Layers in Keras \r\n   - Multi modal Network Topologies with Keras\r\n\r\n---\r\n\r\n# Requirements\r\n\r\nThis tutorial requires the following packages:\r\n\r\n- Python version 3.5\r\n    - Python 3.4 should be fine as well\r\n    - likely Python 2.7 would be also fine, but *who knows*? :P\r\n    \r\n- `numpy` version 1.10 or later: http:\u002F\u002Fwww.numpy.org\u002F\r\n- `scipy` version 0.16 or later: http:\u002F\u002Fwww.scipy.org\u002F\r\n- `matplotlib` version 1.4 or later: http:\u002F\u002Fmatplotlib.org\u002F\r\n- `pandas` version 0.16 or later: http:\u002F\u002Fpandas.pydata.org\r\n- `scikit-learn` version 0.15 or later: http:\u002F\u002Fscikit-learn.org\r\n- `keras` version 2.0 or later: http:\u002F\u002Fkeras.io\r\n- `tensorflow` version 1.0 or later: https:\u002F\u002Fwww.tensorflow.org\r\n- `ipython`\u002F`jupyter` version 4.0 or later, with notebook support\r\n\r\n(Optional but recommended):\r\n\r\n- `pyyaml`\r\n- `hdf5` and `h5py` (required if you use model saving\u002Floading functions in keras)\r\n- **NVIDIA cuDNN** if you have NVIDIA GPUs on your machines.\r\n    [https:\u002F\u002Fdeveloper.nvidia.com\u002Frdp\u002Fcudnn-download]()\r\n\r\nThe easiest way to get (most) these is to use an all-in-one installer such as [Anaconda](http:\u002F\u002Fwww.continuum.io\u002Fdownloads) from Continuum. These are available for multiple architectures.\r\n\r\n---\r\n\r\n### Python Version\r\n\r\nI'm currently running this tutorial with **Python 3** on **Anaconda**\r\n\r\n\r\n```python\r\n!python --version\r\n```\r\n\r\n    Python 3.5.2\r\n\r\n---\t\r\n\t\r\n## Setting the Environment\r\n\r\nIn this repository, files to re-create virtual env with `conda` are provided for Linux and OSX systems, \r\nnamely `deep-learning.yml` and `deep-learning-osx.yml`, respectively.\r\n\r\nTo re-create the virtual environments (on Linux, for example):\r\n\r\n```shell\r\nconda env create -f deep-learning.yml\r\n```\r\n\r\nFor OSX, just change the filename, accordingly.\r\n\r\n### Notes about Installing Theano with GPU support\r\n\r\n**NOTE**: Read this section **only** if after _pip installing_ `theano`, it raises error in enabling the GPU support!\r\n\r\nSince version `0.9` Theano introduced the [`libgpuarray`](http:\u002F\u002Fdeeplearning.net\u002Fsoftware\u002Flibgpuarray) in the stable release (it was previously only available in the _development_ version).\r\n\r\nThe goal of `libgpuarray` is (_from the documentation_) make a common GPU ndarray (n dimensions array) that can be reused by all projects that is as future proof as possible, while keeping it easy to use for simple need\u002Fquick test.\r\n\r\nHere are some useful tips (hopefully) I came up with to properly install and configure `theano` on (Ubuntu) Linux with **GPU** support:\r\n\r\n1) [If you're using Anaconda] `conda install theano pygpu` should be just fine!\r\n\r\nSometimes it is suggested to install `pygpu` using the `conda-forge` channel:\r\n\r\n`conda install -c conda-forge pygpu`\r\n\r\n2) [Works with both Anaconda Python or Official CPython]\r\n\r\n* Install `libgpuarray` from source: [Step-by-step install libgpuarray user library](http:\u002F\u002Fdeeplearning.net\u002Fsoftware\u002Flibgpuarray\u002Finstallation.html#step-by-step-install-user-library)\r\n\r\n* Then, install `pygpu` from source: (in the same source folder)\r\n`python setup.py build && python setup.py install`\r\n\r\n* `pip install theano`.\r\n\r\n\r\nAfter **Theano is installed**:\r\n\r\n```\r\necho \"[global]\r\ndevice = cuda\r\nfloatX = float32\r\n\r\n[lib]\r\ncnmem = 1.0\" > ~\u002F.theanorc\r\n```\r\n\r\n### Installing Tensorflow\r\n\r\nTo date `tensorflow` comes in two different packages, namely `tensorflow` and `tensorflow-gpu`, whether you want to install \r\nthe framework with CPU-only or GPU support, respectively.\r\n\r\nFor this reason, `tensorflow` has **not** been included in the conda envs and has to be installed separately.\r\n\r\n#### Tensorflow for CPU only:\r\n\r\n```shell\r\npip install tensorflow\r\n```\r\n\r\n#### Tensorflow with GPU support:\r\n\r\n```shell\r\npip install tensorflow-gpu\r\n```\r\n\r\n**Note**: NVIDIA Drivers and CuDNN **must** be installed and configured before hand. Please refer to the official \r\n[Tensorflow documentation](https:\u002F\u002Fwww.tensorflow.org\u002Finstall\u002F) for further details.\r\n\r\n\r\n#### Important Note:\r\n\r\nAll the code provided+ in this tutorial can run even if `tensorflow` is **not** installed, and so using `theano` as the (default) backend!\r\n\r\n___**This** is exactly the power of Keras!___\r\n\r\nTherefore, installing `tensorflow` is **not** stricly required!\r\n\r\n+: Apart from the **1.2 Introduction to Tensorflow** tutorial, of course.\r\n\r\n### Configure Keras with tensorflow\r\n\r\nBy default, Keras is configured with `theano` as backend. \r\n\r\nIf you want to use `tensorflow` instead, these are the simple steps to follow:\r\n\r\n1) Create the `keras.json` (if it does not exist):\r\n\r\n```shell\r\ntouch $HOME\u002F.keras\u002Fkeras.json\r\n```\r\n\r\n2) Copy the following content into the file:\r\n\r\n```\r\n{\r\n    \"epsilon\": 1e-07,\r\n    \"backend\": \"tensorflow\",\r\n    \"floatx\": \"float32\",\r\n    \"image_data_format\": \"channels_last\"\r\n}\r\n```\r\n\r\n3) Verify it is properly configured:\r\n\r\n```python\r\n!cat ~\u002F.keras\u002Fkeras.json\r\n```\r\n\r\n    {\r\n    \t\"epsilon\": 1e-07,\r\n    \t\"backend\": \"tensorflow\",\r\n    \t\"floatx\": \"float32\",\r\n    \t\"image_data_format\": \"channels_last\"\r\n    }\r\n\r\n---\r\n\r\n# Test if everything is up&running\r\n\r\n## 1. Check import\r\n\r\n\r\n```python\r\nimport numpy as np\r\nimport scipy as sp\r\nimport pandas as pd\r\nimport matplotlib.pyplot as plt\r\nimport sklearn\r\n```\r\n\r\n\r\n```python\r\nimport keras\r\n```\r\n\r\n    Using TensorFlow backend.\r\n\r\n\r\n## 2. Check installed Versions\r\n\r\n\r\n```python\r\nimport numpy\r\nprint('numpy:', numpy.__version__)\r\n\r\nimport scipy\r\nprint('scipy:', scipy.__version__)\r\n\r\nimport matplotlib\r\nprint('matplotlib:', matplotlib.__version__)\r\n\r\nimport IPython\r\nprint('iPython:', IPython.__version__)\r\n\r\nimport sklearn\r\nprint('scikit-learn:', sklearn.__version__)\r\n```\r\n\r\n    numpy: 1.11.1\r\n    scipy: 0.18.0\r\n    matplotlib: 1.5.2\r\n    iPython: 5.1.0\r\n    scikit-learn: 0.18\r\n\r\n\r\n\r\n```python\r\nimport keras\r\nprint('keras: ', keras.__version__)\r\n\r\n# optional\r\nimport theano\r\nprint('Theano: ', theano.__version__)\r\n\r\nimport tensorflow as tf\r\nprint('Tensorflow: ', tf.__version__)\r\n```\r\n\r\n    keras:  2.0.2\r\n    Theano:  0.9.0\r\n    Tensorflow:  1.0.1\r\n\r\n\r\n\u003Cbr>\r\n\u003Ch1 style=\"text-align: center;\">If everything worked till down here, you're ready to start!\u003C\u002Fh1>\r\n","\u003Cdiv>\n    \u003Ch1 style=\"text-align: center;\">使用Keras和TensorFlow进行深度学习\u003C\u002Fh1>\n    \u003Cimg style=\"text-align: left\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fleriomaggio_deep-learning-keras-tensorflow_readme_2c18e3986a1f.jpg\" width=\"15%\" \u002F>\n\u003Cdiv>\n\u003Cbr>\n\n### 作者：瓦莱里奥·马焦\n\n#### 联系方式：\n\n\u003Ctable style=\"border: 0px; display: inline-table\">\n    \u003Ctbody>\n        \u003Ctr style=\"border: 0px;\">\n            \u003Ctd style=\"border: 0px;\">\n                \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fleriomaggio_deep-learning-keras-tensorflow_readme_da46c48e57f3.png\" style=\"display: inline-block;\" \u002F> \n                \u003Ca href=\"http:\u002F\u002Ftwitter.com\u002Fleriomaggio\" target=\"_blank\">@leriomaggio\u003C\u002Fa>\n\t    \u003C\u002Ftd>\n            \u003Ctd style=\"border: 0px;\">\n                \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fleriomaggio_deep-learning-keras-tensorflow_readme_792d827227e3.png\" style=\"display: inline-block;\" \u002F> \n                \u003Ca href=\"it.linkedin.com\u002Fin\u002Fvaleriomaggio\" target=\"_blank\">valeriomaggio\u003C\u002Fa>\n            \u003C\u002Ftd>\n            \u003Ctd style=\"border: 0px;\">\n                \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fleriomaggio_deep-learning-keras-tensorflow_readme_be8cc437ecf8.png\" style=\"display: inline-block;\" \u002F> \n                valeriomaggio_at_gmail\n            \u003C\u002Ftd>\n       \u003C\u002Ftr>\n  \u003C\u002Ftbody>\n\u003C\u002Ftable>\n\n\n```shell\n\ngit clone https:\u002F\u002Fgithub.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow.git\n```\n\n\n---\n\n## 目录\n\n- **第一部分**：**简介**\n\n    - 人工神经网络入门\n        - 感知器与多层感知器    \n        - 原生纯Python实现\n        - 前向传播、随机梯度下降、反向传播\n        \n    - 深度学习框架简介\n        - Theano入门\n        - TensorFlow入门\n        - Keras入门\n            - 概览与主要特性\n            - `core`层概述\n            - 多层感知器与全连接网络\n                - 使用`keras.models.Sequential`和`Dense`的示例\n            - Keras后端\n    \n- **第二部分**：**监督学习**\n    \n    - 全连接网络与嵌入\n        - MNIST数据集入门\n        - 隐藏层表示与嵌入\n        \n    - 卷积神经网络\n        - 卷积滤波器的意义\n            - ImageNet中的示例    \n        - 可视化卷积网络 \n\n        - 高级卷积网络\n            - Dropout\n            - 最大池化\n            - 批量归一化\n\n        - 实战：MNIST数据集\n            - FC与MNIST\n            - CNN与MNIST\n        \n        - 使用Keras的深度卷积神经网络（参考：`keras.applications`）\n            - VGG16\n            - VGG19\n            - ResNet50\n    - 迁移学习与微调\n    - 超参数优化 \n        \n- **第三部分**：**无监督学习**\n\n    - 自编码器与嵌入\n\t- 自编码器与MNIST\n    \t- word2vec与doc2vec（gensim）与`keras.datasets`\n        - word2vec与CNN\n    \n- **第四部分**：**循环神经网络**\n    - Keras中的循环神经网络 \n        - `SimpleRNN`、`LSTM`、`GRU`\n    - LSTM用于句子生成\n        \n- **第五部分**：**附加材料**：  \n   - Keras中的自定义层 \n   - 使用Keras构建多模态网络拓扑\n\n---\n\n\n# 要求\n\n本教程需要以下软件包：\n\n- Python版本3.5\n    - Python 3.4也可以\n    - 可能Python 2.7也可以，但*谁知道呢*？ :P\n    \n- `numpy`版本1.10或更高：http:\u002F\u002Fwww.numpy.org\u002F\n- `scipy`版本0.16或更高：http:\u002F\u002Fwww.scipy.org\u002F\n- `matplotlib`版本1.4或更高：http:\u002F\u002Fmatplotlib.org\u002F\n- `pandas`版本0.16或更高：http:\u002F\u002Fpandas.pydata.org\n- `scikit-learn`版本0.15或更高：http:\u002F\u002Fscikit-learn.org\n- `keras`版本2.0或更高：http:\u002F\u002Fkeras.io\n- `tensorflow`版本1.0或更高：https:\u002F\u002Fwww.tensorflow.org\n- `ipython`\u002F`jupyter`版本4.0或更高，且支持notebook\n\n（可选但推荐）：\n\n- `pyyaml`\n- `hdf5`和`h5py`（如果你使用Keras中的模型保存与加载功能，则需要）\n- 如果你的机器上有NVIDIA GPU，请安装**NVIDIA cuDNN**。\n    [https:\u002F\u002Fdeveloper.nvidia.com\u002Frdp\u002Fcudnn-download]()\n\n最简单的方法是使用像Continuum的[Anaconda](http:\u002F\u002Fwww.continuum.io\u002Fdownloads)这样的全能安装程序。这些安装程序适用于多种架构。\n\n---\n\n\n### Python版本\n\n我目前在**Anaconda**上使用**Python 3**运行本教程\n\n\n```python\n!python --version\n```\n\n\n    Python 3.5.2\n\n---\n\n\n## 设置环境\n\n在这个仓库中，提供了用于在Linux和OSX系统上用`conda`重建虚拟环境的文件，分别是`deep-learning.yml`和`deep-learning-osx.yml`。\n\n要在Linux上重建虚拟环境（例如）：\n\n```shell\nconda env create -f deep-learning.yml\n```\n\n\n对于OSX，只需相应地更改文件名即可。\n\n### 关于安装带有GPU支持的Theano的注意事项\n\n**注意**：仅当您通过`pip install theano`后，启用GPU支持时出现错误时才阅读本节！\n\n自版本`0.9`起，Theano在稳定版中引入了[`libgpuarray`](http:\u002F\u002Fdeeplearning.net\u002Fsoftware\u002Flibgpuarray)（此前仅在_开发_版本中可用）。\n\n`libgpuarray`的目标是（来自文档）创建一个通用的GPU ndarray（n维数组），可供所有项目复用，并尽可能具有未来兼容性，同时保持简单易用，适合简单需求或快速测试。\n\n以下是一些有用的技巧（希望有用），帮助你在（Ubuntu）Linux上正确安装并配置带有**GPU**支持的`theano`：\n\n1）【如果你使用Anaconda】`conda install theano pygpu`应该就足够了！\n\n有时建议使用`conda-forge`通道安装`pygpu`：\n\n`conda install -c conda-forge pygpu`\n\n2）【适用于Anaconda Python或官方CPython】\n\n* 从源码安装`libgpuarray`：[逐步安装libgpuarray用户库](http:\u002F\u002Fdeeplearning.net\u002Fsoftware\u002Flibgpuarray\u002Finstallation.html#step-by-step-install-user-library)\n\n* 然后，从源码安装`pygpu`：（在同一源码文件夹中）\n`python setup.py build && python setup.py install`\n\n* `pip install theano`。\n\n\n在**Theano安装完毕**后：\n\n```bash\necho \"[global]\ndevice = cuda\nfloatX = float32\n\n[lib]\ncnmem = 1.0\" > ~\u002F.theanorc\n```\n\n### 安装 TensorFlow\n\n截至目前，TensorFlow 提供了两种不同的软件包：`tensorflow` 和 `tensorflow-gpu`。您可以根据需要选择安装仅支持 CPU 的框架，或同时支持 CPU 和 GPU 的框架。\n\n因此，`tensorflow` 未包含在 Conda 环境中，必须单独安装。\n\n#### 仅支持 CPU 的 TensorFlow：\n\n```shell\npip install tensorflow\n```\n\n#### 支持 GPU 的 TensorFlow：\n\n```shell\npip install tensorflow-gpu\n```\n\n**注意**：在安装之前，您必须先安装并配置好 NVIDIA 驱动程序和 CuDNN。有关详细信息，请参阅官方 [TensorFlow 文档](https:\u002F\u002Fwww.tensorflow.org\u002Finstall\u002F)。\n\n#### 重要提示：\n\n本教程中提供的所有代码即使未安装 TensorFlow 也能正常运行，因此您也可以使用 Theano 作为默认后端！\n\n___**这正是 Keras 的强大之处！**___\n\n因此，安装 TensorFlow 并不是必需的！\n\n+ 但请注意，除了 **1.2 TensorFlow 简介** 教程之外。\n\n### 使用 TensorFlow 配置 Keras\n\n默认情况下，Keras 使用 Theano 作为后端。\n\n如果您希望改用 TensorFlow，只需按照以下简单步骤操作即可：\n\n1) 创建 `keras.json` 文件（如果该文件不存在）：\n\n```shell\ntouch $HOME\u002F.keras\u002Fkeras.json\n```\n\n2) 将以下内容复制到该文件中：\n\n```json\n{\n    \"epsilon\": 1e-07,\n    \"backend\": \"tensorflow\",\n    \"floatx\": \"float32\",\n    \"image_data_format\": \"channels_last\"\n}\n```\n\n3) 检查配置是否正确：\n\n```python\n!cat ~\u002F.keras\u002Fkeras.json\n```\n\n    {\n    \t\"epsilon\": 1e-07,\n    \t\"backend\": \"tensorflow\",\n    \t\"floatx\": \"float32\",\n    \t\"image_data_format\": \"channels_last\"\n    }\n\n---\n\n# 测试一切是否正常运行\n\n## 1. 检查导入\n\n```python\nimport numpy as np\nimport scipy as sp\nimport pandas as pd\nimport matplotlib.pyplot as plt\nimport sklearn\n```\n\n```python\nimport keras\n```\n\n    使用 TensorFlow 后端。\n\n\n## 2. 检查已安装的版本\n\n```python\nimport numpy\nprint('numpy:', numpy.__version__)\n\nimport scipy\nprint('scipy:', scipy.__version__)\n\nimport matplotlib\nprint('matplotlib:', matplotlib.__version__)\n\nimport IPython\nprint('iPython:', IPython.__version__)\n\nimport sklearn\nprint('scikit-learn:', sklearn.__version__)\n```\n\n    numpy: 1.11.1\n    scipy: 0.18.0\n    matplotlib: 1.5.2\n    iPython: 5.1.0\n    scikit-learn: 0.18\n\n\n```python\nimport keras\nprint('keras: ', keras.__version__)\n\n# 可选\nimport theano\nprint('Theano: ', theano.__version__)\n\nimport tensorflow as tf\nprint('Tensorflow: ', tf.__version__)\n```\n\n    keras:  2.0.2\n    Theano:  0.9.0\n    Tensorflow:  1.0.1\n\n\n\u003Cbr>\n\u003Ch1 style=\"text-align: center;\">如果到这里一切顺利，您就可以开始啦！\u003C\u002Fh1>","# Deep Learning with Keras and TensorFlow 快速上手指南\n\n## 环境准备\n\n- **Python 版本**：3.5 或以上（推荐使用 Python 3.8+）\n- **推荐环境**：使用 [Anaconda](https:\u002F\u002Fwww.anaconda.com\u002Fproducts\u002Fdistribution)（国内可从 [清华镜像](https:\u002F\u002Fmirrors.tuna.tsinghua.edu.cn\u002Fanaconda\u002F) 下载）\n- **必需依赖**：\n  - numpy ≥ 1.10\n  - scipy ≥ 0.16\n  - matplotlib ≥ 1.4\n  - pandas ≥ 0.16\n  - scikit-learn ≥ 0.15\n  - Keras ≥ 2.0\n  - TensorFlow ≥ 1.0\n- **可选加速**（有 NVIDIA GPU）：\n  - 安装 NVIDIA 驱动 + cuDNN（[官网](https:\u002F\u002Fdeveloper.nvidia.com\u002Fcudnn)）\n  - 使用 `tensorflow-gpu` 替代 `tensorflow`\n\n> 推荐使用 Conda 管理环境，避免手动安装复杂依赖。\n\n## 安装步骤\n\n1. **克隆项目仓库**\n\n```shell\ngit clone https:\u002F\u002Fgithub.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow.git\ncd deep-learning-keras-tensorflow\n```\n\n2. **创建虚拟环境（Linux\u002FmacOS）**\n\n```shell\nconda env create -f deep-learning.yml\n```\n\n> macOS 用户请使用 `deep-learning-osx.yml` 文件。\n\n3. **激活环境**\n\n```shell\nconda activate deep-learning\n```\n\n4. **安装 TensorFlow（根据硬件选择）**\n\n- **CPU 版本**：\n```shell\npip install tensorflow\n```\n\n- **GPU 版本**（需提前安装 cuDNN）：\n```shell\npip install tensorflow-gpu\n```\n\n5. **配置 Keras 使用 TensorFlow 后端**\n\n```shell\nmkdir -p $HOME\u002F.keras\necho '{\n    \"epsilon\": 1e-07,\n    \"backend\": \"tensorflow\",\n    \"floatx\": \"float32\",\n    \"image_data_format\": \"channels_last\"\n}' > $HOME\u002F.keras\u002Fkeras.json\n```\n\n## 基本使用\n\n运行最简单的 MNIST 分类示例：\n\n```python\nimport keras\nfrom keras.models import Sequential\nfrom keras.layers import Dense\nfrom keras.datasets import mnist\n\n# 加载数据\n(x_train, y_train), (x_test, y_test) = mnist.load_data()\nx_train = x_train.reshape(60000, 784).astype('float32') \u002F 255\nx_test = x_test.reshape(10000, 784).astype('float32') \u002F 255\ny_train = keras.utils.to_categorical(y_train, 10)\ny_test = keras.utils.to_categorical(y_test, 10)\n\n# 构建模型\nmodel = Sequential()\nmodel.add(Dense(512, activation='relu', input_shape=(784,)))\nmodel.add(Dense(10, activation='softmax'))\n\n# 编译与训练\nmodel.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])\nmodel.fit(x_train, y_train, batch_size=128, epochs=5, verbose=1)\n\n# 评估\nscore = model.evaluate(x_test, y_test, verbose=0)\nprint('Test accuracy:', score[1])\n```\n\n> 运行后将自动下载 MNIST 数据集，首次运行可能耗时较长，建议使用国内镜像加速（如清华源）。\n\n完成后，你已成功运行第一个 Keras + TensorFlow 模型！","某中型电商公司数据团队希望在不增加人力成本的前提下，快速上线一个商品图像分类系统，用于自动识别用户上传的服装图片（如T恤、裤子、外套），以提升客服工单处理效率。团队仅有2名工程师，熟悉Python但无深度学习经验。\n\n### 没有 deep-learning-keras-tensorflow 时\n- 需要从零手写神经网络结构，调试前向传播与反向传播逻辑，耗时超过3周仍无法收敛\n- 无法高效加载和预处理MNIST之外的图像数据集，图像缩放、归一化、数据增强全靠手动编码，错误频发\n- 想尝试预训练模型如VGG16时，需手动下载权重、配置层结构、处理输入维度，文档零散，调试成本极高\n- 没有统一的训练循环模板，每次修改超参数（如学习率、批量大小）都要重写训练脚本，迭代效率极低\n- 模型训练完成后无法保存或加载，每次重启服务都要重新训练，无法部署到生产环境\n\n### 使用 deep-learning-keras-tensorflow 后\n- 仅用10行代码即可用 `Sequential()` 搭建包含卷积层、Dropout和全连接层的CNN，3天内完成原型开发\n- 直接调用 `keras.preprocessing.image.ImageDataGenerator` 实现自动图像增强与批量加载，无需手动编写数据管道\n- 通过 `keras.applications.VGG16(weights='imagenet')` 一键加载预训练模型，冻结前几层后微调，准确率从62%提升至91%\n- 使用 `model.compile(optimizer='adam', loss='categorical_crossentropy')` 一行配置训练参数，支持回调函数自动保存最佳模型\n- 训练完成后调用 `model.save('clothing_classifier.h5')` 即可持久化模型，通过Flask封装为API，3小时内上线服务\n\ndeep-learning-keras-tensorflow 让非专业AI团队也能在一周内从零构建、训练并部署高精度图像分类系统，彻底释放了深度学习的落地潜力。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fleriomaggio_deep-learning-keras-tensorflow_2c18e398.jpg","leriomaggio","Valerio Maggio","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fleriomaggio_1da2d0b3.jpg","Data Scientist and Community Advocate.",null,"Bristol, UK","https:\u002F\u002Fgithub.com\u002Fleriomaggio",[83,87],{"name":84,"color":85,"percentage":86},"Jupyter Notebook","#DA5B0B",99.7,{"name":88,"color":89,"percentage":90},"Python","#3572A5",0.3,2972,1260,"2026-03-28T00:21:47","MIT","Linux, macOS","可选 NVIDIA GPU，需安装 cuDNN，未说明具体显卡型号和显存大小，CUDA 版本未明确","未说明",{"notes":99,"python":100,"dependencies":101},"建议使用 Anaconda 或 conda 管理环境，通过 deep-learning.yml 文件创建虚拟环境；TensorFlow 可选安装 tensorflow 或 tensorflow-gpu，Keras 默认后端为 Theano，如需使用 TensorFlow 需手动配置 keras.json；安装 GPU 支持前必须预先安装 NVIDIA 驱动和 cuDNN；部分代码可在无 TensorFlow 环境下运行，依赖 Theano 后端","3.5",[102,103,104,105,106,107,108,109,110,111],"numpy>=1.10","scipy>=0.16","matplotlib>=1.4","pandas>=0.16","scikit-learn>=0.15","keras>=2.0","tensorflow>=1.0","ipython>=4.0","jupyter>=4.0","h5py",[13],[114,115,116,117,118,119,120,121,122,123],"tensorflow","python","tutorial","deep-learning","keras","keras-tutorials","keras-tensorflow","cudnn","theano","anaconda","2026-03-27T02:49:30.150509","2026-04-06T07:13:49.471674",[127,132,137,142,147,152,157],{"id":128,"question_zh":129,"answer_zh":130,"source_url":131},9056,"如何正确配置 Theano 以支持 GPU 并避免配置错误？","避免在 theanorc 文件中使用注释（如 device = gpu # switch to cpu...），因为 Theano 会尝试解析注释内容导致报错。正确做法是：1) 使用 Anaconda 安装：conda install theano pygpu；2) 或从源码安装 libgpuarray 和 pygpu，再执行 pip install theano。确保 theanorc 中仅保留有效配置项，如 [global] floatX = float32 和 device = gpu。","https:\u002F\u002Fgithub.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fissues\u002F3",{"id":133,"question_zh":134,"answer_zh":135,"source_url":136},9057,"导入 Keras 时出现 ValueError: invalid literal for int() with base 10: '  # switch to cpu...' 错误如何解决？","该错误是由于 theanorc 配置文件中 device 行包含注释（如 'gpu  # switch to cpu...'）导致 Theano 解析失败。解决方案是删除注释，仅保留纯配置：将 [global] device = gpu  # ... 改为 device = gpu。参考 Issue #3 的解决方案，确保配置文件无任何注释或多余文本。","https:\u002F\u002Fgithub.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fissues\u002F5",{"id":138,"question_zh":139,"answer_zh":140,"source_url":141},9058,"在 macOS 上使用 Matplotlib 时出现 'Python is not installed as a framework' 错误怎么办？","避免此问题的最佳方法是使用 Anaconda Python，它已预配置为框架模式。若已使用系统 Python，可通过 brew install python --framework 重新安装，或在代码中设置后端：import matplotlib; matplotlib.use('Agg')。推荐使用 Anaconda 环境以避免此类系统级依赖问题。","https:\u002F\u002Fgithub.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fissues\u002F16",{"id":143,"question_zh":144,"answer_zh":145,"source_url":146},9059,"使用 pydatait 分支的 deep-learning.yml 安装时出现 pygpu==0.2.1 无法找到的错误如何解决？","pygpu==0.2.1 在默认通道中可能不可用。解决方案是：1) 从 conda-forge 通道安装：conda install -c conda-forge pygpu；2) 或移除 pygpu 依赖，改用 TensorFlow 作为 Keras 后端；3) 手动安装 libgpuarray 后再安装 pygpu。建议使用 master 分支的 yml 文件，其中已移除对 pygpu 的硬性依赖。","https:\u002F\u002Fgithub.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fissues\u002F12",{"id":148,"question_zh":149,"answer_zh":150,"source_url":151},9060,"如何正确创建 Conda 环境以避免 'conda create env' 命令错误？","正确的命令是 conda env create -f deep-learning.yml，而不是 conda create env。注意是 'env create' 而非 'create env'。同时确保使用 -f 参数指定 yml 文件路径。该错误已在 README 中修复，建议始终使用官方文档中的命令格式。","https:\u002F\u002Fgithub.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fissues\u002F1",{"id":153,"question_zh":154,"answer_zh":155,"source_url":156},9061,"Jupyter Notebook 中 Markdown 的内联图片在 GitHub 上不显示如何修复？","该问题已通过 PR 修复，确保图片路径为相对路径且文件已正确提交到仓库。在 GitHub 上显示图片时，需使用完整 URL 或确保图片位于仓库的 assets\u002F 或 images\u002F 目录下，并使用标准 Markdown 语法：![描述](路径)。建议使用绝对链接或上传图片至 GitHub 仓库资源目录。","https:\u002F\u002Fgithub.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fissues\u002F13",{"id":158,"question_zh":159,"answer_zh":160,"source_url":161},9062,"在神经网络中，权重更新为何使用隐藏层激活值（self.ah[j]）？","权重更新公式 change = output_deltas[k] * self.ah[j] 是反向传播的标准实现：output_deltas[k] 是输出层误差梯度，self.ah[j] 是隐藏层第 j 个神经元的激活值，二者相乘得到权重梯度。这是基于链式法则的数学推导，用于计算损失函数对权重的偏导数，而非仅依赖权重本身。","https:\u002F\u002Fgithub.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fissues\u002F9",[163,168,173,178,183,188],{"id":164,"version":165,"summary_zh":166,"released_at":167},106511,"webvalley-2017","\u003Cdiv>\r\n    \u003Ch1 style=\"text-align: center;\">Deep Learning with Keras and Tensorflow\u003C\u002Fh1>\r\n    \u003Cimg style=\"text-align: left\" src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Freleases\u002Fkeras-tensorflow-logo.jpg\" width=\"40%\" \u002F>\r\n\u003Cdiv>\r\n\r\n\u003Cdiv>\r\n    \u003Cimg style=\"text-align: left\" src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Freleases\u002Fwebvalley2017.png\" width=\"40%\" alt=\"WebValley 2017 Logo\" \u002F>\r\n\u003C\u002Fdiv>    \r\n\r\n### Valerio Maggio: _PostDoc Data Scientist @ FBK\u002FMPBA_\r\n\r\n### Contacts:\r\n\r\n\u003Ctable style=\"border: 0px; display: inline-table\">\r\n    \u003Ctbody>\r\n        \u003Ctr style=\"border: 0px;\">\r\n            \u003Ctd style=\"border: 0px;\">\r\n                \u003Cimg src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Ftwitter_small.png\" style=\"display: inline-block;\" width=\"20%\" \u002F> @leriomaggio\r\n            \u003C\u002Ftd>\r\n            \u003Ctd style=\"border: 0px;\">\r\n                \u003Cimg src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Fgmail_small.png\" style=\"display: inline-block;\" width=\"20%\" \u002F> vmaggio@fbk.eu\r\n            \u003C\u002Ftd>\r\n       \u003C\u002Ftr>\r\n  \u003C\u002Ftbody>\r\n\u003C\u002Ftable>\r\n\r\n## Installed Versions\r\n\r\n```python\r\nimport keras\r\nprint('keras: ', keras.__version__)\r\n\r\n# optional\r\nimport theano\r\nprint('Theano: ', theano.__version__)\r\n\r\nimport tensorflow as tf\r\nprint('Tensorflow: ', tf.__version__)\r\n```\r\n\r\n    keras:  2.0.4\r\n    Theano:  0.9.0\r\n    Tensorflow:  1.2.1\r\n\r\n## Outline\r\n\r\n- **Part I**: **Introduction**\r\n\r\n    - Intro to Artificial Neural Networks\r\n        - Perceptron and MLP    \r\n        - naive pure-Python implementation\r\n        - fast forward, sgd, backprop\r\n        \r\n    - Introduction to Deep Learning Frameworks\r\n        - Intro to Theano\r\n        - Intro to Tensorflow\r\n        - Intro to Keras\r\n            - Overview and main features\r\n            - Overview of the `core` layers\r\n            - Multi-Layer Perceptron and Fully Connected\r\n                - Examples with `keras.models.Sequential` and `Dense`\r\n            - Keras Backend\r\n    \r\n- **Part II**: **Supervised Learning**\r\n    \r\n    - Fully Connected Networks and Embeddings\r\n        - Intro to MNIST Dataset\r\n        - Hidden Leayer Representation and Embeddings\r\n        \r\n    - Convolutional Neural Networks\r\n        - meaning of convolutional filters\r\n            - examples from ImageNet    \r\n        - Visualising ConvNets \r\n\r\n        - Advanced CNN\r\n            - Dropout\r\n            - MaxPooling\r\n            - Batch Normalisation\r\n\r\n        - HandsOn: MNIST Dataset\r\n            - FC and MNIST\r\n            - CNN and MNIST\r\n        \r\n        - Deep Convolutiona Neural Networks with Keras (ref: `keras.applications`)\r\n            - VGG16\r\n            - VGG19\r\n            - ResNet50\r\n\r\n    - Transfer Learning and FineTuning\r\n    - Hyperparameters Optimisation \r\n        \r\n- **Part III**: **Unsupervised Learning**\r\n\r\n    - AutoEncoders and Embeddings\r\n\t- AutoEncoders and MNIST\r\n    \t- word2vec and doc2vec (gensim) with `keras.datasets`\r\n        - word2vec and CNN\r\n    \r\n- **Part IV**: **Recurrent Neural Networks**\r\n    - Recurrent Neural Network in Keras \r\n        -  `SimpleRNN`, `LSTM`, `GRU`\r\n    - LSTM for Sentence Generation\r\n\t\t\r\n- **PartV**: **Additional Materials**:  \r\n   - Custom Layers in Keras \r\n   - Multi modal Network Topologies with Keras\r\n","2017-08-22T13:13:05",{"id":169,"version":170,"summary_zh":171,"released_at":172},106512,"pyss2016","\u003Cdiv>\r\n    \u003Ch1 style=\"text-align: center;\">Deep Learning with Keras and Tensorflow\u003C\u002Fh1>\r\n    \u003Cimg style=\"text-align: left\" src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Freleases\u002Fkeras-tensorflow-logo.jpg\" width=\"40%\" \u002F>\r\n\u003Cdiv>\r\n\r\n\u003Cdiv>\r\n    \u003Cimg style=\"text-align: left\" src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Freleases\u002Fpyss2016_logo.png\" width=\"40%\" alt=\"PySS 2016 Logo\" \u002F>\r\n\u003C\u002Fdiv>    \r\n\r\n### Valerio Maggio: _PostDoc Data Scientist @ FBK\u002FMPBA_\r\n\r\n### Contacts:\r\n\r\n\u003Ctable style=\"border: 0px; display: inline-table\">\r\n    \u003Ctbody>\r\n        \u003Ctr style=\"border: 0px;\">\r\n            \u003Ctd style=\"border: 0px;\">\r\n                \u003Cimg src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Ftwitter_small.png\" style=\"display: inline-block;\" width=\"20%\" \u002F> @leriomaggio\r\n            \u003C\u002Ftd>\r\n            \u003Ctd style=\"border: 0px;\">\r\n                \u003Cimg src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Fgmail_small.png\" style=\"display: inline-block;\" width=\"20%\" \u002F> vmaggio@fbk.eu\r\n            \u003C\u002Ftd>\r\n       \u003C\u002Ftr>\r\n  \u003C\u002Ftbody>\r\n\u003C\u002Ftable>\r\n\r\n\r\n# Goal of this Tutorial\r\n\r\n- **Introduce** main features of Keras\r\n    - Plus some introductory overview of Tensorflow\r\n    \r\n- **Learn** how simple and Pythonic is doing Deep Learning with Keras\r\n\r\n- **Understand** how easy is to do basic and *advanced* Deep Learning models in Keras;\r\n    - **Examples and Hand-on Excerises** along the way.\r\n\r\n## Installed Versions\r\n\r\n```python\r\nimport keras\r\nprint('keras: ', keras.__version__)\r\n\r\n# optional\r\nimport theano\r\nprint('Theano: ', theano.__version__)\r\n\r\nimport tensorflow as tf\r\nprint('Tensorflow: ', tf.__version__)\r\n```\r\n\r\n    keras:  1.0.7\r\n    Theano:  0.8.2\r\n    Tensorflow:  0.10.0","2017-08-22T07:54:37",{"id":174,"version":175,"summary_zh":176,"released_at":177},106513,"pydata-london2017","\u003Cdiv>\r\n    \u003Ch1 style=\"text-align: center;\">Deep Learning with Keras and Tensorflow\u003C\u002Fh1>\r\n    \u003Cimg style=\"text-align: left\" src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Freleases\u002Fkeras-tensorflow-logo.jpg\" width=\"40%\" \u002F>\r\n\u003Cdiv>\r\n\r\n\u003Cdiv>\r\n    \u003Cimg style=\"text-align: left\" src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Freleases\u002Fpydata_london.png\" width=\"40%\" alt=\"PyData London 2016 Logo\" \u002F>\r\n\u003C\u002Fdiv>    \r\n\r\n### Valerio Maggio: _PostDoc Data Scientist @ FBK\u002FMPBA_\r\n\r\n### Contacts:\r\n\r\n\u003Ctable style=\"border: 0px; display: inline-table\">\r\n    \u003Ctbody>\r\n        \u003Ctr style=\"border: 0px;\">\r\n            \u003Ctd style=\"border: 0px;\">\r\n                \u003Cimg src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Ftwitter_small.png\" style=\"display: inline-block;\" width=\"20%\" \u002F> @leriomaggio\r\n            \u003C\u002Ftd>\r\n            \u003Ctd style=\"border: 0px;\">\r\n                \u003Cimg src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Fgmail_small.png\" style=\"display: inline-block;\" width=\"20%\" \u002F> vmaggio@fbk.eu\r\n            \u003C\u002Ftd>\r\n       \u003C\u002Ftr>\r\n  \u003C\u002Ftbody>\r\n\u003C\u002Ftable>\r\n\r\n### Library Versions\r\n\r\n```python\r\nimport keras\r\nprint('keras: ', keras.__version__)\r\n\r\n# optional\r\nimport theano\r\nprint('Theano: ', theano.__version__)\r\n\r\nimport tensorflow as tf\r\nprint('Tensorflow: ', tf.__version__)\r\n```\r\n\r\n    keras:  2.0.2\r\n    Theano:  0.9.0\r\n    Tensorflow:  1.0.1\r\n\r\n### Goal\r\n\r\n- Introduce main features of Keras APIs to build Neural Networks.\r\n- Learn how to implement simple and complex Deep Neural Networks Architectures using Keras.\r\n- Discover Keras Implementation and Internals.\r\n    - Note: examples and hands-on exercises will be provided along the way.\r\n\r\n\r\n### Outline in Ten (\\~ish) Notebooks\r\n\r\n1. Multi-layer Fully Connected Networks (and the backends)\r\n2. Hidden Layers features and Embeddings\r\n3. Convolutional Networks\r\n4. Hyperparameter Tuning\r\n5. Cutsom Layers\r\n6. Deep CNN and Residual Networks\r\n7. Transfer Learning and Fine Tuning\r\n8. Recursive Neural Networks\r\n9. AutoEncoders\r\n10. Multi-Modal Networks\r\n","2017-08-22T12:34:35",{"id":179,"version":180,"summary_zh":181,"released_at":182},106514,"pydata-florence2017","\u003Cdiv>\r\n    \u003Ch1 style=\"text-align: center;\">Deep Learning with Keras and Tensorflow\u003C\u002Fh1>\r\n    \u003Cimg style=\"text-align: left\" src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Freleases\u002Fkeras-tensorflow-logo.jpg\" width=\"40%\" \u002F>\r\n\u003Cdiv>\r\n\r\n\u003Cdiv>\r\n    \u003Cimg style=\"text-align: left\" src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Freleases\u002Fpydata_florence.png\" width=\"40%\" alt=\"PyData Florence 2016 Logo\" \u002F>\r\n\u003C\u002Fdiv>    \r\n\r\n### Valerio Maggio: _PostDoc Data Scientist @ FBK\u002FMPBA_\r\n\r\n### Contacts:\r\n\r\n\u003Ctable style=\"border: 0px; display: inline-table\">\r\n    \u003Ctbody>\r\n        \u003Ctr style=\"border: 0px;\">\r\n            \u003Ctd style=\"border: 0px;\">\r\n                \u003Cimg src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Ftwitter_small.png\" style=\"display: inline-block;\" width=\"20%\" \u002F> @leriomaggio\r\n            \u003C\u002Ftd>\r\n            \u003Ctd style=\"border: 0px;\">\r\n                \u003Cimg src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Fgmail_small.png\" style=\"display: inline-block;\" width=\"20%\" \u002F> vmaggio@fbk.eu\r\n            \u003C\u002Ftd>\r\n       \u003C\u002Ftr>\r\n  \u003C\u002Ftbody>\r\n\u003C\u002Ftable>\r\n\r\n### Library Versions\r\n\r\n```python\r\nimport keras\r\nprint('keras: ', keras.__version__)\r\n\r\n# optional\r\nimport theano\r\nprint('Theano: ', theano.__version__)\r\n\r\nimport tensorflow as tf\r\nprint('Tensorflow: ', tf.__version__)\r\n```\r\n\r\n    keras:  2.0.2\r\n    Theano:  0.9.0\r\n    Tensorflow:  1.0.1\r\n\r\n\r\n### Outline at a glance\r\n\r\n- **Part I**: **Introduction to ANN using Tensorflow and Keras**\r\n\r\n    - naive pure-Python implementation\r\n    - fast forward, sgd, backprop\r\n    - Model + SGD using Tensorflow\r\n    - Introduction to Keras main features\r\n        - `keras.layers.core.Dense`\r\n        - `keras.backend`\r\n        - Multi-Layer Perceptron and Fully Connected Networks\r\n\r\n- **Part II**: **Supervised Learning and Convolutional Neural Nets**\r\n    \r\n    - Intro: Focus on Image Classification\r\n    - Intro to ConvNets\r\n    - Advanced CNN\r\n        - Dropout\r\n        - MaxPooling\r\n        - Batch Normalisation\r\n    - Famous Models in Keras\r\n        (ref: `keras.applications`)\r\n    - Transfer Learning\r\n        \r\n- **Part III**: **Unsupervised Learning**\r\n\r\n    - AutoEncoders\r\n    - word2vec & doc2vec (gensim) & `keras.datasets`\r\n        - `Embedding`\r\n\r\n- **Part IV**: **Additional Materials**\r\n\r\n    -  Recurrent Neural Networks: RNN, LSTM, GRU  \r\n    - HandsOn: IMDB\r\n    - Multi-Input\u002FMulti-Output Network Topologies\r\n","2017-08-22T12:18:36",{"id":184,"version":185,"summary_zh":186,"released_at":187},106515,"euroscipy2016","\u003Cdiv>\r\n    \u003Ch1 style=\"text-align: center;\">Deep Learning with Keras\u003C\u002Fh1>\r\n    \u003Cimg style=\"text-align: left\" src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Fkeras-logo-small.jpg\" width=\"10%\" \u002F>\r\n\u003Cdiv>\r\n\r\n\u003Cdiv>\r\n    \u003Ch2 style=\"text-align: center;\">Tutorial @ EuroScipy 2016\u003C\u002Fh2>\r\n    \u003Cimg style=\"text-align: left\" src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Freleases\u002Feuroscipy_2016_logo.png\" width=\"40%\" \u002F>\r\n\u003C\u002Fdiv>\r\n\r\n## Authors: Yam Peleg, Valerio Maggio\r\n\r\n- **Introduce** main features of Keras\r\n- **Learn** how simple and Pythonic is doing Deep Learning with Keras\r\n- **Understand** how easy is to do basic and *advanced* DL models in Keras;\r\n    - **Examples and Hand-on Excerises** along the way.\r\n\r\n### Installed Versions\r\n\r\n```python\r\nimport keras\r\nprint('keras: ', keras.__version__)\r\n\r\n# optional\r\nimport theano\r\nprint('Theano: ', theano.__version__)\r\n\r\nimport tensorflow as tf\r\nprint('Tensorflow: ', tf.__version__)\r\n```\r\n\r\n    keras:  1.0.7\r\n    Theano:  0.8.2\r\n    Tensorflow:  0.10.0rc0\r\n","2017-08-22T09:07:29",{"id":189,"version":190,"summary_zh":191,"released_at":192},106516,"biforum2016","\u003Cdiv>\r\n    \u003Ch1 style=\"text-align: center;\">Deep Learning with Keras and Tensorflow\u003C\u002Fh1>\r\n    \u003Cimg style=\"text-align: left\" src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Freleases\u002Fkeras-tensorflow-logo.jpg\" width=\"40%\" \u002F>\r\n\u003Cdiv>\r\n\u003Cbr\u002F>\r\n\u003Cdiv>\r\n    \u003Cimg style=\"text-align: left\" src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Freleases\u002FBIforum.PNG\" width=\"40%\" alt=\"Bi Forum 2016 Logo\" \u002F>\r\n \u003C\u002Fdiv>    \r\n\r\n### Valerio Maggio: _PostDoc Data Scientist @ FBK\u002FMPBA_\r\n\r\n### Contacts:\r\n\r\n\u003Ctable style=\"border: 0px; display: inline-table\">\r\n    \u003Ctbody>\r\n        \u003Ctr style=\"border: 0px;\">\r\n            \u003Ctd style=\"border: 0px;\">\r\n                \u003Cimg src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Ftwitter_small.png\" style=\"display: inline-block;\" width=\"20%\" \u002F> @leriomaggio\r\n            \u003C\u002Ftd>\r\n            \u003Ctd style=\"border: 0px;\">\r\n                \u003Cimg src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Fleriomaggio\u002Fdeep-learning-keras-tensorflow\u002Fmaster\u002Fimgs\u002Fgmail_small.png\" style=\"display: inline-block;\" width=\"20%\" \u002F> vmaggio@fbk.eu\r\n            \u003C\u002Ftd>\r\n       \u003C\u002Ftr>\r\n  \u003C\u002Ftbody>\r\n\u003C\u002Ftable>\r\n\r\n### Library Versions\r\n\r\n```python\r\nimport keras\r\nprint('keras: ', keras.__version__)\r\n\r\n# optional\r\nimport theano\r\nprint('Theano: ', theano.__version__)\r\n\r\nimport tensorflow as tf\r\nprint('Tensorflow: ', tf.__version__)\r\n```\r\n\r\n    keras:  1.0.7\r\n    Theano:  0.8.2\r\n    Tensorflow:  0.10.0\r\n\r\n\r\n# Goal of this Tutorial\r\n\r\n- **Introduce** main features of Keras\r\n    - Plus some introductory overview of Tensorflow\r\n    \r\n- **Learn** how simple and Pythonic is doing Deep Learning with Keras\r\n\r\n- **Understand** how easy is to do basic and *advanced* Deep Learning models in Keras;\r\n    - **Examples and Hand-on Excerises** along the way.","2017-08-22T07:49:04"]