从零基础到进阶的系统化 AI 学习路径与课程，共 24 条。
面向中文读者的深度学习交互式教材：概念、数学、可运行代码三位一体，从线性回归一路讲到注意力机制与优化算法，配套 B 站视频课程，被国内高校广泛采用。An interactive deep learning book for Chinese readers: concepts, math and runnable code in one place, from linear regression to attention, with companion video courses.
斯坦福吴恩达的机器学习旗舰课程新版本：用 Python 与 scikit-learn 教学，覆盖监督学习、聚类、异常检测、推荐系统与决策树，配套 Labs 与测验，是全球公认的 ML 第一课。Andrew Ng's updated Machine Learning Specialization: Python-based, covering supervised learning, clustering, anomaly detection and recommender systems with hands-on labs.
前 Tesla AI 总监 Andrej Karpathy 的「从零到一」系列：从反向传播手推开始，逐步用 PyTorch 手写 micrograd、makemore、GPT，最终从零实现一个类 GPT-2 的语言模型——理解 Transformer 内部原理的最佳路径。Andrej Karpathy's Zero-to-Hero series: from hand-deriving backprop to building GPT-2 from scratch in PyTorch — the best path to truly understand Transformers.
GitHub 40k+ star 的大模型学习路线图：按 Fundamentals → LLM Scientist → LLM Engineer 三阶段组织，每步附论文、Colab 与资源清单，从原理到量化部署全链路覆盖。A 40k+ star LLM learning roadmap organized in three tiers: Fundamentals, LLM Scientist and LLM Engineer, with papers and Colabs for every step.
Hugging Face 官方学习中心：Transformers、Diffusion、RLHF、Agents 等多门免费课程，全部配套在线可运行环境，是上手 transformer 生态事实标准工具链的官方路径。The official Hugging Face learning hub: free courses on Transformers, Diffusion, RLHF and Agents with runnable notebooks.
Datawhale 出品的中文 LLM 应用开发教程：从 API 调用到 Prompt 工程、RAG 知识库、Vector 检索、LangChain 应用与部署，全程围绕一个「知识库助手」项目实战。Datawhale's Chinese LLM application course: from API calls to prompt engineering, RAG, vector search and LangChain, built around one knowledge-base assistant project.
Jeremy Howard 的实战深度学习课程：第一节课就训练出可用模型，再逐步深入原理。理念是「先跑起来再理解」，配套 fastai 库把最佳实践封装到几行代码。Jeremy Howard's top-down deep learning course: train a working model in lesson one, then dig into how it works. Comes with the fastai library.
PyTorch 官方教程中心：从 60 分钟闪电入门到训练技巧、部署、量化，全链路示例代码可直接在 Colab 运行，学框架的第一手资料。The official PyTorch tutorial hub: from the 60-minute blitz to training recipes, deployment and quantization, all runnable in Colab.
斯坦福研究生级机器学习课程：严格数学推导 + 精炼 notes，覆盖 GLM、SVM、学习理论、EM、强化学习，吴恩达主讲版本是经典中的经典，适合打深基础。Stanford's graduate-level ML course with rigorous math derivations and concise notes, covering GLMs, SVMs, learning theory, EM and RL.
斯坦福 NLP 旗舰课：从词向量到 Transformer、预训练与提示，Manning 主讲，几乎是大模型时代的「前置必修课」，讲义与作业质量双高。Stanford's NLP flagship course from word vectors to Transformers and prompting, taught by Christopher Manning — the prerequisite course for the LLM era.
斯坦福王牌 CV 课程：从图像分类与 kNN 讲到 CNN、检测、分割、3D 与生成模型，课程 notes（中文翻译版流传极广）质量极高，Fei-Fei Li 团队出品。Stanford's flagship computer vision course: from image classification to CNNs, detection, segmentation and generative models, with legendary course notes.
Datawhale 的开源大模型部署手册：环境配置、Llama/Qwen/ChatGLM 等主流开源模型的部署、推理、LoRA 微调全流程，中文环境踩坑指南性质极强。A Chinese hands-on guide to deploying open-source LLMs: environment setup, inference and LoRA fine-tuning for Llama, Qwen, ChatGLM and more.
微软官方 AI Agent 入门课程（11 课，中文讲义）：从工具使用、RAG、规划到多 Agent 设计模式，带代码作业。Microsoft's official beginner course (11 lessons, Chinese notes) covering tool use, RAG, planning and multi-agent design patterns.
周志华《机器学习》：国内 ML 教材的事实标准，覆盖十余种经典算法，语言凝练、体系完整；配合《南瓜书》的公式推导更佳，是考研与面试的高频出处。Zhou Zhihua's Machine Learning (the 'Watermelon Book'): the de-facto standard ML textbook in China, covering classic algorithms with rigor.
吴恩达与 Isabel Kloumann 合作的提示词工程短课：总结、推断、转换、扩展四大范式，配套 Jupyter 实操，两小时建立提示词的工程化思维。A short prompt engineering course by Andrew Ng: four paradigms — summarize, infer, transform, expand — with hands-on Jupyter notebooks.
中文世界最系统的免费提示词教程：从基础结构、进阶技巧到 ChatGPT/ Midjourney 实战，配套可运行示例，适合零基础系统入门。The most systematic free Chinese prompt-engineering tutorial, from basics to practical ChatGPT and Midjourney workflows.
李航《统计学习方法（第二版）》：以统计视角严格推导感知机、SVM、EM、CRF 等经典模型，篇幅精炼、推导完整，与西瓜书并称国内 ML 双璧。Li Hang's Statistical Learning Methods: rigorous derivations of perceptron, SVM, EM and CRF from a statistical perspective; concise and complete.
谷歌官方机器学习速成课：中文界面、约 15 小时、40+ 可视化交互练习与 pandas 实操，节奏快、覆盖广，非常适合快速建立全局认知。Google's official ML Crash Course: Chinese-supported, ~15 hours, 40+ interactive visual exercises — a fast way to see the whole picture.
吴恩达团队的免费短课矩阵：Agent、MCP、RAG、评估、多模态等主题持续更新，每门 1-2 小时可完成，配套 Notebook 实战。Free short courses from Andrew Ng's team covering agents, MCP, RAG, evals and more, each with hands-on notebooks.
roadmap.sh 的 AI 工程师成长路线：从编程基础、数据/模型概念到 LLM 应用、RAG、Agent、评估部署，一张交互式地图看清学习顺序与技能树全貌。An interactive AI Engineer roadmap on roadmap.sh: from programming fundamentals to LLM apps, RAG, agents and evaluation.
DataTalksClub 免费MLOps 实战训练营：实验跟踪、编排、部署、监控、最佳工程实践，9 周做一个端到端项目，从 notebook 走向生产系统的完整训练。A free 9-week MLOps bootcamp: experiment tracking, orchestration, deployment and monitoring — build an end-to-end production ML system.
OpenAI 出品的深度强化学习入门：先讲清 RL 数学符号与算法家族（PPO/SAC 等），再给高质量参考实现，是公认的 Deep RL 第一站。OpenAI's introduction to deep reinforcement learning: clear math notation, algorithm families (PPO, SAC) and high-quality reference implementations.
中文强化学习实践教程：从 MDP 基础到 DQN、策略梯度、SAC，理论与代码交替推进，每章附可运行 Jupyter，适合中文读者入门 RL。A Chinese hands-on RL tutorial from MDP basics to DQN, policy gradients and SAC, with runnable notebooks per chapter.
Hugging Face 官方免费 MCP 课程：从协议原理到开发自己的 MCP Server 与 Agent 集成，配套实战单元。Hugging Face's free official MCP course: protocol fundamentals to building your own MCP servers and agent integrations.