Start Your Deep Learning Journey
This site aggregates the world's top free courses, books, and practice platforms to help you systematically master the core knowledge of Machine Learning and Deep Learning, from scratch to the frontier.
Learning Flow Overview
From a solid foundation to the cutting edge, this is a progressive, interconnected journey. Follow this flowchart to systematically build your knowledge base.
Step 1: Math & Programming Foundations
Master linear algebra, calculus, probability, and Python programming. This is the absolute cornerstone for all subsequent learning.
Step 2: Classic Machine Learning
Learn core algorithms like supervised and unsupervised learning. Understand the statistical principles of model evaluation and selection.
Step 3: Deep Learning Core
Dive into architectures like Neural Networks, CNNs, and RNNs. Build your first deep learning model through code-first courses.
Step 4: Practice & Specialization
Participate in projects on platforms like Kaggle and Hugging Face. Choose a direction to delve deeper into, such as NLP, CV, or MLOps, based on your interests.
Step 5: Advancing to the Frontier
Follow top conferences, read the latest papers, try to reproduce cutting-edge research, and ultimately form your own unique insights.
Part 1: Laying the Foundation - Mathematical Prerequisites
A deep, intuitive understanding of the underlying mathematical and computational principles is the cornerstone of success before calling high-level libraries.
Essence of Linear Algebra - 3Blue1Brown
The best starting point for building linear algebra intuition. Its 'visual-first' approach makes abstract concepts tangible, with unparalleled effectiveness.
Essence of Calculus - 3Blue1Brown
Following the linear algebra series, this one again uses animation and geometric intuition to reveal the core ideas of calculus, like derivatives, integrals, and the chain rule.
Statistics and Probability - Khan Academy
An extremely comprehensive and completely free resource covering all core topics from basic probability to significance testing.
An Introduction to Statistical Learning (ISL)
A foundational work in the field of statistical learning, bridging the gap between statistics and machine learning. Free PDF available.
Mathematics for Machine Learning
A comprehensive book by Deisenroth, Faisal, and Ong, covering linear algebra, calculus, and probability theory required for machine learning. Free online version available.
Part 1: Tools of Implementation - Programming & Algorithms
The tools needed to put mathematical concepts into practice. The ability to write efficient code is key to distinguishing advanced practitioners from beginners.
Python Core Course - Kaggle Learn
A series of concise, free, and hands-on micro-courses providing the most direct path to practical skills, covering Python and Pandas.
Algorithms Specialization - Stanford (Coursera)
An authoritative course in algorithms, covering divide and conquer, data structures, and graph algorithms, which are the cornerstones of computer science.
Google's Tech Dev Guide
An excellent free resource with tutorials, coding problems, and explanations on hash tables, trees, graphs, and Big O notation.
Part 2: Core Curriculum - Machine Learning Principles
Establishing the fundamental concepts of the machine learning field before the 'hype' of deep learning, with an emphasis on the breadth of algorithms and the statistical rigor that underpins them.
Machine Learning Specialization - Andrew Ng
The most famous and influential introductory course in machine learning. Provides a broad introduction to supervised learning, unsupervised learning, etc.
Artificial Intelligence: A Modern Approach (AIMA)
The 'standard' textbook in the field of artificial intelligence, offering a comprehensive introduction to a wide range of topics from search and knowledge representation to machine learning and robotics.
Neural Networks and Deep Learning - Michael Nielsen
A legendary free online book, famous for its clear, first-principles explanations of core concepts like backpropagation.
Part 2: Building Intelligence - Neural Networks & Deep Learning
The transition to modern deep learning, introducing the key figures, core courses, and 'bible-level' textbooks of the field.
Dive into Deep Learning - Mu Li
A unique open-source book and course that emphasizes learning through hands-on practice and runnable code. It covers deep learning models from basics to the cutting edge.
Practical Deep Learning for Coders - fast.ai
A unique top-down teaching philosophy: 'Build state-of-the-art models without a graduate-level math background.' Designed for those with programming experience.
Deep Learning - The 'Flower Book'
Co-authored by Goodfellow, Bengio, and Courville, widely considered the most comprehensive 'bible' textbook in the field.
MIT: Introduction to Deep Learning
An intensive, hands-on bootcamp-style course, updated annually to include cutting-edge topics like LLMs. Available for free on YouTube.
Neural Networks: Zero to Hero - Andrej Karpathy
Taught by Andrej Karpathy himself, this god-tier course implements backpropagation and Transformers from scratch using only Python. Essential for understanding underlying principles.
Neural Networks Course - Geoffrey Hinton
Taught by one of the 'Godfathers of AI', offering a deep, historical, and conceptual understanding of neural networks from a pioneer's perspective.
Pioneers & Thought Leaders
Learn about the giants who have shaped the field of artificial intelligence. Their work and ideas are a valuable treasure for every learner.
Geoffrey Hinton
One of the 'Godfathers of Deep Learning', made foundational contributions to the popularization of the backpropagation algorithm and neural network research, Turing Award winner.
Yann LeCun
Creator of Convolutional Neural Networks (CNNs), which revolutionized the field of computer vision. Chief AI Scientist at Meta, Turing Award winner.
Yoshua Bengio
Another 'Godfather of Deep Learning', did pioneering work on sequence models (especially language models), Turing Award winner.
Authors of 'Attention Is All You Need'
Published the groundbreaking paper that introduced the Transformer architecture, which has become the foundation of modern NLP and large language models.
Andrej Karpathy
An outstanding AI educator and researcher, former Senior Director of AI at Tesla, known for his contributions to computer vision, autonomous driving, and his clear teaching.
Fei-Fei Li
Created the ImageNet dataset, one of the key catalysts that ignited the deep learning revolution. Co-Director of Stanford's Human-Centered AI Institute.
Ian Goodfellow
Invented Generative Adversarial Networks (GANs), opening up a new era of generative models. One of the main authors of the 'Deep Learning' book.
Jeff Dean
Head of Google AI, led the development of large-scale computing systems like MapReduce, BigTable, and TensorFlow, providing the infrastructure for modern AI.
Mu Li
Author of 'Dive into Deep Learning', whose book/course has influenced countless learners with its hands-on, code-first approach. Former Principal Scientist at AWS.
Christopher D. Manning
Head of the Stanford NLP group, author of an NLP textbook, made significant contributions to tree-recursive neural networks and GloVe.
Jürgen Schmidhuber
Co-inventor of Long Short-Term Memory (LSTM), which was the dominant architecture for processing sequential data before the advent of Transformers.
Part 3: The Practitioner's Arena - Platforms, Data & Community
Mastery comes from practice. Here, you can apply your knowledge, find data, and connect with a global community.
Kaggle - Competitions & Datasets
More than just a competition platform, it's a complete learning environment with micro-courses, datasets, and cloud notebooks. The best place to build a portfolio.
Hugging Face - NLP Frontier
The de facto hub for NLP and LLMs. Offers a comprehensive free course on Transformers, fine-tuning, and a massive collection of pre-trained models.
Papers with Code
An excellent platform for discovering the latest machine learning papers and their open-source code implementations. Closely links academic research with engineering practice.
DrivenData
A platform focused on applying data science to solve social good problems. The competitions here are very meaningful if you want your skills to have a positive social impact.
Part 3: Key Tools & Experimentation Platforms
Good tools are prerequisites for success. These platforms and tools provide powerful support for your learning, experimentation, and project deployment.
Google Colaboratory
An essential tool for almost every deep learning practitioner. It provides a free online Jupyter Notebook environment with free access to GPU/TPU resources.
Alibaba Tianchi
A top-tier data science competition platform in China, hosted by Alibaba. Much like a 'Chinese Kaggle', it offers a large number of competitions and datasets from real business scenarios.
ModelScope Community
An open-source model community launched by Alibaba DAMO Academy, hosting a large number of models optimized for Chinese scenarios. A great place to find and use Chinese pre-trained models.
Weights & Biases
A very popular MLOps tool for tracking and visualizing machine learning experiments. It helps you log all key information, greatly facilitating experiment comparison and project collaboration.
TensorFlow Hub
Maintained by Google, a platform specifically for sharing and reusing TensorFlow models. If you are a TensorFlow user, this is an excellent place to find ready-to-use models.
Google AI Studio
An online tool for rapid prototyping and testing of Google's latest generative AI models (like Gemini). The best starting point for experiencing and building LLM-based applications.
Part 4: To the Frontier - Specialization & Research
After mastering the basics, advance to the forefront of the field by specializing in advanced techniques or participating in academic research.
Top Conference Papers - NeurIPS
The Conference on Neural Information Processing Systems is one of the most influential conferences in AI research. All papers are publicly accessible online.
Open Peer Review - OpenReview
Platforms like ICLR use this platform to make the entire peer review process public. An invaluable resource for learning what constitutes high-quality research.
Advanced Specializations on Coursera
Institutions like DeepLearning.AI offer specialized courses in high-demand areas such as NLP, GANs, and autonomous driving.
Recommended Learning Paths
Whether your goal is to become an applied expert, a cutting-edge researcher, or an engineering master, there is a roadmap tailored for you here.
Roadmap A: Applied AI Practitioner / Data Scientist
Foundations Stage
Foundations: Khan Academy (Stats), 3Blue1Brown (LinAlg), Kaggle Learn (Python, Pandas).
Core Machine Learning
Core ML: Andrew Ng's 'Machine Learning' specialization, supplemented by reading the 'ISL' book.
Hands-on Deep Learning
Hands-on DL: Systematically study 'Dive into Deep Learning' (d2l.ai) by Mu Li, completing all chapters and exercises.
Practical Application
Practical Application: Dive into Kaggle micro-courses, complete 2-3 Kaggle competitions, and build a personal portfolio.
Specialization
Specialization: Choose a direction based on interest: fast.ai (General Practice), Hugging Face (NLP), Karpathy's courses (Underlying Principles).
Roadmap B: AI Researcher / PhD Candidate
Foundations Stage
Foundations: Coursera/MIT OCW (Calculus, LinAlg, DSA), MML book (Full math overview).
Core Theory
Core Theory: Hinton's course + Nielsen's book -> Thoroughly read the 'Deep Learning Book' -> Thoroughly read AIMA.
Deep Dive into Principles
Deep Dive into Principles: Study Karpathy's 'Zero to Hero' series, hand-coding core modules for deep understanding.
Modern Practice
Modern Practice: Study MIT's 'Intro to Deep Learning' and Mu Li's 'd2l.ai', cross-validating theory with code.
Engage with the Frontier
Engage with the Frontier: Follow the latest progress on Papers with Code, read top conference papers, and try to reproduce a recent paper.
Roadmap C: AI/MLOps Engineer
Programming & Cloud Basics
Programming & Cloud Basics: Master Python, become familiar with the Linux command line, and complete an entry-level cloud practitioner course from Google AI or AWS.
Core Model Understanding
Core Model Understanding: Take Andrew Ng's 'Machine Learning' course to understand the basic principles of the models you will be deploying and maintaining.
Containerization & Orchestration
Containerization & Orchestration: Learn Docker and Kubernetes. Master building images and deploying applications through official documentation and free tutorials on YouTube.
MLOps Tools & CI/CD
MLOps Tools & CI/CD: Explore MLflow or W&B for experiment tracking, and use GitHub Actions to build automated deployment pipelines, wrapping models as APIs.
Production-Level Project
Production-Level Project: Deploy a Kaggle project or a model from d2l.ai end-to-end in a simulated production environment and implement basic monitoring.
Roadmap D: Data Analyst (ML Focus)
Data Analysis Trifecta
Data Analysis Trifecta: Master SQL (Kaggle), Pandas for data cleaning and transformation (Kaggle), and Matplotlib/Seaborn for data visualization.
Statistical Intuition
Statistical Intuition: Study statistics on Khan Academy and read the first few chapters of ISL, focusing on understanding data distributions, hypothesis testing, and correlation.
Applied Machine Learning
Applied Machine Learning: Complete Kaggle's 'Intro to Machine Learning' course, focusing on using Scikit-learn to solve classification and regression problems.
Business Insight Project
Business Insight Project: Choose a real business dataset (e.g., customer churn, marketing) to perform exploratory data analysis (EDA) and build a simple predictive model to drive business decisions.