Mushroom power-ups for Chinese reinforcement learners

What it does The repo is the online home of EasyRL (蘑菇书), a Chinese-language reinforcement learning course. It packages video lectures—primarily from NTU’s Li Hongyi—into written chapters covering everything from Q-learning and policy gradients to DDPG and AlphaStar, pairing theory with chapter exercises and Jupyter notebook implementations. The content is also available as a printed book from People’s Post and Telecommunications Press, while the web version continues to receive updates such as recent ICLR’25 Minecraft work.

The interesting bit Rather than building yet another RL framework, the project acts as community-powered lecture notes: it absorbs scattered video courses and paper interpretations, then reassembles them into a single linear curriculum with runnable code. The “mushroom” branding is deliberate—the authors want readers to power up like Mario after eating a mushroom.

Key highlights

• Covers a wide arc from tabular methods (Q-learning, SARSA) through deep RL (DQN variants, PPO, A2C, DDPG) to frontiers like AlphaStar and visual RL paper lists.
• Each chapter includes exercises; many chapters link to dedicated Jupyter notebooks implementing the algorithms.
• Maintains both a free, continuously updated web edition and a professionally edited print edition.
• Includes a papers folder with classic RL paper interpretations.
• Licensed under CC BY-NC-SA 4.0, so it is explicitly non-commercial.

Caveats

• Entirely in Chinese; English speakers need not apply unless they are reading for the code only.
• The free PDF is described as an unedited draft, while the print book contains the refined version.
• Code coverage is uneven: some chapters have no linked notebooks.

Verdict Worth bookmarking if you are a Chinese-speaking developer or student looking for a structured, free on-ramp to deep RL with homework and code. Skip it if you want a drop-in Python library or an English reference.