ShangtongZhang/reinforcement-learning-an-introduction

Implements canonical RL algorithms across tabular and function approximation settings—including multi-armed bandits, dynamic programming (policy/value iteration), temporal difference learning (Sarsa, Q-learning), Monte Carlo methods, and n-step bootstrapping—with reproducible experiments that match textbook figures. Uses NumPy for computation and matplotlib for visualization, structured chapter-by-chapter to align with Sutton & Barto's 2nd edition, enabling side-by-side comparison between theoretical concepts and working code implementations.

14,587 stars. No commits in the last 6 months.