From “what is a reward?” to deep, multi-agent, and real-world RL — intuition first (the sister on her bike), then the precise maths, then the algorithms. Built from the author’s own articles and notes.
Start here. What RL is (learn by trial and error from rewards), the agent–environment loop, the return and discounting, and the MDP that frames it all.
The heart of RL. Interaction & feedback, the policy, value functions, the Bellman equation, TD learning, Q-learning, SARSA, exploration vs exploitation, and Monte Carlo vs TD.
Scaling up with neural networks. Function approximation, Deep Q-Networks, policy gradients, actor-critic, PPO, and continuous control (DDPG, TD3, SAC).
The frontier. Model-based RL, reward shaping & reward hacking, multi-agent RL, offline & imitation learning, RLHF, and where RL wins in the real world.
Complete every lesson and claim your personalized certificate of completion.
These are the real, canonical works behind this course — go deeper here.