Chapters

The big picture: what RL is, where it came from, and where you see it today

The building blocks: agents, environments, states, actions, rewards, and policies

The algorithm landscape, your roadmap, and your first hands-on demo

The mathematical framework that formalizes sequential decision-making

Measuring how good states and actions are

The recursive equations that make RL possible

Computing the value of a policy when you know the model

Finding better policies through value functions

Master the exploration-exploitation tradeoff in the simplest RL setting

Personalized decisions based on context features

Learning from experience without waiting for the episode to end

On-policy TD control: learning while following your current policy

Off-policy TD control: learning the optimal policy while exploring

Scaling RL to large state spaces with learned representations

The breakthrough that made deep RL work

Enhancements that make DQN even better

A fundamentally different approach: learning policies directly

The foundational policy gradient algorithm

Combining policy and value learning for stability

The most popular deep RL algorithm in practice

Learning world models for sample-efficient planning

When multiple agents learn and interact together

Learning from logged data without environment interaction

From RLHF to reasoning: how RL transforms language models

Reducing model size and speeding up inference by using lower-precision numbers