The Hundred-Page Deep Reinforcement Learning Book

Understand how intelligent agents learn from experience—and build your own from scratch!

The Hundred-Page Deep Reinforcement Learning Book by Andriy Burkov, author of The Hundred-Page Machine Learning Book and The Hundred-Page Language Models Book, offers a concise yet rigorous introduction to the mathematical foundations and practical implementation of deep reinforcement learning.

Within Burkov’s signature hundred-page format, readers follow the ideas behind modern reinforcement learning algorithms, understand why each improvement is needed, and turn the mathematics into working Python code.

The Hundred-Page Deep Reinforcement Learning Book allows you to:

• Understand the Policy Gradient Theorem and derive the algorithms built upon it
• Implement REINFORCE, Proximal Policy Optimization (PPO), and Deep Q-Networks (DQN) from scratch in PyTorch
• Train agents to land a reusable rocket booster on a moving ocean platform
• Understand how reinforcement learning applies to language models, including learning from human feedback and verifiable rewards

Clear explanations, mathematical derivations, illustrations, and complete implementations connect theory with practice. By using the same rocket-landing task across algorithms, the book lets you compare how different methods learn, behave, and perform.

About the technology

Deep reinforcement learning combines neural networks with learning through interaction. An agent takes actions, observes their consequences, and improves its decisions to maximize reward over time.

These methods help robots learn to move, control systems adapt to changing conditions, and language models improve their responses. Understanding the underlying algorithms gives engineers the foundation to build, evaluate, and troubleshoot systems that learn from experience.

About the book

The Hundred-Page Deep Reinforcement Learning Book builds the subject step by step, beginning with reinforcement learning fundamentals and the Policy Gradient Theorem before developing REINFORCE, actor-critic methods, Generalized Advantage Estimation, and PPO. It then introduces off-policy learning and derives DQN from first principles.

Each new method addresses a limitation of the ideas that came before it. Readers learn why baselines reduce variance, how bootstrapping changes learning, and what makes practical policy optimization possible.

The book also connects these foundations to reinforcement learning for language models, explaining how generated responses become trajectories and how rewards guide training.

What’s inside

• States, actions, rewards, policies, and the reinforcement learning objective
• Policy gradients and the REINFORCE algorithm
• Value functions, actor-critic methods, and Generalized Advantage Estimation
• Proximal Policy Optimization
• Deep Q-learning, experience replay, and target networks
• Complete PyTorch implementations with training results
• Reinforcement learning from human feedback and verifiable rewards
• Mathematical appendices and pointers for further study

Chapters, code, and additional learning resources are available at thedrlbook.com, with companion material at thedrlbook.com/wiki.

About the reader

For software developers, data scientists, machine learning engineers, and technical leaders who want to understand how reinforcement learning works under the hood.

Readers will benefit from Python programming experience and familiarity with basic calculus, probability, linear algebra, and neural networks. Mathematical appendices and companion resources help refresh the foundations along the way.

Endorsements from tech and AI leaders

Stephen Balaban, Co-founder and CTO at Lambda: "Deep reinforcement learning is one of the most important techniques for training frontier models. In this book, Andriy will walk you through the fundamentals of RL and build up to modern applications such as RLHF."

Long Ouyang, Researcher on InstructGPT and ChatGPT at OpenAI: “A cogent and helpful introduction to RL and also a great reference. Read this book!”

More endorsements from AI leaders at thedrlbook.com.