Distributional Reinforcement Learning

The MIT Press

Marc G. Bellemare and Will Dabney and Mark Rowland

This textbook aims to provide an introduction to the developing field of distributional reinforcement learning. The book is available at The MIT Press website (including an open access version).

Distributional Reinforcement Learning

• Table of Contents
• Preface
• 1 Introduction
• 2 The Distribution of Returns
• 3 Learning the Return Distribution
• 4 Operators and Metrics
• 5 Distributional Dynamic Programming
• 6 Incremental Algorithms
• 7 Control
• 8 Statistical Functionals
• 9 Linear Function Approximation
• 10 Deep Reinforcement Learning
• 11 Two Applications and a Conclusion
• Notation
• Bibliography

FAQ

• How do I provide feedback?

  We welcome feedback and questions on all parts of the book (and in particular typographical issues and technical points). The preferred mode of communication is to email us at distributionalrl@gmail.com.

• Why this format for the web version of the book?

  Our agreement with the publisher allows us to make the draft available, but not as a PDF. This format gives access to the work to researchers who cannot readily purchase the published book.

Citing the book

@book{bdr2023,
    title={Distributional Reinforcement Learning},
    author={Marc G. Bellemare and Will Dabney and Mark Rowland},
    publisher={MIT Press},
    note={\url{http://www.distributional-rl.org}},
    year={2023}
}