Reinforcement Learning: An Introduction

Richard S. Sutton, Andrew G. Barto  ·  2018

Reference. Sutton, R. S. & Barto, A. G. (2018). Reinforcement Learning: An Introduction (2nd ed.). MIT Press. ISBN 978-0-262-03924-6.  [link]

  • Claim: An agent can learn good behaviour from reward alone, by estimating the value of states and actions from its own experience.
  • Method: Markov decision processes; dynamic programming, Monte Carlo and temporal-difference learning; function approximation; policy-gradient methods.
  • Matters: The common vocabulary behind every RL post here, and the textbook under the Udacity Deep RL coursework.
  • Connects to: Reinforcement learning · The agent loop: a field guide
  • Code: —
Reinforcement Learning: An Introduction cover
Cover of Reinforcement Learning: An Introduction on the Open Library.