Reinforcement Learning: An Introduction

Přední strana obálky
MIT Press, 26. 2. 1998 - Počet stran: 344
Richard Sutton and Andrew Barto provide a clear and simple account of the key ideas and algorithms of reinforcement learning. Their discussion ranges from the history of the field's intellectual foundations to the most recent developments and applications.

Reinforcement learning, one of the most active research areas in artificial intelligence, is a computational approach to learning whereby an agent tries to maximize the total amount of reward it receives when interacting with a complex, uncertain environment. In Reinforcement Learning, Richard Sutton and Andrew Barto provide a clear and simple account of the key ideas and algorithms of reinforcement learning. Their discussion ranges from the history of the field's intellectual foundations to the most recent developments and applications. The only necessary mathematical background is familiarity with elementary concepts of probability.

The book is divided into three parts. Part I defines the reinforcement learning problem in terms of Markov decision processes. Part II provides basic solution methods: dynamic programming, Monte Carlo methods, and temporal-difference learning. Part III presents a unified view of the solution methods and incorporates artificial neural networks, eligibility traces, and planning; the two final chapters present case studies and consider the future of reinforcement learning.

 

Obsah

The Problem 1
11
Evaluative Feedback
25
The Reinforcement Learning Problem
51
Elementary Solution Methods
87
Monte Carlo Methods
111
TemporalDifference Learning
133
A Unified View
161
Generalization and Function Approximation
193
Planning and Learning
227
Dimensions of Reinforcement Learning
255
Case Studies
261
References
291
Summary of Notation
313
Autorská práva

Další vydání - Zobrazit všechny

Reinforcement Learning
Cornelius Weber,Mark Elshaw,N. Michael Mayer
Omezený náhled - 2008
Reinforcement Learning: An Introduction
Richard S. Sutton,Andrew G. Barto
Omezený náhled - 1998
Reinforcement Learning
Richard S. Sutton
Omezený náhled - 1992
Zobrazit všechny »

Běžně se vyskytující výrazy a sousloví

a₁ action selection action-value function agent algorithm applied artificial intelligence average backgammon backup diagram bandit problem Barto behavior Bellman equation Bellman optimality equation called chapter computation convergence DP methods dynamic programming e-greedy eligibility traces environment estimated value example exploration Figure full backups function approximation goal gradient-descent greedy policy heuristic search idea incremental interaction learning system linear Machine Learning Markov Markov decision process Markov property Monte Carlo methods move n-armed bandit n-step neural networks off-policy on-line on-policy distribution one-step optimal control optimal policy performance planning play player policy evaluation policy improvement policy iteration possible prediction probability Q-learning random walk reinforcement learning methods reinforcement learning problem replacing traces reward robot sample backups Sarsa schedule sequence simulated softmax St+1 state-action pair state-value function step step-size parameter stochastic supervised learning Sutton TD error TD methods TD-Gammon tilings Tsitsiklis update value iteration zero

O autorovi (1998)

Richard S. Sutton is Professor of Computing Science and AITF Chair in Reinforcement Learning and Artificial Intelligence at the University of Alberta, and also Distinguished Research Scientist at DeepMind.

Andrew G. Barto is Professor Emeritus in the College of Computer and Information Sciences at the University of Massachusetts Amherst.

Bibliografické údaje

NázevReinforcement Learning: An Introduction
Adaptive Computation and Machine Learning series
AutořiRichard S. Sutton, Andrew G. Barto
VydavatelMIT Press, 1998
ISBN0262303841, 9780262303842
DélkaPočet stran: 344
  
Exportovat citaciBiBTeX EndNote RefMan