Algorithms for reinforcement learning (2023)

Szepesvári, C. (2010). Algorithms for reinforcement learning. In R. J. Brachman, & T. Dietterich (Eds.), Algorithms for Reinforcement Learning (pp. 1-89). (Synthesis Lectures on Artificial Intelligence and Machine Learning; Vol. 9). https://doi.org/10.2200/S00268ED1V01Y201005AIM009

@inproceedings{f8629d3882e842c1b06ce521d78dd410,

title = "Algorithms for reinforcement learning",

abstract = "Reinforcement learning is a learning paradigm concerned with learning to control a system so as to maximize a numerical performance measure that expresses a long-term objective. What distinguishes reinforcement learning from supervised learning is that only partial feedback is given to the learner about the learner's predictions. Further, the predictions may have long term effects through influencing the future state of the controlled system. Thus, time plays a special role. The goal in reinforcement learning is to develop efficient learning algorithms, as well as to understand the algorithms' merits and limitations. Reinforcement learning is of great interest because of the large number of practical applications that it can be used to address, ranging from problems in artificial intelligence to operations research or control engineering. In this book, we focus on those algorithms of reinforcement learning that build on the powerful theory of dynamic programming. We give a fairly comprehensive catalog of learning problems, describe the core ideas, note a large number of state of the art algorithms, followed by the discussion of their theoretical properties and limitations. Table of Contents: Markov Decision Processes / Value Prediction Problems / Control / For Further Exploration.",

keywords = "Markov Decision Processes, Monte-Carlo methods, PAC-learning, Q-learning, active learning, actor-critic methods, bias-variance tradeoff, function approximation, least-squares methods, natural gradient, online learning, overfitting, planning, policy gradient, reinforcement learning, simulation, simulation optimization, stochastic approximation, stochastic gradient methods, temporal difference learning, two-timescale stochastic approximation",

author = "Csaba Szepesv{\'a}ri",

year = "2010",

doi = "10.2200/S00268ED1V01Y201005AIM009",

language = "אנגלית",

isbn = "9781608454921",

series = "Synthesis Lectures on Artificial Intelligence and Machine Learning",

pages = "1--89",

editor = "Brachman, {Ronald J.} and Thomas Dietterich",

booktitle = "Algorithms for Reinforcement Learning",

}

Szepesvári, C 2010, Algorithms for reinforcement learning. in RJ Brachman & T Dietterich (eds), Algorithms for Reinforcement Learning. Synthesis Lectures on Artificial Intelligence and Machine Learning, vol. 9, pp. 1-89. https://doi.org/10.2200/S00268ED1V01Y201005AIM009

Algorithms for reinforcement learning. / Szepesvári, Csaba.

Algorithms for Reinforcement Learning. ed. / Ronald J. Brachman; Thomas Dietterich. 2010. p. 1-89 (Synthesis Lectures on Artificial Intelligence and Machine Learning; Vol. 9).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Algorithms for reinforcement learning

AU - Szepesvári, Csaba

PY - 2010

Y1 - 2010

N2 - Reinforcement learning is a learning paradigm concerned with learning to control a system so as to maximize a numerical performance measure that expresses a long-term objective. What distinguishes reinforcement learning from supervised learning is that only partial feedback is given to the learner about the learner's predictions. Further, the predictions may have long term effects through influencing the future state of the controlled system. Thus, time plays a special role. The goal in reinforcement learning is to develop efficient learning algorithms, as well as to understand the algorithms' merits and limitations. Reinforcement learning is of great interest because of the large number of practical applications that it can be used to address, ranging from problems in artificial intelligence to operations research or control engineering. In this book, we focus on those algorithms of reinforcement learning that build on the powerful theory of dynamic programming. We give a fairly comprehensive catalog of learning problems, describe the core ideas, note a large number of state of the art algorithms, followed by the discussion of their theoretical properties and limitations. Table of Contents: Markov Decision Processes / Value Prediction Problems / Control / For Further Exploration.

AB - Reinforcement learning is a learning paradigm concerned with learning to control a system so as to maximize a numerical performance measure that expresses a long-term objective. What distinguishes reinforcement learning from supervised learning is that only partial feedback is given to the learner about the learner's predictions. Further, the predictions may have long term effects through influencing the future state of the controlled system. Thus, time plays a special role. The goal in reinforcement learning is to develop efficient learning algorithms, as well as to understand the algorithms' merits and limitations. Reinforcement learning is of great interest because of the large number of practical applications that it can be used to address, ranging from problems in artificial intelligence to operations research or control engineering. In this book, we focus on those algorithms of reinforcement learning that build on the powerful theory of dynamic programming. We give a fairly comprehensive catalog of learning problems, describe the core ideas, note a large number of state of the art algorithms, followed by the discussion of their theoretical properties and limitations. Table of Contents: Markov Decision Processes / Value Prediction Problems / Control / For Further Exploration.

KW - Markov Decision Processes

KW - Monte-Carlo methods

KW - PAC-learning

KW - Q-learning

KW - active learning

KW - actor-critic methods

KW - bias-variance tradeoff

KW - function approximation

KW - least-squares methods

KW - natural gradient

KW - online learning

KW - overfitting

KW - planning

KW - policy gradient

KW - reinforcement learning

KW - simulation

KW - simulation optimization

KW - stochastic approximation

KW - stochastic gradient methods

KW - temporal difference learning

KW - two-timescale stochastic approximation

UR - http://www.scopus.com/inward/record.url?scp=77955790905&partnerID=8YFLogxK

U2 - 10.2200/S00268ED1V01Y201005AIM009

DO - 10.2200/S00268ED1V01Y201005AIM009

M3 - פרסום בספר כנס

AN - SCOPUS:77955790905

SN - 9781608454921

T3 - Synthesis Lectures on Artificial Intelligence and Machine Learning

SP - 1

EP - 89

BT - Algorithms for Reinforcement Learning

A2 - Brachman, Ronald J.

A2 - Dietterich, Thomas

ER -

Szepesvári C. Algorithms for reinforcement learning. In Brachman RJ, Dietterich T, editors, Algorithms for Reinforcement Learning. 2010. p. 1-89. (Synthesis Lectures on Artificial Intelligence and Machine Learning). https://doi.org/10.2200/S00268ED1V01Y201005AIM009