New Book on Reinforcement Learning
Title: ''Reinforcement Learning: State of the Art''
Editors: Marco Wiering and Martijn van Otterlo
Publisher: Springer Verlag
Welcome! If you read this page, you have probably been invited by the editors to contribute a book chapter to the forthcoming new book on reinforcement learning (RL). The editors have carefully selected you as an expert within a subfield of reinforcement learning and hope you accept the invitation to contribute to what should become a new standard book on the field of RL.
Motivation
We are Marco Wiering and Martijn van Otterlo, and we have recently initiated a book project to fill the void that Sutton&Barto, and Bertsekas&Tsitsiklis left after 1998. Because, since a couple of years, it has become increasingly harder to find this one source of literature for your students and colleagues interested in reinforcement learning. The field has been expanding rapidly the last decade, but it is not possible to find one single book in which all contemporary aspects of modern reinforcement learning are included.
Important dates:
30-09-2010 Deadline for submission of the draft chapter
27-01-2011 Feedback about the chapter is given
27-02-2011 Deadline for the camera ready submission
15-04-2011 Publication of the book
Information about writing a chapter
Each chapter consists of between 15 and 25 pages and will have the following characteristics:
About 3/4 of the paper introduces the subfield and its main methods, and surveys the important literature to understand this part of reinforcement learning. (This part of the chapter will provide the necessary contents of the book)
About 1/4 is left open to the expert to exemplify certain aspects or methods using his or her own work. (This part of the chapter will provide the author with means to promote his or her own work, and to stress certain aspects that he/she deems interesting)
We will provide basic formalizations of Markov decision processes and some model-free and model-based algorithms to find value functions and policies, in a first chapter. This means that the expert can safely assume that this is in place, and start from there with formalizing and describing matters.
List of all chapters
(no order imposed, except for the first and last)
==========================
Introduction M. van Otterlo and M. Wiering
Continuous states and actions H. van Hasselt
Relational and first-order knowledge representation M. van Otterlo
Hierarchical approaches B. Hengst
Predictive approaches D. Wingate
Game theory and Multi-agent RL A. Nowe
POMDPs M. Spaan
Decentralized POMDPs F. Oliehoek
RL batch learning M. Riedmiller
Bounds and complexity L. Li
RL for games I. Szita
RL in robotics J. Peters
Bayesian RL N. Vlassis
Least squares policy iteration L. Busoniu
The use of models T. Hester
Transfer learning A. Lazaric
Biological and Psychological Background A. Shah
Evolutionary approaches S. Whiteson
Closing chapter, summarizing, horizons M. Wiering et al.
Links:
Style files Springer, please look at authinst.pdf in the zip-file!
Intro-draft.pdf, note that the notation is important in this chapter. Please have a look at it to see how we would like to use this same notation in the whole book.
http://www.ai.rug.nl/~mwiering/intro.tex shows the intro-chapter tex file.
Last edited on 10-12-2010 by Marco Wiering and Martijn van Otterlo