Conference Information

Workshop Program

# Approximation
# Online Approach
# Multi-Agent RL
# Others
# Exploration
# Invited Talk
# POMDPs
# Hierarchical and
Factored RL
# Decomposition of RL

EWRL- 6
Workshop Program

Thursday, 4 September 2003

09:30 - Registration and Coffee

10:00 - Introductory remarks

10.10 - 11.10 Approximation Session chair: Frédérick Garcia

# 10:10 - A model-based actor-critic algorithm in continuous time and space.
Rémi Coulom, Laboratoire Leibnitz-IMAG, Grenoble, France - <remi-DOT-coulom-AT-free-DOT-fr>

# 10:30 - Error bounds for approximate policy iteration
Rémi Munos, CMAP, Ecole Polytechnique, Palaiseau, France - <remi-DOT-munos-AT-polytechnique-DOT-fr>

# 10:50 - A critic-critic architecture to combine reinforcement and supervised learnings
Fabien Montagne, LIL, Univ. du Littoral Côte d'Opale, Calais, France - <montagne-AT-lil-DOT-univ-littoral-DOT-fr>
Samuel Delepoulle, LIL, Univ. du Littoral Côte d'Opale, Calais, France - <delepoulle-AT-lil-DOT-univ-littoral-DOT-fr>
Philippe Preux, GRAPPA, Univ. Lille 3, Villeneuve d'Asc, France - <ppreux-AT-univ-lille3-DOT-fr>

11:10 - Coffee break

11.40 - 12.20 Online Approach Session chair: Olivier Sigaud

# 11:40 - Online Policy Iteration for several Environments and Reinforcement Functions
Andreas Matt, Institute of Mathematics, Univ. of Innsbruck, Austria - <andreas-DOT-matt-AT-uibk-DOT-ac-DOT-at>
Georg Regensburger, Institute of Mathematics, Univ. of Innsbruck, Austria - <georg-DOT-regensburgert-AT-uibk-DOT-ac-DOT-at>

# 12:00 - On-line search for solving large Markov Decision Processes
Laurent Peret, BIA, INRA, Auzeville, France - <peret-AT-toulouse-DOT-inra-DOT-fr>
Frédérick Garcia, BIA, INRA, Auzeville, France - <fgarcia-AT-toulouse-DOT-inra-DOT-fr>

12:20 - Lunch

14.00 - 15.00 Multi-Agent RL Session chair: Paul Crook

# 14:00 - Reinforcement Learning Techniques for Cooperative Multi-Agent Systems
Maartens Peeters, CML, Vrije Univ. Brussel, Brussel, Belgium - <mjpeeter-AT-vub-DOT-ac-DOT-be>
Katja Verbeeck, CML, Vrije Univ. Brussel, Brussel, Belgium - <kaverbee-AT-vub-DOT-ac-DOT-be>
Ann Nowé, CML, Vrije Univ. Brussel, Brussel, Belgium - <asnowe-AT-info-DOT-vub-DOT-ac-DOT-be>

# 14:20 - Learning to Control Forest Fires with ESP
Marco Wiering, ISG, Institute of Information and Computing Sciences, Utrecht Univ, the Netherlands - <marco-AT-cs-DOT-uu-DOT-nl>
Filippo Mignogna, ISG, Institute of Information and Computing Sciences, Utrecht Univ, the Netherlands - <fmignogn-AT-cs-DOT-uu-DOT-nl>

# 14:40 - A fuzzy adaptive Q-Learning algorithm for expertness based cooperative learning application to hunter-prey problem
H. Rezai-Sardareh, Dept. of Electrical Eng., Ferdowsi Univ. of Mashhad, Iran - <ho_rezaei-AT-yahoo-DOT-com>
M.-R. Akbarzadeh-T., Dept. of Electrical Eng., Ferdowsi Univ. of Mashhad, Iran

15.00 - Coffee break

15.30 - 16.10 Others Session chair: Massoumeh Izadi

# 15:30 - The reinforcement learning data graphical display by means of three way statistical data analysis
Danilo Leone, Univ. of Naples, Faculty of Economics, Mathematics and Statistics Dept., Italy - <danilo.leone-AT-dsaonline-DOT-it>
Marilena Fucili, Univ. of Naples, Faculty of Economics, Mathematics and Statistics Dept., Italy - <fucili-AT-unina-DOT-it>
Rosaria Lombardo, Second Univ. of Naples, Faculty of Economics, Accounting, Management and Quatitative Methods Dept., Italy - <rosaria.lombardo-AT-unina2-DOT-it>

# 15:50 - Learning Bayesian Behaviours for Synthetic Characters
Ronan Le Hy, GRAVIR/IMAG, INRIA Rhônes-Alpes, Montbonnot, France - <lehy-AT-imag-DOT-fr>
Pierre Bessière, GRAVIR/IMAG, INRIA Rhônes-Alpes, Montbonnot, France
Olivier Lebeltel, GRAVIR/IMAG, INRIA Rhônes-Alpes, Montbonnot, France
David Bellot, GRAVIR/IMAG, INRIA Rhônes-Alpes, Montbonnot, France
Anthony Arrigoni, GRAVIR/IMAG, INRIA Rhônes-Alpes, Montbonnot, France

16.10 - Coffe break

16.40 - 17.10 Exploration Session chair: Rémi Coulom

# 16:40 - Optimistic Model Selection in Structure Based Reinforcement Learning
Funlade Sunmola, School of Computer Science, Univ. of Birmingham, Birmingham, UK - <f.t.sunmola-AT-cs-DOT-bham-DOT-ac-DOT-uk>
Jeremy Wyatt, School of Computer Science, Univ. of Birmingham, Birmingham, UK - <j.l.wyatt-AT-cs-DOT-bham-DOT-ac-DOT-uk>

# 17:00 - Exploration in RL Using MDP Characteristics
Bohdana Ratitch, School of Computer Science, McGill Univ., Montreal, Canada - <bohdana-AT-cs-DOT-mcgill-DOT-ca>
Doina Precup, School of Computer Science, McGill Univ., Montreal, Canada - <dprecup-AT-cs-DOT-mcgill-DOT-ca>

20:00 - Social dinner (organized by EWRL-6)

Friday, 5 September 2003

9:40 - Invited talk
Evolutionist Game Theory and Reinforcement Learning
Bernard Walliser, ENPC-CERAS, Paris, France - <walliser-AT-mail-DOT-enpc-DOT-fr>

10:30 - Coffee break

11.00 - 12.00 POMDPs Session chair: Marco Wiering

# 11:00 - Active Perception in Navigation of Partially Observable Grid World
Paul Crook, IPAB, Univ. of Edinburgh, Edinburgh, Scotland - <paulc-AT-dai-DOT-ed-DOT-ac-DOT-uk>
Gillian Hayes, IPAB, Univ. of Edinburgh, Edinburgh, Scotland - <gmh-AT-inf-DOT-ed-DOT-ac-DOT-uk>

# 11:20 - Planning Algorithm for Predictive State Representations
Massoumeh T. Izadi, School of Computer Science, McGill Univ., Montreal, Canada - <mtabae-AT-cs-DOT-mcgill-DOT-ca>
Doina Precup, School of Computer Science, McGill Univ., Montreal, Canada - <dprecup-AT-cs-DOT-mcgill-DOT-ca>

# 11:40 - Generalized Dynamic Programming for Decentralized POMDPs
Daniel S. Bernstein, Dept. of Computer Science, Univ. of Massachusetts, Amherst, USA - <bern-AT-cs-DOT-umass-DOT-edu>
Shlomo Zilberstein, Dept. of Computer Science, Univ. of Massachusetts, Amherst, USA - <shlomo-AT-cs-DOT-umass-DOT-edu>

12:00 - Lunch

14.00 - 16.30 Hierarchical and Factored RL Session chair: Ann Nowé

# 14:00 - Hierarchical Reinforcement Learning Based on Automatic Discovery of Subgoals and Specialization of Subpolicies
Bram Baker, IAS, Univ. of Amsterdam, Amsterdam, The Netherlands - <bram-AT-science-DOT-uva-DOT-nl>
Jürgen Schmidhuber, IDSIA, Lugano, Switzerland - <juergen-AT-idsia-DOT-ch>

# 14:20 - An Overview of Current Issues in Exploration in Reinforcement Learning
Jeremy Wyatt, School of Computer Science, Univ. of Birmingham, Birmingham, UK - <J.L.Wyatt-AT-cs-DOT-bham-DOT-ac-DOT-uk>

# 14:40 - Efficient Reinforcement Learning using Relational Aggregation
Martijn Van Otterlo, TKI, Dept. of Computer Science, Enshede, the Netherlands - <otterlo-AT-cs-DOT-utwente-DOT-nl>

15:00 - Coffee break

15.30 - 17.40 Decomposition of RL Session chair: Bohdana Ratitch

# 15:30 - Learning to control real systems with parallel Dyna-Q
Guillaume Laurent, Laboratoire d'Automatique de Besançon; Besançon, France - <glaurent-AT-ens2m-DOT-fr>

# 15:50 - A Self-Made Agent Based on Action-Selection
Olivier Buffet, LORIA, Nancy, France - <buffet_AT_loria-DOT-fr>
Alain Dutech, LORIA, Nancy, France - <dutech-AT-loria-DOT-fr>

16:10 - Final Discussion (trying to end at 17.30)

For any information concerning EWRL'6, send an email to Alain Dutech or Iadine Chadès

For any other information concerning the European Workshops on RL, you can also send an email to Marco Dorigo: mdorigo-AT-ulb-AT-ac-AT-be

Last update, July 24th 2003, by Iadine Chadès.