EWRL-5 Home Page

WORKSHOP PRELIMINARY PROGRAM

Utrecht, The Netherlands, October 5-6, 2001


Friday 5 October 2001

  • 09:30 - Registration and Coffee

  • 10:00 - Introductory remarks

  • 10.10 - 10.30 Policy Search Session chair: Marco Wiering

  • 10:10 - Policy Search using a State-Policy Evaluation Function
    Malcolm Strens, QinetiQ Center for Robotics & Machine Vision, mjstrens@QinetiQ.com

  • 10:30 - Gradient-based Reinforcement Planning in Policy-Search Methods
    Ivo Kwee, Marcus Hutter and Juergen Schmidhuber, Idsia, Lugano, ivo@idsia.ch

  • 10:50 - Policy Improvement for Several Environments
    Andreas Matt and Georg Regensburger, University of Innsbruck,andreas.matt@uibk.ac.at and georg.regensburger@uibk.ac.at

  • 11:10 - Coffee break

  • 11.40 - 12.40 POMDPs and Combinatorial Optimization Session chair: Marco Dorigo

  • 11:40 - Learning to use Contextual Information for Solving Partially Observed Markov Decision Problems
    Alain Dutech and Bruno Scherrer, Loria/Inria, France, dutech@loria.fr

  • 12:00 - Reinforcement learning in non-Markovian domains using LSTM recurrent neural networks
    Bram Bakker,Dept. of Psychology, Leiden University, bbakker@fsw.leidenuniv.nl

  • 12:20 - Reinforcement Learning of Combinatorial Optimization Problem with Ant algorithm
    Nicolas Meuleau, MIT, nm@ai.mit.edu

  • 12:40 - Lunch

  • 14.00 - 14.40 Function Approximation Session chair: Jeremy Wyatt

  • 14:00 - Hippocampal Spatial Model for State Space Representation in Robotic Reinforcement Learning
    Angelo Arleo and Wulfram Gerstner, EPFL Lausanne Switzerland, angelo.arleo@ep fl.ch

  • 14:20 - Reinforcement Learning and the Perception of Time Intervals
    Jon Shapiro, University of Manchester, jls@cs.man.ac.uk

  • 14:40 - Coffee break

  • 15.10 - 16.10 Multi-Agent RL Session chair: Juergen Schmidhuber

  • 15:10 - A Markov Model for Dyadic Interaction Learning
    Walter Gutjahr and Anselm Eder, University of Vienna, Austria, walter.gutjahr @univie.ac.at

  • 15:30 - Learning Fair Periodical Policies
    Katja Verbeeck, Ann Now\'e, and Johan Parent. Vrije Universiteit Brussel, Belgium, kaverbee@vub.ac.be

  • 15:50 - A non supervised multi-reinforcement agents architecture to model the development of behavior of living organisms
    Philippe Preux, C. Cassagnabere, S. Delepoulle and J-C Darcheville, Laboratoire d'Informatique du Littoral, France, philippe.preux@lil.univ-littoral.fr

  • 16:10 - Coffee break

  • 16.40 - 17.20 Exploration Session chair: Malcolm Strens

  • 16:40 - Advances in exploration control in reinforcement learning
    Jeremy Wyatt and Funlade Summola, University of Birmingham, jlw@cs.bham.ac.uk

  • 17:00 - The Curse of Optimism
    Stuart Reynolds, The University of Birmingham, sir@cs.bham.ac.uk

  • 20:00 - Social dinner (organized by EWRL-5)

  • Saturday 6 October 2001

  • 9:40 - Robotic Reinforcement learning and the use of teacher signals (Invited talk)
    Leslie Kaelbling, MIT, USA

  • 10:30 - Coffee break

  • 11.00 - 12.20 Hierarchy Session chair: Stuart Reynolds

  • 11:00 - Looking for Scalable Agents
    Olivier Buffet and Alain Dutech, LORIA/INRIA, France, buffet@loria.fr

  • 11:20 - Using Multi-step Actions for Faster Reinforcement Learning
    Ralf Schoknecht and Martin Riedmiller, University of Karlsruhe, Germany, schokn@ira.uka.de

  • 11:40 - Learning Digger using Hierarchical Reinforcement Learning for Concurrent Goals
    Kurt Driessens and Hendrik Blockeel, Katholieke Universiteit Leuven, Belgium, kurt.driessens@cs.kuleuven.ac.be

  • 12:00 - ATD and AQ-learning: reward baseline reinforcement learning algorithms for discounted reward problems
    Frederick Garcia, INRA/BIA, France, fgarcia@toulouse.inra.fr

  • 12:20 - Lunch

  • 13.50 - 14.50 Different Session chair: Leslie Kaelbling

  • 13:50 - Experience Stack Reinforcement Learning: An Online Forward lambda-Return Method
    Stuart Reynolds, The University of Birmingham, sir@cs.bham.ac.uk

  • 14:10 - Kolmogorov Complexity and universal learners
    Juergen Schmidhuber, Idsia, Switzerland, juergen@idsia.ch

  • 14:30 - Universal Sequential Decisions in Unknown Environment
    Marcus Hutter, Idsia, Switzerland, marcus@idsia.ch

  • 14:50 - Coffee break

  • 15.20 - 16.20 Applications Session chair: Ralf Schoknecht

  • 15:20 - Scheduling with adaptive agents - an empirical evaluation
    Werner Hunger and Martin Riedmiller, University of Karlsruhe, Germany, riedml@ira.uka.de

  • 15:40 - Order Acceptance with Reinforcement Learning
    Marisela Mainegra Hing and Aart van Harten, University of Twente, The Netherlands, M.MainegraHing@sms.utwente.nl

  • 16:00 - Applying Reinforcement Learning To ITS For Adapting Learning Situations
    Abdellah Bennane, I. Michiels, B. Manderick, and T. D'Hondt, Vrije Universiteit Brussel, Belgium, abennane@vub.ac.be

  • 16:20 - Final Discussion