Connectionists: Second Annual Reinforcement Learning Competition

[Shimon Whiteson]

[Apologies for multiple postings]

===============================================
   The Second Annual Reinforcement Learning Competition
                     at ICML-08 in Helsinki, Finland
              Announcement and Call for Participants
===============================================

The Second Annual Reinforcement Learning Competition invites  
researchers from around the world to apply their latest methods to a  
suite of exciting and diverse challenge problems.  The aim of the  
competition is to facilitate direct comparisons between various  
learning methods on important and realistic domains.  We believe such  
a competition can stimulate the development and verification of  
increasingly practical algorithms.

This year's event will feature well-known benchmark domains as well  
as more challenging problems of real-world complexity.  The  
competition domains are:

-Mountain Car: Perhaps the most well-known reinforcement learning  
benchmark task, in which an agent must learn how to drive an  
underpowered car up a steep mountain road.

-Tetris: The hugely popular video game, in which four-block shapes  
must be manipulated to form complete lines when they fall.

-Helicopter Hovering: A simulator, based on the work of Andrew Ng and  
collaborators, which requires an agent to learn to control a hovering  
helicopter.

-Keepaway: A challenging task, based on the RoboCup soccer simulator,  
that requires a team of three robots to maintain possession the ball  
while two other robots attempt to steal it.

-Real-Time Strategy: A game, based on popular real-time strategy  
games, which poses exciting new challenges for the reinforcement  
learning community.

-Polyathlon: The agent will face a set of potentially unrelated MDPs  
with minimal task knowledge and no prior training.

In addition, this year's competition will utilize new evaluation  
paradigms designed to encourage algorithms that generalize well to  
previously unseen tasks.  In particular, each domain will be  
paramaterized and competition parameters will differ from those seen  
by the participants while developing their entries. As a result, only  
learning algorithms that are robust across a range of parameters can  
expect to perform well in the competition.

The competition will end with an event at the 2008 International  
Conference on Machine Learning in Helsinki, Finland, at which the  
winners will be announced.  Competitors will be invited to attend and  
present their methods.  The event will also feature invited speakers  
and discussions about the best way to perform empirical comparisons  
in reinforcement learning and the future of the competition.

Website
======

To learn more about the competition, details of the domains, and how  
to get started, visit the competition website at:

http://rl-competition.org

Timeline
=======

1 November, 2007: Public release of competition training software

1 December, 2007: Public release of competition test software;  
competitors can begin submitting results

1 July 2008: Deadline to submit competition results

6-9 July 2008: Competition event at ICML in Helsinki, Finland

Organizing Committee
==================
Shimon Whiteson, Universiteit van Amsterdam (Chair)
Adam White, University of Alberta
Rich Sutton, University of Alberta
Doina Precup, McGill University
Peter Stone, University of Texas at Austin
Michael Littman, Rutgers University
Nikos Vlassis, Technical University of Crete
Martin Riedmiller, Universität Osnabrück

Technical Committee
=================
Brian Tanner, University of Alberta (Chair)
Marc Lanctot, University of Alberta
Pieter Abbeel, Stanford University
Matt Taylor, University of Texas at Austin
Shivaram Kalyanakrishnan, University of Texas at Austin
Leah Hackman, University of Alberta
Mark Lee, University of Alberta
Jordan Frank, McGill University
Nick Jong, University of Texas at Austin