EWRL3 (1997) | European Workshops on Reinforcement Learning

EWRL-3 location —————

Departement Informatique
INSA: INSTITUT NATIONAL DES SCIENCES APPLIQUEES DE RENNES
20 av. des Buttes de Coësmes,
35043 RENNES Cédex,
France

How to get to INSA

From train station (northern exit), walk along J. Janvier avenue up to Vilaine river. Turn right (Quai Richemont): the bus station (lines #16, #16express or #40express) is 20 m away. Get off at “INSA” bus stop.

New! line #41express is direct from train station to the Campus (week days).
Departures: 7h34, 7h45, 8h03 and 8h15.
Back to train sation (“INSA” bus stop): 16h05, 16h42, 17h10, 17h48, 18h08 and 18h08.

From “Place de la Mairie” take bus #6 and get off at “Tournebride” bus stop; then take bus #16 (or #40Ex, #41Ex, #16Ex) and get off at “INSA”.

From city center, take bus #16 or #40Ex from “Republique” or “Pont Pasteur”.

From INSA bus stop, follow the arrows !!

How to get to Rennes

Conference Information

List of Hotels

The participants should book directly the Hotels. Accomodation at INSA at competitive prices will be provided (this information will be added in the future). Otherwise, a list is provided in the following.

Rennes hotels

Some links …

Announcement

Reinforcement learning (RL) is a growing research area within the machine learning community. To try to build an European community of people working in reinforcement learning, and to give visibility to the current situation in the old continent, we are running a now biennal series of workshops. EWRL-3 will take place this year (1997) in Rennes, France The first two editions of the workshop took place in 1994 in Brussels, Belgium (EWRL-1, the First European Workshop on Reinforcement Learning), and in 1995 in Milano, Italy (EWRL-2, the Second European Workshop on Reinforcement Learning). The two-day workshop will be organized as follows. The first morning will be opened by a plenary talk given by Andrew Moore.

The rest of the workshop will be dedicated to presentations given by selected workshop participants. The length of presentations will be decided after we have some feedback on the number of participants.

The workshop will be restricted to researchers active in the field and to their students.

An inscription fee of FF 1,000 (roughly $170 at current change) will cover local organization expenses, coffee breaks, and a social dinner on Monday evening.

Please communicate ASAP your intention to participate by means of the intention form below (e-mail is preferred: mdorigo@ulb.ac.be).
Otherwise send intention forms to: Marco Dorigo, IRIDIA, Université Libre de Bruxelles, Avenue Franklin Roosvelt 50, CP 194/6, 1050 Bruxelles, Belgium.

Preliminary Program

Monday 13 October 1997

08:30 – Registration
09:00 – Introductory remarks
09:15 – Commercial Reinforcement Learning (Invited talk)
Andrew MOORE, Smith Hall 221, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh PA 15213, USA. awm@cs.cmu.edu
10:15 – Coffee break
10:45 – Combining Dynamic Programming and Neurocontrol: Some Basic Issues
Hugues BERSINI, Mauro BIRATTARI and Gianluca BONTEMPI, IRIDIA, Université Libre de Bruxelles, Bruxelles, Belgium
11:15 – Distributing Local Models for Reinforcement Learning
Tomas LANDELIUS, Linköping University, Linköping, Sweden
Get a copy of the talk slides (1826K)
12:00 – Comparing Reinforcement Learning Algorithms Applied to LCS
Andrea BONARINI, PM AI&R, Politecnico di Milano, Italy
12:40 – Free time for lunch
(Lunch is up to the attendees. A list of restaurants and cafeterias in the neighborhood will be provided.)
14:10 – Fuzzy Q-Learning
Pierre-Yves GLORENNEC and Lionel JOUFFE, INSA de Rennes, France
14:50 – Distributed Reinforcement Agents for Adaptive Routing in Communication Networks
Gianni DI CARO and Marco DORIGO, IRIDIA, Université Libre de Bruxelles, Bruxelles, Belgium
Here you can find related papers.
15:30 – Coffee break
15:50 – Parallel Q-learning with an Application to Telecommunication
Ann NOWE and K. STEENHAUT, Vrije Universiteit Brussel, Belgium
16:25 – Synthesis of Collective Behavior by Multiple Agents via Temporal and Structural Credit Assignments
Norihiko ONO, Faculty of Engineering, University of Tokushima, Tokushima, Japan
17:00 – The Theoretical Aspect of QT-Learning and its Using for the Optimal Management of a Water Network
Nathalie CHATENET, CEMAGREF, Groupement de Bordeaux, Cestas, France
17:35 – Reinforcement Learning in Finite Horizon: Optimality Criteria Equivalence and the ODE Method
Frederick GARCIA, Station de Biometrie et d’Intelligence Artificielle, INRA, France
20:30 – Social dinner (organized by EWRL-3)

Tuesday 14 October 1997

09:00 – Gradient Descent Method for Model Based Reinforcement Learning
Vincenzo CAIRONI, PM AI&R, Politecnico di Milano, Italy
09:35 – Reinforcement Learning in the Continuous Case: Approximating the Value Function by Finite-elements and Finite-differences Methods
Remi MUNOS, Laboratoire d’Ingenierie des Systemes Complexes, CEMAGREF, Antony, France
10:10 – Coffee break
10:40 – Combining an Exploration/Exploitation Method with State Generalization
Edwin DE JONG, VUB AI Lab, Brussels, Belgium
Get a copy of the talk slides (509K)
Here you can find a related paper.
11:15 – Learning Control for Continuous Mimo Dynamical Systems Using Neuro Dynamic Programming
Martin RIEDMILLER, Universität Karlsruhe, Germany
11:50 – Optimal Active Suspension Of Vehicles Using Adaptive Critic Methods
Birgit KOSLIK, Zentrum Mathematik, Technische Universität, München, Germany
12:25 – Free time for lunch
(Lunch is up to the attendees. A list of restaurants and cafeterias in the neighborhood will be provided.)
14:00 – Prediction Based Reinforcement Learning for Complex, Dynamic, and Chaotic Environments
Sadayoshi MIKAMI, Laboratory of Chaotic Systems Engineering, Research Group of Complex Systems Engineering, Hokkaido University, Japan
14:35 – Adaptive Climate Control in an Office Environment
Olle GÄLLMO and Patrik LÖGDAHL, Uppsala university, Dept. of Computer Systems, Sweden
15:10 – Coffee break
15:30 – Exploration of Multi-States Environments: Local Measures and Back-Propagation of Uncertainty
Nicolas MEULEAU, Cemagref, Laboratoire d’Ingenierie des Systemes Complexes, Antony cedex, France
16:05 – Final discussion

Organizing and Program Committee

Hugues Bersini and Marco Dorigo, IRIDIA, ULB, Brussels, Belgium
Pierre Yves Glorennec, Lionel Jouffe, INSA de Rennes, France
Eric Anquetil, IRISA de Rennes, France

List of Participants with emails

Intention Form

INTENTION FORM

Third European Workshop on Reinforcement Learning (EWRL-3)

Rennes, France, October 13-14, 1997

I intend to participate and will give a presentation with the following tentative title:

Family Name:

First Name:

Institution:

Address:

Phone No.:

Fax No.:

E-mail: