Personal tools
Distributed Strategic Learning in Dynamic Robust Games
| What |
|
|---|---|
| When |
Nov 10, 2010 from 03:00 PM to 04:00 PM |
| Where | Engr IV Maxwell Room 57-124 |
| Add event to calendar |
|
Hamidou Tembine
Supelec, Paris, France
Wednesday, November 10, 2010 at 3:00pm
Engr IV Maxwell Room 57-124
Abstract
We consider finite number of potential players in a dynamic and uncertain environment. Each player can be in active mode or in sleep mode. At each time, a random set of players interact. Each of the active players updates their strategies and estimations after measurement/observations of the delayed noisy payoffs which depends on the strategy profiles and the state of the nature. We provide a novel heterogeneous delayed combined fully distributed payoff and strategy reinforcement learning (Delayed-CODIPAS-RL) to capture realistic behaviors of the players behavior allowing for the possibility of limited capabilities and computations but also for heterogeneity and constrained updating. We analyze the asymptotic behavior of the dynamic robust game under average payoffs. In contrast to the standard learning analysis which are limited to finite number of players, we extend the Delayed-CODIPAS-RL to a class of mean field learning. The convergence to an explicit mean field dynamics is provided when the size of the system goes to infinity. We then introduce adaptive hybrid learning where each player can change its learning patterns during the long-run interactions. Global convergence to equilibria for specific classes of games are discussed.
Biography
Hamidou Tembine is assistant professor at Ecole Superieure d'Electricite (Supelec, Gif-sur-Yvette, France). His main research interests are evolutionary games and their applications. Hamidou Tembine received in 2006 Master degrees in applied mathematics from Ecole Polytechnique (Palaiseau, France) and in pure mathematics from University Joseph Fourier (Grenoble, France). He received the PhD degree in Computer Science entitled population games with networking pplications from University of Avignon in 2009. From 2007 to 2009, he has been research assistant at the Computer Science Department of University of Avignon and teacher assistant at University of Aix-Marseille. He has been visiting researcher at University of McGill (Montreal, Quebec, Canada), Ecole Polytechnique de Montreal (Quebec, Canada), University of Illinois at Urbana-Champaign (UIUC, US), Ecole Polytechnique Federale de Lausanne (EPFL, Switzerland) and University of Wisconsin (Madison, US).
