Michael Kearns Professor of Computer and Information Science Department
                             University of Pennsylvania
                             National Center Chair in Resource Management and Technology
                             Secondary Appointment, Operations and Information Management, Wharton 

"Graphical Models and Game Theory"

Judea Pearl's work established the fundamental insight that articulating a network structure of statistical dependence could have deep representational and algorithmic benefits. In the case of Bayesian networks and related models, edges represent the probabilistic influence of one random variable on another in a possibly complex joint distribution. Inspired by the work of Pearl and others, in this decade a number of researchers have developed similar models, algorithms and theory for strategic, rather than probabilistic, interactions. Network structure now represents the influence that one player or agent has on the payoffs of others in a multi-party game or economy. We survey such "graphical models for game theory", including algorithms for equilibrium computation, connections to probabilistic inference, and their recent role in the resolution of the complexity of finding Nash equilibria.

Thursday, April 17th, 2008
10:40am -  11:20 pm
IRCS

3401 Walnut Street, Room 400A

For more information regarding our speaker please visit:

http://www.cis.upenn.edu/~mkearns/