Matthew Bourque : Policy Improvement for Stochastic Games

Posted by Gabriel Conant , part of the Graduate Student Colloquium.

At

April 29, 2013, 4:15 p.m.

In

SEO 636

Abstract

Policy improvement is a well known and popular algorithm for solving Markov decision processes. It has been extended to solve some classes of discounted stochastic games. I will discuss the extension of policy improvement to solve average-payoff stochastic games with additive rewards and additive transitions.