X. Lu, H. M. Schwartz and S. N. Givigi (2011) "Policy Invariance under Reward Transformations for General-Sum Stochastic Games", Volume 41, pages 397-406

PDF | PostScript | doi:10.1613/jair.3384

We extend the potential-based shaping method from Markov decision processes to multi-player general-sum stochastic games. We prove that the Nash equilibria in a stochastic game remains unchanged after potential-based shaping is applied to the environment. The property of policy invariance provides a possible way of speeding convergence when learning to play a stochastic game.

Click here to return to Volume 41 contents list