A Generalized Reinforcement-Learning Model: Convergence and Applications Export

@techreport{citeulike:5971642, abstract = {Reinforcement learning is the process by which an autonomous agent uses its experience interacting with an environment to improve its behavior. The Markov decision process (MDP) model is a popular way of formalizing the reinforcement-learning problem, but it is by no means the only way. In this paper, we show how many of the important theoretical results concerning reinforcement learning in MDPs extend to a generalized MDP model that includes MDPs, two-player games and MDPs under a worst-case optimality criterion as special cases. The basis of this extension is a stochastic-approximation theorem that reduces asynchronous convergence to synchronous convergence. Keywords: Reinforcement learning, Q-learning convergence, Markov games}, address = {Providence, RI, USA}, author = {Littman, Michael L. and Szepesv\'{a}ri, Csaba}, citeulike-article-id = {5971642}, citeulike-linkout-0 = {http://portal.acm.org/citation.cfm?id=864918}, institution = {Brown University}, keywords = {1996, game-theory, reinforcement-learning}, posted-at = {2009-10-19 17:53:28}, priority = {5}, publisher = {Brown University}, title = {A Generalized Reinforcement-Learning Model: Convergence and Applications}, url = {http://portal.acm.org/citation.cfm?id=864918}, year = {1996} }

TY - RPRT ID - citeulike:5971642 L3 - citeulike-article-id:5971642 N2 - Reinforcement learning is the process by which an autonomous agent uses its experience interacting with an environment to improve its behavior. The Markov decision process (MDP) model is a popular way of formalizing the reinforcement-learning problem, but it is by no means the only way. In this paper, we show how many of the important theoretical results concerning reinforcement learning in MDPs extend to a generalized MDP model that includes MDPs, two-player games and MDPs under a worst-case optimality criterion as special cases. The basis of this extension is a stochastic-approximation theorem that reduces asynchronous convergence to synchronous convergence. Keywords: Reinforcement learning, Q-learning convergence, Markov games AD - Providence, RI, USA TI - A Generalized Reinforcement-Learning Model: Convergence and Applications PB - Brown University KW - 1996 KW - game-theory KW - reinforcement-learning AU - Littman, Michael AU - Szepesvári, Csaba PY - 1996/// UR - http://portal.acm.org/citation.cfm?id=864918 ER -