Title: Learning in a Class of Non-Zero-Sum Two-Person Games Part I
Abstract: We consider a finite two person game where the players move in turn. Neither player knows the payoffs of the other, but each player has an a-priori-probability distribution for the possible "types" of his opponent and is able to form an opinion of this a-priori-probability of his opponent. Each player follows the Bayesian rule of maximizing his payoff expectation. "Learning" in this context means that a player narrows down the possible "types" of his opponent by observing and analyzing his moves in the course of the game. Thus he is able to correct his a-priori-probabilities and to improve his decisions. Asymptotically he may know exactly against whom he is playing. It is shown that a unique solution results which under certain conditions converges to the subgame perfect Nash equilibrium point of the game under perfect information.
Creation-Date: May 1991
Unfortunately this paper is not available online. Please contact us to order a hardcopy.
SFB 303 Homepage
17.02.1998, © Webmaster