A String Kernel Density Estimation Algorithm for Repeated Games

Renan Motta GoulartSaul de Castro LeiteRaul Fonseca Neto

Repeated games is a branch of game theory, where a game can be played several times by the players involved. In this setting, it is assumed that the players do not always play in the optimal Min-Max strategy and that they may be willing to collaborate. A learn's task is then to detect this behavior in its opponents and play as to maximize its own profits. In this paper, it is pro- posed a novel String Kernel Density Estimation (SKDE) algorithm that makes on-line short term prediction in two player's game and can rapidly adapt to changing strategies by the opponent. We show, in the experiments section, that the algorithm performs well against a set of opponent strategies and is supe- rior in respect to memory consumption and computational time to the Entropy Learning Prune Hypothesis Space (ELPH) algorithm.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: