COMPENSATORY FDNF-FCNF NEUROFUZZY SYSTEMS FOR FUZZY GAMES

In this paper, a new adaptive fuzzy reasoning technique using compensa tory DNF-CNF operations is proposed according to the compensatory prin ciple and interval valued fuzzy sets. The compensatory FDNF-FCNF neuro fuzzy system using the control-oriented fuzzy neurons and the decision -oriented fuzzy neurons can not only adjust fuzzy membership functions bur also optimize the adaptive fuzzy reasoning by using the compensat ory learning algorithm. This system can effectively be used to learn t he fuzzy rules of two players in a game from given data, then transfor ms a local game to a global game, and finally makes better fuzzy moves based on the global game. In addition, simulations have indicated tha t the convergence speed of the compensatory FDNF-FCNF learning algorit hm is faster than that of the conventional backpropagation algorithm.