Stochastic Adaptive Behavior for Prisoner's Dilemma