Adaptive dynamics in games played by heterogeneous populations (Q1566893)

There are two theorems conveying the following information. (1) When the probability of imitators is small, the outcome of the collective ``learning'' process is close to the dynamically stable Nash equilibria of the game, and the asymptotic behaviour is similar to the case, where agents always choose best replies. (2) The second theorem deals with the game which has a unique, interior equilibrium. If at least one of the coordinates of the mixed equilibrium equals \(\frac 12\), then no matter however small \(\varepsilon\) is, the process does not converge to a neighbourhood of the equilibrium with positive probability provided that the same size is sufficiently large. A suitable generalization of the Bendixson stability theory combined with stochastic approximation theory is employed.