Learning in multilevel games with incomplete information. II

摘要

Multilevel games are abstractions of situations where decisionnmakers are distributed in a network environment. In Part I of thisnpaper, the authors present several of the challenging problems thatnarise in the analysis of multilevel games. In this paper a specific setnup is considered where the two games being played are zero-sum games andnwhere the decision makers use the linear reward-inaction algorithm ofnstochastic learning automata. It is shown that the effective game matrixnis decided by the willingness and the ability to cooperate and is anconvex combination of two zero-sum game matrices. Analysis of thenproperties of this effective game matrix and the convergence of thendecision process shows that players tend toward noncooperation in thesenspecific environments. Simulation results illustrate this noncooperativenbehavior