ANALYSIS OF THE BEHAVIOR OF MULTILEVEL HIERARCHICAL SYSTEMS OF LEARNING AUTOMATA AND THEIR APPLICATION FOR MULTIMODAL FUNCTIONS OPTIMIZATION

Multilevel hierarchical systems of learning stochastic automata are in vestigated in this paper. The system under consideration consists of s eueral levels of automata with different number of outputs. Each autom aton is a variable-structure stochastic automaton. A learning scheme w hich is based on the Bush-Mosteller reinforcement scheme is used to ad just the probabilities associated with the actions of the automata of the hier archical learning system. Boolean and continuous automaton in puts have been considered. Convergence and convergence rate analysis a re presented. The optimization of multimodal functions using this mult ilevel system of automata is also described. Simulation results indica te the effectiveness of such multilevel learning systems.