Global Impact Balancing in the Hierarchic Genetic Search

Abstract

The new Globally Balanced Hierarchic Genetic Strategy (GB-HGS) was introduced as a tool for solving difficult global optimization problems. This strategy provides a multi-deme economic stochastic search with an adaptive accuracy that allows many local extremes of the objective to be found. The strategy was designed according to the Multi Agent System (MAS) paradigm. The novelty of GB-HGS derives from its control of the search impact performed by various demes on the basis of the global information gathered and exchanged among the computing agents. This mechanism is applied together with the local profiling of the computational process already used in the previous versions of hierarchic genetic computations. The new strategy exhibits better efficiency, especially in the second phase of computations, when the promising regions containing the global extremes are encountered.