Decision strategies in evolutionary optimization

Abstract

This paper describes a sequel of the previous experiments on real number genetic algorithm behavior [2], [3]. A particular example of multicriteria optimization is discussed. The behavior of the two previously explored genetic algorithms is compared with a simple evolutionary algorithm. The main idea of the experiments is to stimulate the algorithm to find the Pareto set without measuring dominance and non-dominance. The implication of maxi-min decision method is affecting optimization so that the final solutions lay closer to the Pareto set than those obtained without any decision method. This theoretical concept is tested and analyzed graphically by picturing populations after a certain number of generations. The differences in the algorithm behavior and causes of such differences are explained. © Springer-Verlag 2001.