すみ分け型進化戦略を用いた免疫アルゴリズムの動的制御

抄録

For the problems that have plural optimal solutions like as scheduling problems or simultaneous equa-tions, it is desired to get plural solutions simultaneously. It has been reported that Immune Algorithm (IA) is an useful method to solve such problems. But various IA parameters such as crossover rate should be set optimally. About this setting problem of the IA parameters, the methods to control the IA in which the IA parameters are changed dynamically by using Meta GA have been reported. However, they had a problem of long searching time though they had good performance. In contrast with it, Evolution Strategy (ES) has the system as that each ES individual in the population evolves independently because the individuals evolve basically by mutations without crossovers. So searching time can be shortened when ES is used as Meta algorithm. In this paper, a method of dynamic control of the IA by using the ESNM (Evolution Strategy with Niche Method) is proposed in order to realize efficient IA search. And this method is applied to simultaneous nonlinear equations, and its performance is compared with that obtained by usual method.