A group work inspired generation alternation model of real-coded GA
Takatoshi Niwa, Koya Ihara, Shōhei Kato · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2019
Genetic algorithms (GAs) are stochastic optimization methods that mimic the evolution of living organisms. Among them, GA which uses real-coded genes is called Real-Coded GA (RCGA). Several generation alternation models have been proposed to improve RCGA. Problem decomposition is a known method for improving search performance. Therefore, in this study, a practical, real-world method of problem decomposition was considered. Here, it was considered that the individual evaluation method utilized in group work could be applied to the individual evaluation of RCGA. We focused on the group work which practice problem decomposition in the real world. In group work, each of members collaborates with the others and is evaluated comprehensively from the group's outcome and personal contribution to the group. Therefore, in this study a model is proposed to introduce the concept of group work into a generation alternation model of RCGA. We evaluated the search performance of the proposed method using benchmark functions and confirmed the performance improvement by comparing with the conventional methods.