Multi-strategy Genetic Algorithm for Self-Configuring Solving of Complex Optimization Problems

Many complex optimization problems require a modification of the general evolutionary algorithm (EA) according to the given features of the problem. There exist a great variety of EAs that represent different search strategies for many classes and subclasses of optimization problems. Real-world problems may combine several features that are not known beforehand, thus there is no information about what EA to choose and what EA´s settings to apply for efficient problem solving. This study presents a novel metaheuristic for designing multi-strategy EA based on the hybrid of the island model, cooperative and competitive co evolution schemes. The approach controls interactions of EAs and leads to the self-configuring solving of problems with a priori unknown structure. Two examples of implementations of the approach for multi-objective and non-stationary optimization are discussed. The results of numerical experiments for benchmark problems from CEC competitions are presented. The proposed approach has demonstrated the efficiency comparable with other well-studied techniques for multi-objective and non-stationary optimization. And it does not require the participation of the human-expert, because it operates in an automated, self-configuring way.