A dual dynamic migration policy for island model genetic algorithm

Abstract

The common problem in island model is the way to migrate individual from one to another island, or usually called as migration policy. Previous researches in this movement protocol could be categorized into two different approaches, diversity preservation based such as migration protocol in LIMGA and better island pursuing based such as new dynamic migration policy. The main purpose of this works is to introduce a brand-new migration mechanism called as Dual Dynamic Migration Policy (DDMP) for island model GA. DDMP will take the advantage from result pursuer of dynamic migration policy and convergence avoider of LIMGA's migration protocol. The experiment result shows that DDMP could give great result while carrying out the general optimization cases. It could produce the best score result for all cases among previous island model migration methods. This work also compares DDMP with the-best-known-so-far solution for the problem set.