A War-Based Parallel Genetic Algorithm with Variable Population-Size for Multimodal Optimization without Constraints

The purpose of this paper is to show a new parallel genetic algorithm based on war. Each population has its own resources and keep consuming it until exhaust it, then the population go after new resources which are obtained by means of war. This mechanism allows the variation of the winning population size bringing back both the resources and the best individuals to its own population. The proposed algorithm was compared with an ordinary GA and a serial version of itself aiming to get quality results, comparisons, speed up and efficiency. Moreover, we compare the proposed algorithm with the parallel island model, proving that our approach is more efficient in the set of the considered multimodal benchmarks. All comparisons are based on statistical test for supporting our results.