Multi-search differential evolution algorithm

The differential evolution algorithm (DE) has been shown to be a very simple and effective evolutionary algorithm. Recently, DE has been successfully used for the numerical optimization. In this paper, first, based on the fitness value of each individual, the population is partitioned into three subpopulations with different size. Then, a dynamically adjusting method is used to change the three subpopulation group sizes based on the previous successful rate of different mutation strategies. Second, inspired by the "DE/current to pbest/1", three mutation strategies including "DE/current to cbest/1", "DE/current to rbest/1" and "DE/current to fbest/1" are proposed to take on the responsibility for either exploitation or exploration. Finally, a novel effective parameter adaptation method is designed to automatically tune the parameter F and CR in DE algorithm. In order to validate the effectiveness of MSDE, it is tested on ten benchmark functions chosen from literature. Compared with some evolution algorithms from literature, MSDE performs better in most of the benchmark problems.