Efficient differential evolution using speciation for multimodal function optimization

In this paper differential evolution is extended by using the notion of speciation for solving multimodal optimization problems. The proposed species-based DE (SDE) is able to locate multiple global optima simultaneously through adaptive formation of multiple species (or subpopulations) in a DE population at each iteration step. Each species functions as a DE by itself. Successive local improvements through species formation can eventually transform into global improvements in identifying multiple global optima. In this study the performance of SDE is compared with another recently proposed DE variant Crowding DE. The computational complexity of SDE, the effect of population size and species radius on SDE are investigated. SDE is found to be more computationally e cient than CrowdingDE over a number of benchmark multimodal test functions.