Enhance Differential Evolution with Random Walk

Abstract

This paper proposes a novel differential evolution (DE) algorithm with random walk (DE-RW). Random walk is a famous phenomenon universally exists in nature and society. As random walk is an erratic movement that can go in any direction and go to any place, it is likely that this mechanism can be used in search algorithm to bring in diversity. We apply the random walk mechanism into conventional DE variants with different parameters. Experiments are conducted on a set of benchmark functions with different characteristics to demonstrate the advantages of random walk in avoiding local optima. Experimental results show that DE-RWs have general better performance than their corresponding conventional DE variants, especially on multimodal functions.