Self-adaptive differential evolution algorithm with discrete mutation control parameters

TLDR: A self-adaptive DE algorithm with discrete mutation control parameters (DMPSADE) is proposed that was compared with 8 state-of-the-art DE variants and 3 non-DE algorithms by using 25 benchmark functions and indicates that the average performance of DMPSADE is better than those of all other competitors.

Abstract: In DMPSADE, control parameters and mutation strategies could be automatically adjusted.We first proposed a new encoding for parameter control in DE algorithm.Roulette wheel is used to implement the selection of mutation strategies. Generally, the optimization problem has different relationships (i.e., linear, approximately linear, non-linear, or highly non-linear) with different optimized variables. The choices of control parameters and mutation strategies would directly affect the performance of differential evolution (DE) algorithm in satisfying the evolution requirement of each optimized variable and balancing its exploitation and exploration capabilities. Therefore, a self-adaptive DE algorithm with discrete mutation control parameters (DMPSADE) is proposed. In DMPSADE, each variable of each individual has its own mutation control parameter, and each individual has its own crossover control parameter and mutation strategy. DMPSADE was compared with 8 state-of-the-art DE variants and 3 non-DE algorithms by using 25 benchmark functions. The statistical results indicate that the average performance of DMPSADE is better than those of all other competitors.