Performance analysis of DE mutation schemes for constrained LSGO problems

Abstract

Abstract Differential Evolution (DE) is one of the most popular evolution algorithms (EAs) for numerical optimization problems with various difficulty levels. Mutation scheme makes a strong influence in DE performance. It produces new trial vectors based on the current population to improve the previous best-found solution. In this paper, we have analyzed the performance of a wide range of various mutation schemes for early proposed ε-CC-SHADE algorithm. We have used scaled benchmark set of constrained large-scale global optimization (cLSGO) problems to investigate ε-CC-SHADE performance. The new benchmark set is based on IEEE CEC 2017 Competition on Constrained Real-Parameter Optimization. The experimental results have shown the dependence of EA performance on the selected mutation scheme for solving optimization problems with high dimension. Based on the numerical results, we have made conclusions for choosing mutation scheme to increase DE performance for cLSGO.