Empirical Adaptation of Control Parameters in Differential Evolution Algorithm

Abstract

Proper settings of two control parameters in differential evolution (DE) algorithm play a crucial role when solving various optimisation problems. There are a lot of investigated adaptive mechanisms sampling the values of these parameters differently. In the proposed adaptive approach, efficient and inefficient settings of both DE control parameters are considered. A sampling of the control parameters is performed using the performance of parameters’ settings achieved in a preliminary experimental study. The proposed approach is applied to a frequently used adaptive jDE algorithm. The statistical analysis of the results provides better information about the efficiency of the proposed technique. More proper settings in the adaptation of DE control parameters provide better performance than a more general approach when solving CEC 2017 problems.