Multiple individual guided differential evolution with time varying and feedback information-based control parameters

Abstract

Differential evolution (DE) is a simple and efficient metaheuristic algorithm for solving global optimization problems. It is used widely in various fields due to its concise structure and strong search ability. In this study, an extended version of the DE named multiple individual guided differential evolution (MGDE) is proposed. The MGDE is distinguished by introducing a novel mutation strategy based on multiple guiding individuals of the DE population to manage the diversity and convergence. The base vector of the mutation strategy is defined as a center of guiding individuals and the difference vectors are assigned to perform a search towards one of the top fitted and top diversified individuals available in the population. The control parameters of the DE are adjusted in a way to provide a suitable transition from exploration to exploitation and to utilize the information of recent success history of evolution. The performance of the proposed MGDE is evaluated on three different benchmark sets including IEEE CEC2014, IEEE CEC2017, and IEEE CEC2011 of real-world problems. Different performance metrics such as average and standard deviation of fitness, ranking of algorithms, Wilcoxon signed-rank test, and convergence analysis are used to analyze and compare the results of the MGDE with several other metaheuristic algorithms. Comparison attest that the proposed MGDE algorithm is highly competitive with the other metaheuristic algorithms.