Differential evolution with two-level adaptive mechanism for numerical optimization

This paper presents a novel differential evolution algorithm by designing two prediction-based muta-tion operators and a two-level adaptive mechanism. To accelerate the search efficiency of algorithm and enhance the suitability of integrating multiple mutation operators with different characteristics, two new mutation operators are first developed to strengthen the local search ability by making full use of the neighborhood information of each individual to predict its promising region and guide the search process. Meanwhile, a two-level adaptive mechanism is designed to meet the search requirements of both individual and population by utilizing the history search information of its neighbors and that of the whole population simultaneously to assign a more suitable mutation operator for it. Moreover, an enhanced restart operation is further adopted to avoid the invalid search during the evolutionary process. Differing from the existing DE variants, the proposed algorithm incorporates the prediction process in mutation to enhance the exploitation, and simultaneously uses the local information of each individual and the global information of population to adaptively adjust the search capability. Then it could effectively improve the search efficiency and balance the exploration and exploitation. Compared with 21 typical algorithms, the numerical results on the benchmark functions from both IEEE CEC2014 and CEC2017 show that the proposed algorithm has better performance. (c) 2022 Elsevier B.V. All rights reserved.

Automated summary

This paper presents a novel differential evolution algorithm that utilizes prediction and adaptive mechanisms to improve search efficiency and balance exploration and exploitation, demonstrating better performance.