Differential evolution with distributed direction information based mutation operators: an optimization technique for big data

Abstract

With the rapid advance in networking, data storage, and data collection technique, big data is fast expanding in various scientific and engineering fields, such as physical, social and biological sciences. Thanks to solving difficult optimization problems without detailed prior knowledge, evolutionary algorithm (EA) has become a powerful optimization technique for dealing with complex problems in big data. This study focuses on differential evolution (DE), which is one of the most successful and popular EAs and distinguishes from other EAs with its mutation mechanism. However, for the mutation operators of most DE algorithms, the base and difference vectors are always randomly selected from the whole population, where the population information is not utilized effectively. In this study, a novel DE framework with distributed direction information based mutation operators (DE-DDI) is proposed. In DE-DDI, the distributed topology is employed to create a neighborhood for each individual in the population first and then the direction information derived from the neighbors is introduced into the mutation operator of DE. Therefore, the neighborhood and direction information are fully utilized to exploit the regions of better individuals and guide the search to the promising area. In order to test the performance of the proposed algorithm, DE-DDI is applied to several original DE algorithms, as well as the advanced DE variants. The results clearly indicate that DE-DDI is able to improve the performance of the DE algorithms studied.