A shade-based algorithm for large scale global optimization

Abstract

During the last decade, large-scale global optimization has been a very active research area not only because of its many challenges but also because of its high applicability. It is indeed crucial to develop more effective search strategies to explore large search spaces considering limited computational resources. In this paper, we propose a new hybrid algorithm called Global and Local search using Success-History Based Parameter Adaptation for Differential Evolution (GL-SHADE) which was specifically designed for large-scale global optimization. Our proposed approach uses two populations that evolve differently allowing them to complement each other during the search process. One is in charge of exploring the search space while the other is in charge of exploiting it. Our proposed method is evaluated using the CEC’2013 large-scale global optimization (LSGO) test suite with 1000 decision variables. Our experimental results show that the new proposal outperforms one of the best hybrid algorithms available in the state of the art (SHADEILS) in the majority of the test problems adopted while being competitive with respect to several other state-of-the-art algorithms when using the LSGO competition criteria adopted at CEC’2019.