Elitist artificial bee colony with dynamic population size for multimodal optimization problems

Many real-world problems can be formulated as a multimodal optimization problem (MMOP), and metaheuristic algorithms used in solving MMOP have to find multiple optimal points simultaneously. The key requirement for dealing with such problems is to balance exploration capability in global space and exploitation in multiple optimal spaces. Artificial bee colony (ABC), a metaheuristic algorithm, is designed to find only a single global optimum and cannot solve the MMOP. In this paper, we propose an ABC variant named “Elitist ABC with Dynamic Population Size” to cope with multimodal optimization problems. It has a dynamic population size strategy and uses a search equation selection strategy powered by elite members. The dynamic population size strategy enhances the exploration capability of the algorithm. The search equation selection strategy determines the appropriate search behavior for a particular problem instance at runtime. Thus, exploitation and exploration behaviors can be adjusted adaptively. In addition, candidate optimum peaks, that are overlooked in the original ABC algorithm, are memorized with elite population members. The proposed algorithm has been tested on multimodal optimization problems presented at CEC 2013. The algorithm has been compared with ten state-of-the-art multimodal optimization algorithms and the top 25 algorithms participating in the CEC competition on multimodal function optimization between 2013 and 2020. Experimental results have shown that the proposed algorithm is superior to many new algorithms and can compete with top-level algorithms.