GaMeDE2 — improved Gap–based Memetic Differential Evolution applied to Multimodal Optimization

This paper presents an improved Gap–based Memetic Differential Evolution (GaMeDE2), the modification of the GaMeDE method, which took second place in the GECCO 2020 Competition on Niching Methods for Multimodal Optimization. GaMeDE2 has reduced complexity, fewer parameters, redefined initialization, selection operator, and removed processing phases. The method is verified using standard benchmark function sets (classic ones and CEC2013) and a newly proposed benchmark set comprised of deceptive functions. A detailed comparison to state-of-the-art methods (like HVCMO and SDLCSDE) is presented, where the proposed GaMeDE2 outperforms or gives similar results to other methods. The document is concluded by discussing various insights on the problem instances and the methods created as a part of the research.