An Enhanced Multi-Population Ensemble Differential Evolution

MPEDE integrates multiple effective strategies to solve optimization problems. However, there is still some room to improve the optimization performance of it. In this work, we introduce an enhanced multi-population ensemble DE (eMPEDE). In the proposed algorithm, an improved mutation strategy "rand-to-mpbest/1" replaces "rand/1" in MPEDE to balance the exploration and exploitation, which utilizes multiple best solutions to guide searching. Moreover, an improved parameter adaptation method is employed to alleviate premature convergence by using success-history based adaptation. The experiments on CEC2005 benchmark problems are executed, including a comparison with other peer competitors. The experimental results reveal the capability of eMPEDE to generate more competitive results compared to MPEDE and other peer competitors.