A Univariate Marginal Distribution Resampling Differential Evolution Algorithm with Multi-Mutation Strategy

Mutation strategy is an important issue of differential evolution (DE). The efforts of developing new mutation strategies have received more attention. In this paper, a univariate marginal distribution resampling differential evolution algorithm with multi-mutation strategy (UMDE-MS). Univariate marginal distribution algorithm continuous (UMDAc) has been proved to be efficient in nonseparable problems and it is less influenced by the correlation between variables. In the suggested algorithm, a univariate marginal distribution resampling mechanism inspired by this feature of UMDAc is proposed to reinitialize the population when evolution comes to a standstill. Moreover, two novel adaptive mutation strategies are proposed and integrated to build a mutation strategy pool. A parameter pool which consists of four popular F and Cr settings is employed in UMDE-MS. Finally, the proposed algorithm is tested on the CEC2019 100-Digit challenge benchmark problems. Experimental results demonstrate the effectiveness of UMDE-MS compared with other state-of-the-art algorithms.