A Self-Adaptive Differential Evolution with Dynamic Selecting Mutation Strategy

Abstract

A self-adaptive differential evolution with dynamic selecting mutation strategy (DSMSDE) is proposed to improve the performance of differential evolution algorithm by three improvements. Mutation strategies are dynamically selected, and the successfully updated individuals are stored into the archive, which is beneficial for improving the convergence performance. A mechanism that is related to the best individual at the current population is employed to help the stagnation solutions to get rid of local minima. Self-adaptive parameters control is used to accelerate the convergence speed. DSMSDE is compared with the other state-of-the-art algorithms, and they are tested on nine benchmark functions. Experimental results show that DSMSDE has higher accuracy, faster speed and better reliability.