Enhanced Differential Evolution by Dynamic Selection Framework of Mutation Operator

Abstract

Muatation operator is the core operators in differential evolution(DE). Many researchers have proposed various versions of improved muatation operators. Most of them balance the exploration and exploitation in a constant way, which means that the importance between exploration and exploitation is invariable in the whole evolution process. However, exploration is more important in the initial evolution stage and exploitation is more important in the latter evolution stage. So we propose a framework named $\varepsilon$ ranking based mutation operators in DE $\pmb{(\varepsilon \mathrm{R}-\mathrm{DE})}$ in this paper. The framework balance exploration and exploitation in a dynamic way according to the evolution stages. Experiments on 20 benchmark functions of CEC2013 indicate that $\pmb\varepsilon \mathbf{R}$ -DE improves the original DE and advanced DE very well.