Differential evolution with explicit control of diversity for constrained optimization
Gabriel Vázquez, Carlos Segura · 2020
Evolutionary Algorithms (EAs) have been quite successful both in constrained and unconstrained single-objective optimization. However, in the constrained case little attention has been paid to controlling the diversity explicitly with the aim of avoiding premature convergence. This paper proposes Differential Evolution with Clustering-based Diversity (DECD). DECD uses a novel replacement strategy that controls the diversity explicitly. Particularly, the replacement strategy forces the selection of some distant individuals but it simultaneously allows the survival of some other individuals that form clusters to promote intensification. The experimental validation shows the important benefits provided by the explicit control of diversity.