Enhanced differential evolution with modified parent selection technique for numerical optimisation

Abstract

Differential evolution (DE) is considered to be one of the most prominent evolutionary algorithms for numerical optimisation. However, it may suffer from the slow convergence rate, especially in the late stage of evolution progress. The reason might be that the parents in the mutation operator are randomly selected from the parent population. To remedy this limitation and to enhance the performance of DE, in this paper, a modified parent selection technique is proposed, where the parents in the mutation operator are chosen based on their previous successful experiences. The major advantages of the proposed parent selection technique are its simplicity and generality. It does not destroy the simple structure of DE, and it can be used in most DE variants. To verify the performance of the proposed technique, it is integrated into the classical DE algorithm and three advanced DE variants. Thirteen widely used benchmark functions are used as the test suite. Experimental results indicate that the proposed technique is able to enhance the performance of the classical DE and advanced DE algorithms in terms of both the quality of final solutions and the convergence rate.