Self-learning differential evolution algorithm for dynamic polycentric problems

Abstract

A novel self-learning differential evolution algorithm is proposed to solve dynamical multi-center optimization problems.The approach of re-evaluating some specific individuals is used to monitor environmental changes.The proposed self-learning operator guides the evolutionary group to a new environment,meanwhile maintains the stable topology structure of group to maintain the current evolutionary trend.A neighborhood search mechanism and a random immigrant mechanism are adapted to make a tradeoff between algorithmic convergence and population diversity.The experiment studies on a periodic dynamic function set suits are done,and the comparisons with peer algorithms show that the self-learning differential algorithm outperforms other algorithms in term of convergence and adaptability under dynamical environment.

Keywords

• evolutionary computation;dynamic optimization;self-learning mechanism;differential evolution