Towards better population sizing for differential evolution through active population analysis with complex network

Abstract

This research paper presents an analysis of the population activity in Differential Evolution algorithm (DE) during the optimization process. A state-of-art DE variant – Success-History based Adaptive DE (SHADE) is used and the population activity is analyzed through Complex Network (CN) created from mutation, crossover and selection steps. The analysis is done on the CEC2015 benchmark set and possible future research directions for the population sizing are suggested.