Title: Differential evolution algorithm with dynamic population scheme

Authors: Xinyu Zhou; Mingwen Wang; Jianyi Wan

Addresses: School of Computer and Information Engineering, Jiangxi Normal University, Nanchang 330022, China ' School of Computer and Information Engineering, Jiangxi Normal University, Nanchang 330022, China ' School of Computer and Information Engineering, Jiangxi Normal University, Nanchang 330022, China

Abstract: Differential evolution (DE) is an efficient population-based evolutionary algorithm for solving optimisation problems. Its control parameters have significant influence on performance. In the past few years, how to adjust the scaling factor F and crossover probability CR has attracted a lot of attention. However, few works have been focused on the parameter of population size. So in this paper, we designed a dynamic population scheme to manage the population size of DE. This scheme mainly consists of two parts: the logistic model and opposition-based learning. The first part is dedicated to calculate how many new individuals should be added into the population, while the second part is used to generate these new individuals. A series of experiments is conducted on 25 well-known benchmark functions including shifted and rotated ones. Results show that our approach shows promising performance.

Keywords: differential evolution; population size; logistic modelling; opposition-based learning; OBL; dynamic population.

DOI: 10.1504/IJWMC.2016.077212

International Journal of Wireless and Mobile Computing, 2016 Vol.10 No.3, pp.261 - 271

Available online: 23 Jun 2016 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article