Int. J. of Wireless and Mobile Computing   »   2015 Vol.8, No.2

 

 

Title: Top-k elites based oppositional differential evolution

 

Authors: Jianming Zhang; Weifeng Pan; Jingjing Wu; Jing Wang

 

Addresses:
School of Computer Science and Information Engineering, Zhejiang Gongshang University, Hangzhou 310018, China
School of Computer Science and Information Engineering, Zhejiang Gongshang University, Hangzhou 310018, China
School of Computer Science and Information Engineering, Zhejiang Gongshang University, Hangzhou 310018, China
School of Software and Communication Engineering, Jiangxi University of Finance and Economics, Nanchang 330013, China

 

Abstract: Opposition-based Differential Evolution (ODE) is a new DE variant with faster convergence speed and more robust search abilities than the classical DE. It utilises the concept of opposition and simultaneously evaluates an estimate and its corresponding opposite estimate. Numerical results have shown that the selection of the symmetry point between the estimate and the opposite estimate affects the performance of ODE variants. To make full use of the information included in the elites, this paper presents a novel DE variant, Top-k elites-based Oppositional Differential Evolution (TEODE), which is based on a new opposition-based learning strategy using the top-k elites (TEOBL) in the current generation and employs similar schemes of ODE for population initialisation and generation jumping with TEOBL. Experiments are conducted on 17 benchmark functions. The results confirm that TEODE outperforms classical DE, ODE and COODE (opposition-based differential evolution using the current optimum).

 

Keywords: differential evolution; opposition-based learning; function optimisation; top-k elites; convergence speed; search abilities.

 

DOI: 10.1504/IJWMC.2015.068634

 

Int. J. of Wireless and Mobile Computing, 2015 Vol.8, No.2, pp.166 - 174

 

Submission date: 12 Jul 2014
Date of acceptance: 28 Aug 2014
Available online: 28 Mar 2015

 

 

Editors Full text accessAccess for SubscribersPurchase this articleComment on this article