Please use this identifier to cite or link to this item:
Title: Force-imitated particle swarm optimization using the near-neighbor effect for locating multiple optima
Authors: Liu, L
Yang, S
Wang, D
Keywords: Particle swarm optimization;Multimodal optimization problem;Near-neighbor effect;Force-imitated particle dynamics
Issue Date: 2012
Publisher: Elsevier Inc
Citation: Information Sciences, 182(1), 139 - 155, Jan 2012
Abstract: Multimodal optimization problems pose a great challenge of locating multiple optima simultaneously in the search space to the particle swarm optimization (PSO) community. In this paper, the motion principle of particles in PSO is extended by using the near-neighbor effect in mechanical theory, which is a universal phenomenon in nature and society. In the proposed near-neighbor effect based force-imitated PSO (NN-FPSO) algorithm, each particle explores the promising regions where it resides under the composite forces produced by the “near-neighbor attractor” and “near-neighbor repeller”, which are selected from the set of memorized personal best positions and the current swarm based on the principles of “superior-and-nearer” and “inferior-and-nearer”, respectively. These two forces pull and push a particle to search for the nearby optimum. Hence, particles can simultaneously locate multiple optima quickly and precisely. Experiments are carried out to investigate the performance of NN-FPSO in comparison with a number of state-of-the-art PSO algorithms for locating multiple optima over a series of multimodal benchmark test functions. The experimental results indicate that the proposed NN-FPSO algorithm can efficiently locate multiple optima in multimodal fitness landscapes.
Description: Copyright @ Elsevier Inc. All rights reserved.
ISSN: 0020-0255
Appears in Collections:Publications
Computer Science
Dept of Computer Science Research Papers

Files in This Item:
File Description SizeFormat 
Fulltext.pdf355.36 kBAdobe PDFView/Open

Items in BURA are protected by copyright, with all rights reserved, unless otherwise indicated.