A modified particle swarm optimization algorithm based on dynamic learning factors and sharing method

To improve the global convergence ability and rate of particle swarm optimization, an improved particle swarm optimization algorithm based on dynamic learning factors and sharing method is proposed. The inertia weight factor of the algorithm decreases non-linearly, and the learning factor changes dy...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Zhejiang da xue xue bao. Journal of Zhejiang University. Sciences edition. Li xue ban Jg. 43; H. 6; S. 696 - 700
Hauptverfasser: Tan, Yifeng, Sun, Tingting, Xu, Xinming
Format: Journal Article
Sprache:Chinesisch
Veröffentlicht: Zhejiang University Press 01.11.2016
Schlagworte:
Algorithms
Control
Convergence
Fitness
Machine learning
Nonlinear dynamics
Optimization
Swarm intelligence
共享适应度
动态
学习因子
粒子群算法
ISSN:1008-9497
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:To improve the global convergence ability and rate of particle swarm optimization, an improved particle swarm optimization algorithm based on dynamic learning factors and sharing method is proposed. The inertia weight factor of the algorithm decreases non-linearly, and the learning factor changes dynamically with the descending. A sharing fitness function is introduced on the basis of dynamic regulation. When the algorithm is stagnated without reaching termination, part of the particles will be selected according to the distance between particles and optimal solution. The chosen particles will be re-initialized as a new swarm and be evaluated by sharing fitness. The old and new swarms follow their own local solutions respectively until the end of the iteration. Simulation results of four typical multimodal functions show that the modified algorithm can greatly enhance the rate of the optimal solution searching and improve the global convergence performance of PSO.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
ISSN:1008-9497
DOI:10.3785/j.issn.1008-9497.2016.06.014