基于动态多子群协作QPSO算法的RBFNN优化  

Optimization of RBFNN Based on Dynamic Multiple Sub-population Collaboration Quantum-behaved Particle Swarm Optimization Algorithm

作　　者：卫晓娟[1] 李宁洲[1] 周学舟 丁杰[1] 丁旺才[1] 

机构地区：[1]兰州交通大学机电工程学院,甘肃兰州730070

出　　处：《兰州交通大学学报》2014年第3期98-103,共6页Journal of Lanzhou Jiaotong University

基　　金：国家自然科学基金(11162007);甘肃省自然科学基金(1308RJZA149)

摘　　要：提出了一种动态多子群协作QPSO算法(Dynamic Multiple Sub-population Collaboration Quantum-behaved Particle Swarm Optimization,简称DMPQPSO),该方法动态构建各子群,并采用混沌策略分2个阶段优化QPSO,同时对各子群的收缩扩张系数分别进行自适应调整.采用该方法优化RBFNN,并将DMPQPSO算法与标准PSO和QPSO算法对比,仿真实验验证了该方法的优化效果.A Dynamic Multiple Sub-population Collaboration Quantum-behaved Particle Swarm Optimization is proposed for parameters identification of RBFNN.The method dynamically builds each sub-population,and the chaotic strategy is adopted to optimize the Quantum-behaved Particle Swarm Optimization （QPSO）algorithm in the two stages of search process,at the same time,the contraction expansion coefficient of the algorithm is adjusted adaptively in the evolutionary process according to the fitness of each particle.The proposed method is used to optimize RBFNN,and compared with standard PSO and QPSO.The simulation results show that the optimized effect is enhanced.

关 键 词：DMPQPSO 减聚类算法 RBFNN 非线性问题处理性能 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]