一种改进RBF-PSO算法的极值寻优方法
An extremal optimization method of improved RBF-PSO algorithm
徐富强1
作者信息
摘要
如何在有限的实验数据下寻找最优实验条件与实验结果,一直是研究人员关心的问题.本文提出了一种基于RBF神经网络和改进的PSO算法的极值寻优方法.该方法利用径向基(RBF)神经网络结构简单、可调参数少、训练简洁且收敛速度快等特点,将有限的实验结果和对应的实验条件逼近为某一非线性函数,再利用具有收敛快和通用性强的改进粒子群优化算法(PSO)结合最佳RBF网络寻找最优值.文章通过3个实例验证并与常见的BP-PSO算法进行比较,表明改进的RBF-PSO算法达到较好的寻优效果,该算法具有较好的稳定性和应用性.
Abstract
How to get the best experimental condition and experimental results from limited experimental data is a problem researchers have concerned for a long time. The paper proposes an optimization method basing on RBF neural networks and improved PSO algorithm: in virtue of adopting radial-basis function(RBF) neural networks' simple structure, few adjustable parameters, brevity and fast convergence speed, the method approximates limited experimental results and their corresponding experimental data to some nonlinear function and uses the improved Particle Swarm Optimization ( PSO) algorithm, which has fast convergence as well as versatility, with the combination of the best RBF networks to find the best value. The paper will also illustrate the better optimal effect, stability and validity of the improved RBF-PSO algorithm by means of three example confirmations and a comparative study with the usual BP-PSO algorithm.
关键词
极值寻优/RBF神经网络/PSO算法Key words
Extremal optimization/RBF neural network/PSO algorithm引用本文复制引用
基金项目
巢湖学院自然科学基金(XLY-201101)
巢湖学院自然科学基金(XLY-201102)
出版年
2012
NETL
NSTL
维普
万方数据