论文部分内容阅读
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.
How to find optimal experimental conditions and experimental results under limited experimental data has always been a concern for researchers.This paper presents an extremum optimization method based on RBF neural network and improved PSO algorithm.The method uses radial RBF neural network has the advantages of simple structure, few adjustable parameters, simple training and fast convergence speed. The finite experimental results and the corresponding experimental conditions are approximated to a certain nonlinear function, and then they have the advantages of fast convergence and high universality (PSO) is combined with the best RBF network to find the optimal value.Through the verification of three examples and comparison with the common BP-PSO algorithm, this paper shows that the improved RBF-PSO algorithm can achieve better optimization results , The algorithm has good stability and applicability.