采用人群搜索算法的径向基函数神经网络参数整定方法

Parameter Adjusting Method of Radial Basis Function Neural Network Using Seeker Optimization Algorithm

针对径向基函数(RBF)神经网络的逼近结构中,对权值、基宽和中心向量的初始值等参数的选取不当,导致系统的鲁棒性变差、收敛精度降低,甚至不再收敛的问题,提出一种基于人群搜索算法的RBF神经网络的参数整定方法.以基于遗传算法和基于粒子群算法的RBF神经网络参数整定方法为对比条件,采用MATLAB软件进行实验与分析.结果表明:应用人群搜索算法去优化RBF神经网络的初始参数,能有效地提升RBF神经网络的逼近精度,验证了该算法的可行性.

In the approximation structure of radial basis function(RBF)neural network,its improper parameters such as weight,base width and initial value of the center vector will leads to poor system robustness and reduced convergence accuracy,or even no convergence.Aiming at this problem,the parameter adjusting method of RBF neural network based on seeker optimization algorithm is proposed.And using the parameter adjusting method of RBF neural network based on genetic algorithm and particle swarm optimization as comparison,the experiment and analysis are completed by MATLAB software.The results show that the seeker optimization algorithm can optimize the initial parameters of RBF neural network,so that the approximation accuracy of RBF neural network is effectively improved,and the feasibility of the algorithm is verified.