基于OLS算法的RBF网设计方法与实现  

Design Method and Implementation of RBF Network Based on OLS Algorithm

作　　者：刘金玲[1] 

机构地区：[1]徐州广播电视大学铜山分校,江苏徐州221116

出　　处：《计算机时代》2010年第7期21-22,25,共3页Computer Era

摘　　要：在RBF网络学习中用随机数据点作为径向基函数中心,计算神经网络的权,这样的结果往往会造成网络过大或者结果不理想,从而无法满足要求。径向基函数中心的选择基于正交最小二乘法,逐个找出径向基函数的中心,构建出良好的网络。该算法选择的中心最大值不会产生数字的病态问题。正交最小二乘法策略简单、有效,设计的径向基函数网络效果很好。In RBF network learning,using random data points as radial basis function centers to compute neural network weights,the result is often large network or unsatisfactory,and thereby cannot meet the requirements. Based on orthogonal least squares method,selecting radial basis function centers can find the centers one by one to build good network,and the maximum of the center selected by the algorithm can avoid ill-conditioned problem. The orthogonal least squares strategy is simple and effective,and the designed radial basis function network has a very good effect.

关 键 词：基函数 最小二乘法 神经网络 回归 正交 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]