基于聚类的RBF-LBF串联神经网络学习算法  

Learning algorithm for cascade RBF-LBF neural networks based on clustering

作　　者：唐勇智[1] 葛洪伟[1] 

机构地区：[1]江南大学信息工程学院,江苏无锡214122

出　　处：《计算机应用》2007年第12期2916-2918,共3页journal of Computer Applications

摘　　要：为提高网络的泛化能力,研究了单层RBF神经网络和LBF网络组成的RBF-LBF串联神经网络,并提出了一种基于模式聚类的RBF-LBF串联神经网络的学习算法。该算法分别对单层RBF网络和LBF网络的输入进行模式聚类,以确定网络的初始结构,然后通过调整错分样本的类别,使之部分重叠或合并核函数。经双螺旋线问题仿真实验证明,该算法确具有很好的泛化能力且只需较短的训练时间。To enhance the generalization ability of the network, a kind of cascade RBF-LBF neural networks consisting of single-lager Radial Basis Function （RBF） and Linear Basis Function （LBF） networks was proposed, and a pattern-based clustering of cascade RBF-LBF neural networks structure learning algorithm was presented. The input of RBF and LBF neural networks was trained by using pattern clustering to determine the initial structure of the network. Then the samples were adjusted in wrong class to overlap or merge the kernel function. Experiment of two spirals problem proves that the algorithm has higher generalization ability and needs shorter training time.

关 键 词：泛化能力 神经网络 径基函数 模式聚类 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]