摘要
提出一种确定RBF网络隐含层神经元和权值的有效方法。该方法将自动聚类算法与对称距离相结合优化每个隐含层神经元的中心向量;利用伪逆方法确定隐层神经元到输出神经元的权值。实验结果表明:该方法比自动聚类算法有更好的分类能力。
This paper proposed a new method that can efficiently determine hide layer neurons and weights of radial basis function network (RBFN). The method combined auto - clustering algorithm (ACA) and symmetry distance to improve each hide layer neuron ; And the weight from hide layer neu- ron to output layer neuron were determined to use pseud - inverse method. The simulation display that the proposed method can obtain better results than ACA.
出处
《微处理机》
2006年第4期48-49,52,共3页
Microprocessors
关键词
径向基函数神经网络
自动聚类
对称距离
伪逆
Radial basis function network
Auto - clustering
Symmetry distance
Pseudo - inverse