基于APC-Ⅲ算法在线训练径向基神经网络

ON-LINE Training RBF Networks Based On APC-ⅢCluster Algorithm

针对径向基函数神经网络隐层节点难以确定的问题,提出了基于APc-Ⅲ聚类算法的思想来在线训练网络结构的方法,使用聚类算法对输入样本数据进行模式分类,以此自适应地调整RBF神经网络的结构和参数,解决了网络结构的在线优化问题。应用该方法进行非线性系统的实时辩识,仿真结果证明了本文提出的在线训练算法的优越性。

This paper describes online training algorithm based on APC-Ⅲ cluster algorithm to construct Radial Basis Function neural network. By this algorithm, it is self-adaptively and optimally adjusted the structure and parameters of the hidden layer of the RBFNN. The superiority of this arithmetic is proved by a simulation example applied to a realtime identification of non-linear systems.