基于改进自构形学习算法的RBF网络结构优化

Based on Selfstructure Learning Algorithm Structural Optimization of RBF Network

RBF网络中的隐层神经元的数目直接影响着整个网络的性能和效率,因而对RBF网络的结构优化是一个非常必要的环节。本文先采用分步式训练构造初始RBF网络,然后利用改进的神经网络自构形学习算法对所构造的RBF网络的隐层进行优化,最后通过实验结果的分析与对比,验证改进的神经网络自构形学习算法对RBF网络优化的有效性。

The number of hide layer neural units in RBF neural network affects the whole network's quality and efficiency directly.So it is a very necessary step to optimize the structure of RBF network.The thesis constructs an initial RBF network firstly by training in steps,then optimizes the network's hide layer with modified neural network selfstructure learning algorithm.Finally,the thesis proves the validity of the modified neural network selfstructure learning algorithm for structural optimization of RBF network by analyzing and comparing the expermental results.