论文部分内容阅读
从信息扩散的角度对RBF(RadialBasisFunction)网络进行了分析.从理论上证明了RBF网络具有信息扩散功能,并分析了其网络的物理意义.说明了根据正态扩散的择近原则确定RBF网络中规划因子的合理性,也说明了在训练样本数目大时用聚类方法确定中心参数的优越性.实验结果证明这种方法设计的RBF网络分类器性能优于一般RBF网络分类器.
The Radial Basis Function (RBF) network is analyzed from the perspective of information diffusion. It is theoretically proved that RBF network has the function of information diffusion and the physical meaning of its network is analyzed. The reasonability of planning factors in RBF network is proved according to the principle of proximity of normal diffusion, and the superiority of clustering method in determining center parameters when the number of training samples is large is illustrated. The experimental results show that the RBF network classifier designed by this method performs better than the general RBF network classifier.