摘要
利用多层网络BP算法对复杂的函数进行逼近来讨论BP网络训练过程的训练误差和检验误差的关系,详细对隐节点的个数及取最优隐节点分布、复相关系数、检验误差以及网络输入为维数与计算复杂度的关系进行讨论,最后分析并检验所得关系的正确性。
The multi-layer network BP algorithm is used to discuss the relationship between the training error and the examination error by the complicated function approximation. The number of the hiding nodes, the distribution of the optional hiding nodes, the complex related coefficient, the examination error, the input dimension of the network and the degree of the computation complexity are discussed in detail. The results show that the CR = ε[（n＋ 1）H]^R/n is valid in 64.3 %.
出处
《成都信息工程学院学报》
2006年第1期70-73,共4页
Journal of Chengdu University of Information Technology
关键词
隐节点
复相关系数
网络输入维数
计算复杂度
hiding nodes
complex related coefficient
input dimension of network
computation complexity
