摘要
由于误差函数的高维复杂性 ,BP网络在目前的应用中存在训练速度慢、甚至导致网络系统瘫痪的问题 ,针对训练中的归一化问题、隐层节点数的选取、样本数目的增减法。
As higt-dimensional complexity of error function,'BP'network exists slow training speed in the present application,even leads to palsy of network system,in accordance with normalization problem in the training,it was made researches on selection of implicit layer node,add-and-subtract method of sample number,determination of the whole learning efficiency and training algorithm.
出处
《信息技术》
2002年第1期4-6,共3页
Information Technology
