一种新型BP神经网络的生成算法  

A Novel Training Method for BP Neural Network

作　　者：赵春雨[1] 韩学军[1] 丛恒斌[2] 闻邦椿[2] 

机构地区：[1]东北大学黄金学院 [2]东北大学机械工程学院

出　　处：《黄金学报》1999年第2期106-109,共4页Gold Journal

摘　　要：提出了一种新型的神经网络的训练方法，包括输入层节点的自动寻找技术和隐含层节点的自动生成技术．为了验证该算法的正确性，利用该算法对同向回转观机驱动振动系统进行了神经网络逼近分析，计算机模拟结果表明该算法可以准确地确定输入层和隐含层节点的数目．A novel training method for BP neural network is proposed in this paper,which involves automatic finding the nodes of input layer and automatic generation of hidden layers.In order to verify the eHiciency of the proposed algorithm, a computer simulation is carried out, that a BP neural network is used to study the dynamics of vibrating machine with twin-motor rotating in the same direction. The results of computer simulation show that the proposed methed can effectively find the nodes of input layer and generate the nodes of hidden layers.

关 键 词：神经网络 输入层节点 隐含层节点 生成算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]