摘要
提出了适用于单层神经网络快速学习算法分析的一种新模型——广义系统辨识模型,分析了Karayiannis的快速BP算法.研究结果表明:利用所提出的新模型。
The single layer NN fast learning algorithm is a gradient descent method. By adjusting the connection weights of the neural network, the value of the generalized error function for the training of neural network is a minimum. A generalized system identifying model for the single layer neural network is proposed. It is suitable for the statistical analysis of the fast backpropagation algorithm. By means of the model proposed, the mean weight behavior and generalized error of Karayiannisfast BP algorithm are investigated. It is shown that ,although the weights grow unbounded, the fast BP algorithm will perform the training quickly.
出处
《华中理工大学学报》
CSCD
北大核心
1996年第8期21-23,共3页
Journal of Huazhong University of Science and Technology
关键词
神经网络
辨识模型
快速学习算法
neural network
identifying model
fast learning algorithm