摘要
分别从BP网络的学习步长,学习速率自适应调整算法的参数,动量法和自适应学习速率结合起来算法的参数3方面讨论了改进BP参数对网络识别能力的影响;在确定BP网络的隐含层节点个数的过程中提出了BP神经网络自适应学习算法,使得隐层节点的选取动态实现。仿真实验表明,该改进是可行的。
This paper discusses the influence of BP parameter to the recognition capability of network through varying the step of BP network, parameters of adjusting learning rate adaptively and parameters of integrating momentum method and self-adapting method. During the course of determining the number of nodes in the hidden units of BP network, the self-adapting algorithm of BP neural network has been proposed and it makes the selection of hidden units dynamically in the course of training and improves the adaptive ability of network. The simulation results show the improvement is effective.
出处
《计算机工程》
CAS
CSCD
北大核心
2005年第B07期154-156,共3页
Computer Engineering
关键词
人工神经网络
BP算法
自适应
Artificial neural network
BP algorithm
Self-adapting