摘要
介绍BP神经网络结构和学习方法,针对误差反向传播神经网络模型学习收敛速度慢、容易陷入局部极小点等缺点,本文对BP网络模型进行了改进。对原始数据采用非线性的归一化函数,提出一种更加有效的学习率改进算法,提高了网络的收敛速度,采用了一种新的权值及阈值初始化方法,以避免训练时误差陷入局部极小解,并对改进BP算法与传统的BP算法进行比较,验证了该算法的优越性。
This text introduce the BP neural network structure and study method. Carry on an improvement to BP neural network to overcome the shortcomings of the standard BP neural network. Study ofnolinear normalization of original data from objects a method of improving the convergence speed of BP algorithm by using new study rate parameter. And compare the improved algorithm with the former algorithm, the experiment is presented to show that the improved algorithm is effective.
出处
《自动化与仪器仪表》
2007年第6期77-80,共4页
Automation & Instrumentation
关键词
BP网络
改进算法
故障诊断
学习率参数
BP Neural network
Improved algorithm
Fault diagnosis
Study rate parameter