论文部分内容阅读
针对BP网络学习速率和动量项参数较难选取以及学习过程中学习效率较为低下的问题,提出BP网络的改进算法模型—AB网络模型,来选取学习速率和动量项的参数值,即通过一个为给定先验知识的A网,动态调节另一个执行实际应用的B网中的学习速率和动量项的参数值,并以此提高整个网络的学习效率.实验结果表明,通过AB网络自适应调整参数的算法比普通BP算法的学习效率大大提高.在实际应用中,我们可以通过AB网络自适应调节的方法,对学习速率参数和动量项参数进行合适的选取.
Aiming at the problems that BP network learning rate and momentum term parameters are difficult to select and the learning efficiency is low in learning process, an improved algorithm model of BP network -AB network model is proposed to select learning velocity and momentum parameter values, namely, Given the prior knowledge of A network, dynamically adjust the other implementation of the actual application of the B network learning rate and momentum term parameter values, and thus improve the learning efficiency of the entire network.The experimental results show that through the AB network adaptive adjustment The algorithm of parameter is much higher than that of ordinary BP algorithm.In the practical application, we can make appropriate choice of learning rate and momentum parameters through adaptive adjustment of AB network.