摘要
本文引入模糊控制机制,对文献[1]的具有正态模糊网络参数的前馈式模糊神经网络学习算法进行改进,提出了一种效率更好的F-BP学习算法。在此算法中,采用近似模糊推理技术来确定网络的学习率、动量因子、加速系数三个学习参数,使得这些学习参数在网络的学习过程中,根据学习时间的长短、误差大小及误差变化情况,进行动态调整,从而提高学习效率。最后,通过实例考查了F-BP学习算法的性能,并讨论了学习参数的调整对学习效率的影响。
This paper,by introducing the fuzzy control mechanism, improved the learning algorithm of the neural network with bell - shaped fuzzy parameter in [ 1 ] . We proposed a more efficient F - BP learning algorithm, hi this algorithm, we cited the approximate fuzzy inference technique to determine the three learning coefficients of the network. they include the learning - rate coefficient, the momentum coefficient and the accelerator coefficient, and adjusted these learning coefficients at the same time by the learning time,the error and the change of the error.Thereby,which in above improved the learning efficiency. At last, we examine and analyse the ability of the proposed learning algorithm, and discussed the influence of adjusting these there learning coefficients on learning efficiency.
出处
《计算机应用与软件》
CSCD
2000年第8期21-31,共11页
Computer Applications and Software
