摘要
BP网络在实际应用中的最大障碍就是学习速度慢这一问题。本文在详细分析与论证的基础上得出:在学习过程中,通过对sigmoid函数形状的动态调整可有效地提高BP网络的学习速度,而且这一方法实现简单、易于理解、具有普遍适用性。同时本文对近期发表的有关文献中所提出的一些不恰当的观点和结论进行了修正和补充。
The greatest difficulty for BP network to apply in practice is the low learning speed. In this paper, based on detailed analysis and experiments, we get that: In the process of learning, the learning speed can be improved effectively through adjusting the form of Sigmoid function dynamicaly.At the same time, we point out some unsuitable view points and conclusions in some papers published recently.
出处
《安徽大学学报(自然科学版)》
CAS
1998年第3期64-67,共4页
Journal of Anhui University(Natural Science Edition)
基金
安徽省教委重点项目基金