摘要
从BP神经网络的优点与缺点入手,对于其所存在的问题进行了探讨,并根据BP神经网络的问题,较全面地罗列了数种对BP神经网络训练优化的方法,如:数据的初始化、初始权值、隐层数的确定方法,传递函数、学习率和权值修改的优化方法,以及其它优化方法·
The paper introduces problems of BP neural network begin with the advantage and disadvantage of it and have enumerated more than ten kinds methods to optimize the BP neural network. For instance, initialization of data, definite methods of initial weights and hidden layers, optimize methods of transmit function, learning rate and right value revise, and others.
出处
《怀化学院学报》
2006年第2期126-130,共5页
Journal of Huaihua University
关键词
神经网络
BP算法
优化方法
neural network
BP algorithm
improvement methods
