| 关于BP网络学习算法中步长上界的估计 |
| |
| 引用本文： | 贺江峰,车海平,陈增强,袁著祉.关于BP网络学习算法中步长上界的估计[J].系统工程与电子技术,1995(12). |
| |
| 作者姓名： | 贺江峰 车海平 陈增强 袁著祉 |
| |
| 作者单位： | 南开大学计算机与系统科学系 |
| |
| 摘 要： | 本文首先给出了ＢＰ算法中关于步长取值使目标学习函数递减的充分条件，然后，具体给出了步长的一个上界估计值，最后，对于估计结果做了进一步讨论。本文给出的结果对于构造各种收敛的ＢＰ网络算法具有重要意义。
|
| 关 键 词： | 计算机技术，网络，＋神经网络，算法。 |
Estimation of a Superbound of Step Factor in the BPN Learning Algorithm |
| |
| He Jiangfeng,Che Haiping,Chen Zengqiang and,Yuan Zhuzhi.Estimation of a Superbound of Step Factor in the BPN Learning Algorithm[J].System Engineering and Electronics,1995(12). |
| |
| Authors: | He Jiangfeng Che Haiping Chen Zengqiang and Yuan Zhuzhi |
| |
| Abstract: | In the paper,a sufficient condition on the step factor guaranteeing decrease of the objective function is proposed firstly,then,a superbound value of step factor is estimated,finally a more discussion on the estimation is made.The results that we propose give a guarantee for convergence ofgradient-based BPN Learning algorithms. |
| |
| Keywords: | NN BPN Learning algorithm Step factor |
| 本文献已被 CNKI 等数据库收录！ |
|
