| 神经网络中处理鞍点的LMBP改进算法 |
| |
| 引用本文： | 李炯城,肖恒辉,李桂愉.神经网络中处理鞍点的LMBP改进算法[J].计算机工程,2012,38(23):173-176,180. |
| |
| 作者姓名： | 李炯城 肖恒辉 李桂愉 |
| |
| 作者单位： | 广东省电信规划设计院有限公司广州市无线网络优化重点工程中心,广州,510630 |
| |
| 基金项目： | 广东省教育部产学研结合基金资助项目，广州市软件(动漫)产业发展基金资助项目 |
| |
| 摘 要： | 针对目前神经网络中的Levenberg-Marquardt反向传播(LMBP)算法在训练过程中有可能迭代到鞍点的问题，提出一种能有效克服鞍点的LMBP改进算法。计算鞍点处雅克比矩阵的正特征值对应的特征向量并将其作为新的搜索方向。通过实例对比传统LMBP算法与改进LMBP算法的效果，证明改进的算法能有效地脱离鞍点并进一步收敛到极小点处。
|
| 关 键 词： | 神经网络 Levenberg-Marquardt反向传播算法 鞍点 雅克比矩阵 海森矩阵 高斯-牛顿法 |
| 收稿时间： | 2012-01-17 |
Improvement Algorithm of Levenberg-Marquardt Back Propagation for Saddle Point in Neural Network |
| |
| LI Jiong-cheng
,
XIAO Heng-hui
,
LI Gui-yu.Improvement Algorithm of Levenberg-Marquardt Back Propagation for Saddle Point in Neural Network[J].Computer Engineering,2012,38(23):173-176,180. |
| |
| Authors: | LI Jiong-cheng XIAO Heng-hui LI Gui-yu |
| |
| Affiliation: | (Key Wireless Network Optimization Center of Guangzhou, Guangdong Planning and Designing Institute of Telecommunications Co., Ltd., Guangzhou 510630, China) |
| |
| Abstract: | Aiming at the Levenberg-Marquardt Back Propagation(LMBP) algorithm of neural network sometimes converges to the saddle point during training process, an improved LMBP algorithm which can overcome the saddle point effectively is proposed. All eigenvectors for all the positive eigenvalue of Jacobi matrix are calculated as new searching directions. The improved LMBP algorithm is proved that it can get out of saddle point, and it iterates to minima effectively by an example of comparing with the traditional LMBP algorithm and the improved one. |
| |
| Keywords: | neural network Levenberg-Marquardt Back Propagation(LMBP) algorithm saddle poin Jacobi matrix Hessian matrix Gauss- Newton algorithm |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
| 点击此处可从《计算机工程》浏览原始摘要信息 |
|
点击此处可从《计算机工程》下载全文 |
