| 网络泛化能力与随机扩展训练集 |
| |
| 引用本文： | 杨慧中,卢鹏飞,张素贞,陶振麟. 网络泛化能力与随机扩展训练集[J]. 控制理论与应用, 2002, 19(6): 963-966 |
| |
| 作者姓名： | 杨慧中 卢鹏飞 张素贞 陶振麟 |
| |
| 作者单位： | 1. 江南大学通信与控制工程学院,江苏,无锡,214036
2. 华东理工大学自动化研究所,上海,200237 |
| |
| 摘 要： | 针对神经网络的过拟合和泛化能力差的问题, 研究了样本数据的输入输出混合概率密度函数的局部最大熵密度估计, 提出了运用Chebyshev不等式的样本参数按类分批自校正方法, 以此估计拉伸样本集, 得到新的随机扩充训练集. 使估计质量更高, 效果更好. 仿真结果证明用这种方法训练的前馈神经网络具有较好的泛化性能.
|
| 关 键 词： | 前馈神经网络 泛化能力 最大局部熵密度函数 Chebyshev不等式 |
| 文章编号： | 1000-8152(2002)06-0963-04 |
| 收稿时间： | 2001-11-28 |
| 修稿时间： | 2002-07-01 |
Generalization of networks and random expanded training sets |
| |
| YANG Hui-zhong+,+,+,+ (. College of Communication , Control Engineering,Southern Yangtze University,Jiangsu Wuxi,China; . Research Institute of Automation,East China University of Science,Shanghai,Chin,LU Peng-fei,ZHANG Su-zhen and TAO Zhen-lin. Generalization of networks and random expanded training sets[J]. Control Theory & Applications, 2002, 19(6): 963-966 |
| |
| Authors: | YANG Hui-zhong+,+,+,+ (. College of Communication & Control Engineering,Southern Yangtze University,Jiangsu Wuxi,China . Research Institute of Automation,East China University of Science,Shanghai,Chin,LU Peng-fei,ZHANG Su-zhen TAO Zhen-lin |
| |
| Affiliation: | College of Communication & Control Engineering, Southern Yangtze University,Jiangsu Wuxi 214036,China;College of Communication & Control Engineering, Southern Yangtze University,Jiangsu Wuxi 214036,China;Research Institute of Automation, East China University of Science, Shanghai 200237,China;Research Institute of Automation, East China University of Science, Shanghai 200237,China |
| |
| Abstract: | Aiming at the problems of over-fitting and generalization for neural networks, the locally most entropic colored Gaussian joint input-output probability density function (PDF) estimate is studied, and a new method by means of Chebyshev inequality is proposed to self-revise respectively according to every cluster. In terms of the method, a random expanded training set is obtained. The simulation results illustrate that generalization of feed forward neural networks using the expanded training sets is greatly improved. |
| |
| Keywords: | feed forward neural networks generalization locally most entropic probability density function Chebyshev inequality |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
| 点击此处可从《控制理论与应用》浏览原始摘要信息 |
|
点击此处可从《控制理论与应用》下载全文 |
|
