| 基于Kalman滤波的基本Elman网络训练新算法 |
| |
| 引用本文： | 李小兵,田玉松,邱天爽.基于Kalman滤波的基本Elman网络训练新算法[J].大连理工大学学报,2009,49(2):276-281. |
| |
| 作者姓名： | 李小兵 田玉松 邱天爽 |
| |
| 作者单位： | 大连理工大学电子与信息工程学院,辽宁,大连,116024 |
| |
| 基金项目： | 国家自然科学基金资助项目(60372081,30570475,60872122);;辽宁省科学技术基金资助项目(2001101057);;教育部博士学科点专项科研基金资助项目(20050141025) |
| |
| 摘 要： | Kalman滤波算法应用于基本Elman网络学习时,收敛速度较快,但收敛精度往往不高;而基于梯度下降的BP算法可以以很高的精度实现输入输出的非线性映射,但在极值点处收敛速度缓慢.针对上述问题,提出一种将Kalman滤波算法应用于基本Elman网络的新学习训练算法.该算法结合Kalman滤波算法和基于梯度下降的BP算法的优点来训练网络,以基本Elman网络隐层单元输出作为非线性系统的状态变量,通过Kalman滤波算法实现状态变量的快速准确跟踪,然后通过梯度下降法修正权值以保证精度.另外,在训练过程中,通过增加训练样本的信息内容来提高网络收敛的精度.仿真结果表明了该算法的有效性.
|
| 关 键 词： | Kalman滤波 梯度下降算法 基本Elman网络 |
A new training algorithm based on Kalman filter of essential Elman network |
| |
| LI Xiaobing,TIAN Yusong,QIU Tianshuang.A new training algorithm based on Kalman filter of essential Elman network[J].Journal of Dalian University of Technology,2009,49(2):276-281. |
| |
| Authors: | LI Xiaobing TIAN Yusong QIU Tianshuang |
| |
| Institution: | School of Electronic and Information Engineering;Dalian University of Technology;Dalian 116024;China |
| |
| Abstract: | In the application of essential Elman network,the Kalman filter algorithm has high convergence speed but low precision;on the contrary,the BP algorithm with gradient descend technique has high precision to map the output with input nonlinearly,but usually with low convergence speed at extreme point.Therefore,a new algorithm based on Kalman filter is proposed to accomplish essential Elman network training.The advantages of Kalman filter and BP algorithm are combined to optimize the new algorithm.The outputs ... |
| |
| Keywords: | Kalman filter gradient descend algorithm essential Elman network |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
| 点击此处可从《大连理工大学学报》浏览原始摘要信息 |
| 点击此处可从《大连理工大学学报》下载免费的PDF全文 |
