| BP神经网络快速收敛算法研究 |
| |
| 引用本文： | 王赟松,褚福磊,何永勇,郭丹. BP神经网络快速收敛算法研究[J]. 农业机械学报, 2004, 35(6): 182-184 |
| |
| 作者姓名： | 王赟松 褚福磊 何永勇 郭丹 |
| |
| 作者单位： | 清华大学精密仪器与机械学系,博士后,100084,北京市;清华大学精密仪器与机械学系,教授,博士生导师;清华大学精密仪器与机械学系,副教授 |
| |
| 基金项目： | 教育部跨世纪优秀人才培养基金资助项目,中国博士后科学基金资助项目 (项目编号 :2 0 0 40 3 5 0 0 5 9),山东省自然科学基金资助项目(项目编号 :Y2 0 0 2 F17) |
| |
| 摘 要： | 以标准BP算法为基础,应用Levenberg—Marquardt优化算法,提出了一种快速收敛的BP算法——LMBP算法。仿真结果表明,与标准BP算法及其他改进形式比较,LMBP算法收敛速度大大提高,稳定性并未降低,这为BP神经网络应用于实时性要求高的场合(如在线检测)提供了算法基础。该算法的缺点是计算量大,所需计算机内存大,不适合大型网络的计算。
|
| 关 键 词： | 神经网络 LMBP算法 Levenberg-Marquardt优化算法 |
| 修稿时间： | 2003-03-31 |
New and Rapidly Converging BP Neural Network Algorithm |
| |
| Wang Yunsong Chu Fulei He Yongyong Guo Dan. New and Rapidly Converging BP Neural Network Algorithm[J]. Transactions of the Chinese Society for Agricultural Machinery, 2004, 35(6): 182-184 |
| |
| Authors: | Wang Yunsong Chu Fulei He Yongyong Guo Dan |
| |
| Affiliation: | Tsinghua University |
| |
| Abstract: | The feature of a slow converging of a standard BP neural network algorithm is the main cause of its limitation for a wide application. A new BP algorithm, Levenberg Marquardt back propagation algorithm, which could converge rapidly, was developed in this paper based on the least square method and the standard BP algorithm. The experimental result proved that LMBP converged very fast and had a good stability compared with the standard BP algorithm, which meat that LMBP algorithm was suitable for a big network with hundreds of weight values or high training precision demand. The superiority of the algorithm over others is notable if the computer memory is large enough. |
| |
| Keywords: | Neural network LMBP algorithm Levenberg-Marquardt optimization |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
