| 一种快速有效的神经网络新算法 |
| |
| 引用本文： | 程正群,李歧强,钱积新.一种快速有效的神经网络新算法[J].浙江大学学报(自然科学版 ),2000,34(3):338-340. |
| |
| 作者姓名： | 程正群 李歧强 钱积新 |
| |
| 作者单位： | [1]浙江大学工业控制技术研究所 [2]山东工业大学自动化系 |
| |
| 摘 要： | 基于输出层函数为线性函数的层前馈神经网络，结合自适应步长和动量解耦的伪牛顿算法及迭代最小二乘法导出了一种混合算法。伪真证明该混合算法自适应能力强，计算量小，收敛速度快，是一种有效的工程实用算法。
|
| 关 键 词： | 前馈神经网络 解耦 BP算法 伪牛顿算法 混合算法 |
| 修稿时间： | 1997-12-12 |
A fast novel algorithm for neural networks |
| |
| CHENG Zheng-qun, LI Qi-qiang, QIAN Ji-xin.A fast novel algorithm for neural networks[J].Journal of Zhejiang University(Engineering Science),2000,34(3):338-340. |
| |
| Authors: | CHENG Zheng-qun LI Qi-qiang QIAN Ji-xin |
| |
| Abstract: | On the basis of both adaptive BP algorithm and Newton's method, Quasi Newton algorithm with adaptive decoupled step and momentum (QNADSM) for feed-forward neural networks is derived. Based on three-layered feed-forward neural networks with linear output, a hybrid algorithm is proposed. The simulation results demonstrated that the hybrid algorithm is an effectively realistic algorithm. |
| |
| Keywords: | feed-forward neural networks de-couple BP algorithms Quasi Newton algorithms iterative least square methods |
| 本文献已被 CNKI 维普 等数据库收录！ |
