| 一种前馈神经网络算法 |
| |
| 引用本文： | 王祖麟,王丽霞.一种前馈神经网络算法[J].科技广场,2004(8):51-53. |
| |
| 作者姓名： | 王祖麟 王丽霞 |
| |
| 作者单位： | 江西理工大学,机电工程学院,赣州,341000 |
| |
| 摘 要： | 神经网络由于其非线性处理能力强，性能稳定等特点得到了广泛应用和研究，主要应用于模式识别、信号处理、知识工程、专家系统、优化组合、机器人控制等。神经网络中使用最为广泛的就是前馈神经网络，其网络权值学习算法中影响最大的就是误差反向传播算法(back-propagation简称BP算法)。BP算法存在局部极小点，收敛速度慢等缺点，基于优化理论的Levenberg-Marquardt算法忽略了二阶项，该文讨论当误差不为零或者不为线性函数即二阶项S(W)不能忽略时的Hesse矩阵的近似计算，进而训练网络。
|
| 关 键 词： | 前馈神经网络算法 网络权值学习算法 误差反向传播算法 Hesse矩阵 收敛速度 |
A Modified Feed-Forword Neural Network LM Algorithm |
| |
| Wang Zhuling Wang lixia.A Modified Feed-Forword Neural Network LM Algorithm[J].Science Mosaic,2004(8):51-53. |
| |
| Authors: | Wang Zhuling Wang lixia |
| |
| Abstract: | Neural network is widely studied for its nonlinear disposal capacity and good stability. Pattern Recognition, Signal Processing, Knowledge Engineering, Expert System, Combinatorial Optimum, and Robot Control are all concerning about neural network design. For feed-forward neural network ,BP algorithm is among the most important neural network algorithms. BP algorithm has its local minima and its slow training speed. Second order item is omitted in Levenberg-Marquardt algorithm based on optimal theory. This paper presents an approximation calculation for Hesse matrix to train networks. |
| |
| Keywords: | neural network back-propagation algorithm Hesse matrix |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
