| 一种基于并行策略的BP改进算法 |
| |
| 作者单位： | 山东师范大学信息科学与工程学院 |
| |
| 摘 要： | 介绍了BP神经网络的基本结构及原理,分析了其收敛慢的原因。为加快其收敛速度,结合带动量梯度下降法提出一种新的算法(PBBP),用多个学习速率不同但结构相同的网络进行并行训练,在每次迭代后都根据误差找出处于最佳状态的网络,并使其它网络的训练参数作适当变化再进行下一次迭代,直到整个网络的误差减小到允许范围内或达到训练次数要求,加快了其收敛速度,能够很好地脱离平坦区。通过在Matlab里编程进行仿真实验证明,该算法是可行的。
|
| 关 键 词： | 神经网络 BP算法 并行 |
An Improved BP Algorithm Based on Parallel |
| |
| GUO Yan-wei,WANG Hong-guo,WANG Xin,YU Hui. An Improved BP Algorithm Based on Parallel[J]. Microcomputer Development, 2008, 0(10) |
| |
| Authors: | GUO Yan-wei WANG Hong-guo WANG Xin YU Hui |
| |
| Abstract: | Analysed why the momentum BP algorithm is slow in convergence,describes a new method for the improvement of BP algorithm,named PBBP(parallel based BP algorithm).A method using many networks which are different in learning speed to train by parallel.Then speed up learning process and get out of flat area.The simulation experiments by programing in Matlab show that the improved algorithm is feasible. |
| |
| Keywords: | neural network BP algorithm parallel |
| 本文献已被 CNKI 等数据库收录！ |
|
