| 一种递归神经网络的快速并行算法 |
| |
| 引用本文： | 李鸿儒,顾树生.一种递归神经网络的快速并行算法[J].自动化学报,2004,30(4):516-522. |
| |
| 作者姓名： | 李鸿儒 顾树生 |
| |
| 作者单位： | 1.东北大学信息科学与工程学院,沈阳 |
| |
| 基金项目： | 国家“十五”重点攻关计划项目(2001BA401A06-0.4)资助~~ |
| |
| 摘 要： | 针对递归神经网络BP(Back Propagation)学习算法收敛慢的缺陷,提出一种新的递归
神经网络快速并行学习算法.首先,引入递推预报误差(RPE)学习算法,并且证明了其稳定性；
进一步地,为了克服RPE算法集中运算的不足,设计完整的并行结构算法.本算法将计算分配
到神经网络中的每个神经元,完全符合神经网络的并行结构特点,也利于硬件实现.仿真结果表
明,该算法比传统的递归BP学习算法具有更好的收敛性能.理论分析和仿真实验证明,该算法
与RPE集中运算算法相比可以大大节省计算时间.
|
| 关 键 词： | 递归神经网络 递推预报误差 并行算法 集中运算 |
| 收稿时间： | 2002-10-8 |
| 修稿时间： | 2002年10月8日 |
A Fast Parallel Algorithm for a Recurrent Neural Network |
| |
| LI Hong-Ru,GU Shu-Sheng.A Fast Parallel Algorithm for a Recurrent Neural Network[J].Acta Automatica Sinica,2004,30(4):516-522. |
| |
| Authors: | LI Hong-Ru GU Shu-Sheng |
| |
| Affiliation: | 1.School of Information Science and Engineering,Northeastern University,Shenyang |
| |
| Abstract: | Because the recurrent BP algorithm suffers from the drawback of slow con-vergence, a new fast parallel learning algorithm for recurrent neural networks is pro-posed. First, the recursive predictive error(RPE) learning algorithm for recurrent neural networks is introduced, and its stability is demonstrated. Furthermore, in order to overcome the disadvantage of the centralized computing of RPE learning algorithm, a parallel structure algorithm is derived. In the new parallel learning algorithm, the com-putation is distributed to each neuron in the network, which is coherent with the mas-sively parallel nature of the network, and also convenient for hardware implementation. Simulation results show that better convergence performance of the proposed algorithm over the traditional recurrent BP algorithm. Meanwhile, theoretical analysis and simu-lation results show that the new parallel algorithm also cut lots of computational time compared with the RPE centralized computing algorithm. |
| |
| Keywords: | Recurrent neural networks recursive predictive error parallel algorithm centralized computing |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
| 点击此处可从《自动化学报》浏览原始摘要信息 |
|
点击此处可从《自动化学报》下载全文 |
|
