| 一种局部回归神经网络的在线学习算法 |
| |
| 引用本文： | 江小平,姚天任. 一种局部回归神经网络的在线学习算法[J]. 华中科技大学学报(自然科学版), 2005, 33(5): 1-3 |
| |
| 作者姓名： | 江小平 姚天任 |
| |
| 作者单位： | 华中科技大学,电子与信息工程系,湖北,武汉,430074;华中科技大学,电子与信息工程系,湖北,武汉,430074 |
| |
| 摘 要： | 针对目前局部回归神经网络误差函数在线计算复杂的缺陷，利用信号流图(SFG)基本理论，通过分析信号流图(SFG)和转置信号流图(ASFG)，将神经网络的误差导数的信号流图(SFG)和转置信号流图(ASFG)分别级联在原始信号流图(SFG)和转置信号流图(ASFG)上，构成单输出自回归神经网络．依据因果非线性时变系统流图计算仅仅与网络拓扑结构有关的理论，推导了一种与网络结构无关的在线后向BP学习算法，较好地解决了对任意结构的局部回归神经网络的在线学习问题．仿真结果表明了本算法的有效性．
|
| 关 键 词： | 局部回归神经网络 在线BP算法 信号流图 |
| 文章编号： | 1671-4512(2005)05-0001-03 |
| 修稿时间： | 2004-06-15 |
An on-line learning algorithm of neural networks of local recurrent |
| |
| Jiang Xiaoping,Yao Tianren. An on-line learning algorithm of neural networks of local recurrent[J]. JOURNAL OF HUAZHONG UNIVERSITY OF SCIENCE AND TECHNOLOGY.NATURE SCIENCE, 2005, 33(5): 1-3 |
| |
| Authors: | Jiang Xiaoping Yao Tianren |
| |
| Affiliation: | Jiang Xiaoping Yao TianrenJiang Xiaoping Doctoral Candidate, Dept of Electronics & Information Eng.,Huazhong Univ. of Sci. & Tech.,Wuhan 430074,China. |
| |
| Abstract: | Using the signal-flow-graph (SFG) basic theories and analyzing the SFG and adjoint signal-flow-graph (ASFG), we connected the SFG of the gradient of the cost function and its ASFG respectively to the original SFG and ASFG, adjoined the overall systems to a target SFG, constituting a kind of alone-oupput recurrent neural networks which depends only on the functional relationships between the branch variables. The simulations of non-linear dynamic systems identification also were presented to assess the performance of the algorithm in the application of local recurrent neural networks. |
| |
| Keywords: | local recurrent neural networks on-line BP learning signal-flow-graph |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
