| 样本重叠映射的柔合算法 |
| |
| 引用本文： | 徐静波 周美华. 样本重叠映射的柔合算法[J]. 上海工程技术大学学报, 1997, 11(4): 17-20 |
| |
| 作者姓名： | 徐静波 周美华 |
| |
| 作者单位： | 上海工程技术大学计算机及电子电气工程系!上海200335(徐静波)，中国纺织大学!上海200051(周美华) |
| |
| 摘 要： | 在小脑模型关节控制器CMAC的ALbus算法中，可由批量学习的输入产生各个权在更新时的柔合因数。这样能改善样本逐一学习法的遗忘现象，又能有较快的收敛速度，对非线性函数的仿真结果，表明了改进算法的特点。
|
| 关 键 词： | 神经网络 小脑模型 泛化 收敛性 hash编码 |
A FLEXIBLE ALGORITHM FOR SAMPLES'' OVERLAPPING MAP |
| |
| Xu Jingbo, Zhou Meihua. A FLEXIBLE ALGORITHM FOR SAMPLES'' OVERLAPPING MAP[J]. Journal of Shanghai University of Engineering Science, 1997, 11(4): 17-20 |
| |
| Authors: | Xu Jingbo Zhou Meihua |
| |
| Affiliation: | Xu Jingbo; Zhou Meihua |
| |
| Abstract: | Based on ALbus algorithm, the flaxible factor for each modified weight can be obtained from batch input matrix during CMAC learning process. The forgetting phenomenon of the one by one learning method can be diminished while the speed of convergent can be increased. Simulation result for a nonlinear function shows the advantage of this improved method. |
| |
| Keywords: | Neural networks Cerebellar model articulation controller(CMAC)Generalization Convergent Hash coding |
| 本文献已被 CNKI 维普 等数据库收录！ |
