| 一种实时求解矩阵平方根的新型递归神经网络 |
| |
| 引用本文： | 易称福,王华金,张小红. 一种实时求解矩阵平方根的新型递归神经网络[J]. 南方冶金学院学报, 2011, 0(3) |
| |
| 作者姓名： | 易称福 王华金 张小红 |
| |
| 作者单位： | 江西理工大学信息工程学院; |
| |
| 基金项目： | 江西省教育厅资助项目(GJJ11128,GJJ11470) |
| |
| 摘 要： | 不同于梯度法的设计思想,提出了一种实时求解矩阵平方根的新型递归神经网络模型.与传统梯度神经网络模型相比,新型模型是用隐性动力学方程表示的,而梯度神经网络模型的动力学方程是显性的.另外,新型神经网络模型定义的误差函数是无界的,而梯度神经网络定义的基于标量的非负能量函数.通过一个计算机的仿真实例(采用幂S型激励函数)表明,两种神经网络模型均能有效地求解任意给定的非奇异正定矩阵的平方根.
|
| 关 键 词： | 矩阵平方根 神经网络 梯度法 误差函数 能量函数 |
A Type of New Recurrent Neural Networks for Solving True Time Matrix Square Root |
| |
| YI Chen-fu,WANG Hua-jin,ZHANG Xiao-hong. A Type of New Recurrent Neural Networks for Solving True Time Matrix Square Root[J]. Journal of Southern Institute of Metallurgy, 2011, 0(3) |
| |
| Authors: | YI Chen-fu WANG Hua-jin ZHANG Xiao-hong |
| |
| Affiliation: | YI Chen-fu,WANG Hua-jin,ZHANG Xiao-hong(Faculty of Information Engineering,Jiangxi University of Science and Technology,Ganzhou 341000,China) |
| |
| Abstract: | Being different from the designed idea of gradient algorithm,a type of new recurrent neural networks for the solution of matrix square root in real-time is presented in the paper.Compariing with the conventional gradient-based neural networks(GNN),the new neural networks are expressed by implicit dynamic equations,while the equation of GNN is explicit.In addition,the new neural networks are defined by an indefinite error function,rather than a scalar-valued nonnegative energy function,which is usually assoc... |
| |
| Keywords: | matrix square root neural networks gradient algorithm error function energy function |
| 本文献已被 CNKI 等数据库收录！ |
|
