| 信用评级的简约神经网络算法 |
| |
| 引用本文： | 赵禹骅,李栋龙,李可柏.信用评级的简约神经网络算法[J].计算机工程与应用,2006,42(23):201-203,224. |
| |
| 作者姓名： | 赵禹骅 李栋龙 李可柏 |
| |
| 作者单位： | 同济大学环境科学与工程学院,上海,200092;广西工学院信息与计算科学系,广西,柳州,545006;东南大学经济管理学院,南京,210096 |
| |
| 摘 要： | 运用经济资源的“边际效用递减”原理,分析了信用评级知识的非线性特点,着眼于神经网络算法的结构、函数和收敛算法三部件逻辑独立性,分析了经典神经网络算法拓扑结构的复杂性引致的算法参数调整过度复杂问题,提出了简约神经网络的拓扑结构,证明了在全部结点函数线性且全部隐层结点函数过原点的条件下经典神经网络与简约神经网络具有等价性,设计了基于简约网络的算法,算法结果获得了较高的拟合精度。
|
| 关 键 词： | 经典神经网络 简约神经网络 非线性 信用等级 |
| 文章编号： | 1002-8331-（2006）23-0201-03 |
| 收稿时间： | 2006-01 |
| 修稿时间： | 2006-01 |
The Contracted Neural Network Algorithm of Credit Rating Evaluation |
| |
| Zhao Yuhua,Li Donglong,Li Kebai.The Contracted Neural Network Algorithm of Credit Rating Evaluation[J].Computer Engineering and Applications,2006,42(23):201-203,224. |
| |
| Authors: | Zhao Yuhua Li Donglong Li Kebai |
| |
| Affiliation: | 1 School of Environmental Science and Engineering,Tongji University,Shanghai 200092; 2 Department of Information and Computer Science,Guangxi University of Technology,Liuzhou,Guangxi 545006; 3 School of Economics and Management,Southeast University,Nanjing 210096 |
| |
| Abstract: | In this paper,the nonlinear of credit rating evaluation is analyzed,by the law of diminishing marginal utility.From the view of the relation on the structure,function and convergent algorithm in neural network algorithm,it is analyzed that the complexity of adjusting its parameters from the complexity of the structure of classical neural network.Then,the structure of constructed neural network is advanced,the equipollence of two kinds of neural network is proved if all crunodes functions are linear and all inner crunodes functions pass origin,and the constructed neural network algorithm is designed.The case indicates the contracted neural network algorithm gets to a preferable precision. |
| |
| Keywords: | classical neural network contracted neural network nonlinear credit rating |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
