|
|||||
|
|
| 可分解马尔可夫网络参数学习中的噪声平滑 | |
| 引用本文： | 王双成,李小琳,刘风霞.可分解马尔可夫网络参数学习中的噪声平滑[J].控制与决策,2009,24(4). |
| 作者姓名： | 王双成 李小琳 刘风霞 |
| 作者单位： | 1. 上海立信会计学院信息科学系,南京,210093 2. 南京大学,商学院,南京,210093 3. 上海立信会计学院工商管理系,上海,201600 |
| 基金项目： | 国家自然科学基金，上海市教委重点学科和卜海市教委科研创新重点项目 |
| 摘 要： | 对于存在噪声的可分解马尔可夫网络参数学习.目前可通过调整可分解马尔可夫网络的结构来增强其抗噪声能力,但调整后的结构往往会降低网络的有效性和可靠性,而且不能从源头上减少噪声对参数的影响.针对该问题,首先建立基于可分解马尔可夫网络理论的邻域最优预测;然后以变量为基本单位,使用预测值平滑噪声.实验结果显示,使用平滑后的数据进行参数学习,其可靠性能够得到显著的改进. |
| 关 键 词： | 可分解马尔可夫网络 参数学习 邻域预测 噪声平滑 |
Noise smoothing in learning the parameters of decomposable Markov networks |
|
| WANG Shuang-cheng,LI Xiao-lin,LIU Feng-xia.Noise smoothing in learning the parameters of decomposable Markov networks[J].Control and Decision,2009,24(4). | |
| Authors: | WANG Shuang-cheng LI Xiao-lin LIU Feng-xia |
| Affiliation: | 1a.Department of Information Science;1b.Department of Business Administration;Shanghai Lixin University of Commerce;Shanghai 201600;China;2.School of Business;Nanjing University;Nanjing 210093;China. |
| Abstract: | At present, the influence of noise can be reduced by simplifying the structure in the parameter learning of decomposable Markov network. But regulated structure will decrease the network reliability and the influence of noise can not be decreased or removed from root. In this paper, an effective method of smoothing noise is presented. Firstly, optimal neighborhood prediction is built based on the theory of decomposable Markov network. Then, a variable is taken as a basic unit and the noise can be smoothed b... |
| Keywords: | Decomposable Markov network Parameter learning Neighborhood prediction Noise smoothing |
| 本文献已被 CNKI 万方数据 等数据库收录！ | |