| 基于模糊最大散度差判别准则的聚类方法 |
| |
| 引用本文： | 皋军,王士同.基于模糊最大散度差判别准则的聚类方法[J].软件学报,2009,20(11):2939-2949. |
| |
| 作者姓名： | 皋军 王士同 |
| |
| 作者单位： | 1. 江南大学,信息工程学院,江苏,无锡,214122;盐城工学院,信息工程学院,江苏,盐城,224001;浙江大学,CAD&CG国家重点实验室,浙江,杭州,310027
2. 江南大学,信息工程学院,江苏,无锡,214122;浙江大学,CAD&CG国家重点实验室,浙江,杭州,310027 |
| |
| 基金项目： | Supported by the National Natural Science Foundation of China under Grant Nos.60773206, 60903100, 90820002(国家自然科学基金); the National High-Tech Research and Development Plan of China under Grant Nos.2007AA1Z158, 2006AA10Z313 (国家高技术研究发展计划(863)); the National Defense Research Foundation of China under Grant No.A1420461266 (国防应用基础研究基金); the Jiangsu Provincial Innovation Project of Graduate Students of China under Grant No.CX09B-175Z (江苏省普通高校研究生科研创新计划); the Open Project Program of the State Key Laboratory of CAD&CG, Zhejiang University of China under Grant No.A0802 (浙江大学CAD&CG国家重点实验室开放课题) |
| |
| 摘 要： | 基于最大散度差判别准则提出了一种模糊最大散度差准则,并根据模糊最大散度差准则提出一种聚类方法(fuzzy maximum scatter difference discriminant criterion based clustering algorithm,简称FMSDC).该方法通过迭代优化方法实现聚类的同时还可以实现特征降维.该方法首先在最大散度差判别准则中引入模糊概念;然后通过具体原则设定模糊最大散度差判别准则中的参数η,从而在一定程度上降低了由参数η引起的敏感性;最后分别根据模糊隶属度μik、最优鉴别矢量ω进行聚类和特征降维.实验结果表明,FMSDC方法不但具有基本的聚类功能,而且具有较好的鲁棒性和较强的特征降维能力.
|
| 关 键 词： | 模糊最大散度差判别准则 鉴别矢量 降维 模糊聚类 鲁棒性 |
| 收稿时间： | 3/3/2008 12:00:00 AM |
| 修稿时间： | 7/9/2008 12:00:00 AM |
Fuzzy Maximum Scatter Difference Discriminant Criterion Based Clustering Algorithm |
| |
| GAO Jun and WANG Shi-Tong.Fuzzy Maximum Scatter Difference Discriminant Criterion Based Clustering Algorithm[J].Journal of Software,2009,20(11):2939-2949. |
| |
| Authors: | GAO Jun and WANG Shi-Tong |
| |
| Affiliation: | GAO Jun1,2,3 ,WANG Shi-Tong1,3 1(School of Information Engineering,Jiangnan University,Wuxi 214122,China) 2(School of Information Engineering,Yancheng Institute of Technology,Yancheng 224001,China) 3(State Key Laboratory of CAD&CG,Zhejiang University,Hangzhou 310027,China) |
| |
| Abstract: | In this paper, a fuzzy scatter difference discrimininant criterion is presented. Based on this criterion, fuzzy clustering algorithm FMSDC (fuzzy maximum scatter difference discriminant criterion based clustering algorithm) is also presented. The proposed algorithm reduces dimensionality while clustering by iterative optimizing procedure. First, it introduces the fuzzy concept into maximum scatter difference discriminant criterion; then the parameter Η in the fuzzy criterion is appropriately determined based on specific principles so that the sensibility aroused by parameter η can be decreased to some extent; At last clustering and reducing dimensionality are realized according to fuzzy membership μ_(ik) and optional discriminant vector ω, respectively. Experimental results demonstrate the proposed method FMSDC is not only capable of clustering but also robust and capable of reducing dimensionality. |
| |
| Keywords: | fuzzy maximum scatter difference discriminant criterion discriminant vector dimensionality reduction fuzzy clustering robust |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
| 点击此处可从《软件学报》浏览原始摘要信息 |
|
点击此处可从《软件学报》下载全文 |
|
