一种潜在信息约束的非负矩阵分解方法  

Potential Information Restrained Nonnegative Matrix Factorization

作　　者：高新波[1] 王笛[1] 王秀美[1] 

机构地区：[1]西安电子科技大学电子工程学院,西安710071

出　　处：《数据采集与处理》2014年第1期11-18,共8页Journal of Data Acquisition and Processing

基　　金：国家杰出青年科学基金(61125204)资助项目;国家自然科学基金(61172146;61100158)资助项目;陕西省重点科技创新团队(2012KCT-02)资助项目

摘　　要：传统的非负矩阵分解方法没有充分利用数据间的内在相似性,从而影响了算法的性能。为此,本文提出一种潜在信息约束的非负矩阵分解方法。该方法首先利用迭代最近邻方法挖掘原始数据的潜在信息,然后利用潜在信息构造数据之间的相似图,最后将相似图作为约束项求得非负矩阵的最优分解。相似图的约束使得非负矩阵分解在降维过程中保持了原始数据之间的相似性关系,进而提高了非负矩阵分解的判别能力。图像聚类实验结果表明了该方法的有效性。Traditional NMF method does not fully utilize the internal similarity among original data, thus the performance of dimensionality reduction is limited. To this end, a new nonnega- tive matrix factorization algorithm restrained by the regularization of potential information is proposed. Firstly, the potential information is mined via the iterative nearest neighbor. Then the potential information is utilized to construct similarity graph of data set. Finally, the simi- larity graph is incorporated as a regularization term to preserve the relationship between origi- nal data in the decomposition process of nonnegative matrix. The regularization term keeps the similarity between the original data in the process of dimensionality reduction, which can im- prove the discriminant ability of nonnegative matrix factorization algorithm. Thorough experi- ments on standard image databases show the superior performance of the proposed method.

关 键 词：数据降维 非负矩阵分解 潜在信息 相似图 迭代最近邻 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]