广义的监督局部保留投影算法

王晓明; 王士同

doi:10.3724/SP.J.1146.2008.00946

广义的监督局部保留投影算法

doi: 10.3724/SP.J.1146.2008.00946

基金项目:

国家863计划项目(2007AA1Z158，2006AA10Z313)和国家自然科学基金(60704047)资助课题

计量
- 文章访问数: 3606
- HTML全文浏览量: 95
- PDF下载量: 1201
- 被引次数: 0
出版历程
- 收稿日期: 2008-07-24
- 修回日期: 2009-03-09
- 刊出日期: 2009-08-19

Generalized Supervised Locality Preserving Projection

摘要

摘要: 针对监督的局部保留投影算法(Supervised Locality Preserving Projection，SLPP)在小样本情况下矩阵的奇异性问题，该文提出了一种广义的监督局部保留投影算法(Generalized Supervised Locality Preserving Projection，GSLPP)。GSLPP在大样本情况下等价于SLPP，在小样本情况下却可以等价转换到一个低维空间中来求解，从而有效解决了小样本问题。最后，实验结果验证了该方法的有效性。
- 特征提取;局部保留投影;监督局部保留投影
Abstract: Supervised Locality Preserving Projection (SLPP) is a generalization of Locality Preserving Projection (LPP) in the case of supervised learning. In this paper the drawback of SLPP in the high-dimensional and small sample size case is pointed out, and a new algorithm called Generalized Supervised Locality Preserving Projection (GSLPP) is proposed. The relationship between SLPP and GSLPP is theoretically analyzed. In the small sample size case GSLPP can be solved equivalently in lower-dimensionality space. Finally, the effectiveness of the proposed algorithm is verified by experimental results.

HTML全文

参考文献(1)

宋枫溪, 高秀梅, 刘树海等. 统计模式识别中的维数削减与低损降维[J]. 计算机学报, 2005, 28(11): 1915-1922.Song Feng-xi, Gao Xiu-mei, and Liu Shu-hai, et al..Dimensionality reduction in statistical pattern recognitionand low loss dimensionality reduction [J]. Chinese Journal ofComputers, 2005, 28(11): 1915-1922.[2]He X and Niyogi P. Locality preserving projections [C]. Proc.Conf. Advances in Neural Information Processing Systems,Vancouver, Canada, 2003: 585-591.[3]Kokiopoulou E and Saad Y. Orthogonal neighborhoodpreserving projections: A projection-based dimensionalityreduction technique [J].IEEE Transactions on PatternAnalysis and Machine Intelligence.2007, 29(12):2143-2156[4]申中华, 潘永惠, 王士同. 有监督的局部保留投影降维算法[J].模式识别与人工智能, 2008, 21(2): 233-239.Shen Zhong-hua, Pan Yong-hui, and Wang Shi-tong. Asupervised locality preserving projection algorithm fordimensionality reduction [J]. Pattern Recognition andArtificial Intelligence, 2008, 21(2): 233-239.[5]Tao Q, Wu G W, and Wang J. The theoretical analysis ofFDA and applications [J].Pattern Recognition.2006, 39(6):1199-1204[6]杨键, 杨静宇, 叶晖等. Fisher线性鉴别分析的理论研究及其应用[J]. 自动化学报, 2003, 29(4): 482-493.Yang Jian, Yang Jing-yu, Ye Hui, et al.. Theory of fisherlinear discriminant analysis and its application [J]. ActaAutomatic Sinica, 2003, 29(4): 482-493.[7]Liu J, Chen S C, and Tan X Y. A study on three lineardiscriminant analysis based methods in small sample sizeproblem[J].Pattern Recognition.2008, 41(1):102-116[8]Zhuang X S and Dai D Q. Improved discriminate analysis forhigh-dimensional data and its application to face recognition[J].Pattern Recognition.2007, 40(5):1570-1578[9]Cai D, He X, and Han J, et al.. Orthogonal laplacianfaces forface recognition [J].IEEE Transactions on Image Processing.2006, 15(11):3608-3614[10]Turk M and Pentland A. Eigenfaces for recognition [J].Journal of Cognitive Neuroscience.1991, 3(1):71-86[11]Masashi Sugiyama. Dimensionality reduction of multimodallabeled data by local Fisher discriminant analysis [J]. Journalof Machine Learning Research, 2007, 8(5): 1027-1061.[12]Jiang L X, Cai Z H, and Wang D H, et al.. Survey ofimproving k-nearest-neighbor for classification [C]. FourthInternational Conference on Fuzzy Systems and KnowledgeDiscovery (FSKD), Haikou, China, Aug 24-27, 2007: 679-683.

施引文献

资源附件(0)

访问统计