期刊文献+

基于核函数的PCA-L1算法 被引量:4

PCA-L1 Algorithm Based on Kernel Function
下载PDF
导出
摘要 主成分分析方法由于使用了L2范数,因此对异常值较敏感。针对该问题,提出一种基于核函数的L1范数主成分分析方法。运用核函数将原始数据映射到核空间中得到核矩阵,再利用L1范数使距离函数达到最小。实验结果表明,该算法具有旋转不变性,对异常值和非线性问题具有稳定性,且正确识别率较高。 Because of using L2 norm,Principal Component Analysis(PCA) method is sensitive to outliers.So this paper proposes a PCA method based on kernel function and L1 norm.It maps original data to kernel space to get a kernel matrix,and utilizes kernel function and L1 norm to minimize the distance function.Experimental result shows that the algorithm is invariant to rotations and robust to outliers and nonlinear problems,and it has higher correct recognition rate.
出处 《计算机工程》 CAS CSCD 北大核心 2011年第22期174-175,178,共3页 Computer Engineering
基金 国家自然科学基金资助项目(61003169)
关键词 PCA-L1算法 L1范数 核主成分分析 核函数 人脸识别 PCA-L1 algorithm L1 norm Kernel Principal Component Analysis(KPCA) kernel function face recognition
  • 相关文献

参考文献6

  • 1Jolliffe I T.Principal Component Analysis[M].New York,USA: Springer-Verlag,1986. 被引量:1
  • 2Baccini A,Besse P,Falguerolles A D.A L1-norm PCA and a Heuristic Approach[C]//Proc.of Conference on Ordinal and Symbolic Data Analysis.Berlin,Germany: Springer,1996. 被引量:1
  • 3Ke Q,Kanade T.Robust Subspace Computation Using L1 Norm[R].Carnegie Mellon University,Tech.Rep.: CMU-CS- 03-172,2003. 被引量:1
  • 4Kwak N.Principal Component Analysis Based on L1-norm Maximization[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2008,30(9): 1672-1680. 被引量:1
  • 5许亮,张小波.一种快速核特征提取方法及其应用[J].计算机工程,2009,35(24):26-28. 被引量:2
  • 6Newman D J,Hettich S,Blake C L,et al.UCI Repository of Machine Learning Databases[EB/OL].[2010-10-22].http://www.ics.uci.edu/-mlearn/MLRepository.html. 被引量:1

二级参考文献4

  • 1Vapnik V N. Statistical Learning theory[M]. New York, USA: John Wiley and Sons Inc., 1998. 被引量:1
  • 2Scholkopf B, Smola A, Mubller K R. Nonlinear Component Analysis as a Kernel Eigenvalue Problem[J]. Neural Computer, 1998, 10(6): 1299-1319. 被引量:1
  • 3Baudat G, Anouar F. Feaure Vector Selection and Projection Using Kernels[J]. Neurocomputing, 2003, 55(1): 21-38. 被引量:1
  • 4肖健华,吴今培.基于核的特征提取技术及应用研究[J].计算机工程,2002,28(10):36-38. 被引量:27

共引文献1

同被引文献36

  • 1胡永生,冀小平.基于阈值的图像分割方法的研究[J].科技情报开发与经济,2007,17(2):171-172. 被引量:9
  • 2张向东,李波.基于Gabor小波变换和PCA的人脸识别方法[J].电子科技,2007,20(4):72-74. 被引量:9
  • 3WRIGHT J, PENG Yi-gmg, MA Yi. Robust principal component analysis : exact recovery of corrupted low-rank matrices by convex opti- mization[ C ]//Advance in Neural Information Processing Systems. 2009 : 2080- 2088. 被引量:1
  • 4WU Xiao-jun,JOSEF K,YANG Jing-yu,et al. A new direct LDA (D- LDA) algorithm for feature extraction in flce recognition [ C ]//Proe of the 17th International Conference on Pattern Recognition. 2004:545- 548. 被引量:1
  • 5HYVARINEN A, OJA E. Independent component analysis: algo- rithms and applications [Jl. Noural Networks,2000,'13(4-5) :411- 430. 被引量:1
  • 6ZHAO Ying,ATKESON C G. Implementing projection pursuit learning [J]. IEEE Trans on Neural Networks,1996,7(2) :362-373. 被引量:1
  • 7NICOLAS G, FRANCOIS G. A multilevel approach Ior nonnegative matrix faetorization [ J ]. Journal of Computational and Applied Mathematics, 2012,236 ( 7 ) : 1708-1723. 被引量:1
  • 8BURGES C J C. Dimension reduction:a guided tour[ J ]. Founda- tions and Trends in Machine Learning,2009,2(4) :275-365. 被引量:1
  • 9AILON N, CHAZELLE B. Faster dimension reduction [ J ]. Communi- cations of the ACM,2010,53(2) :97-104. 被引量:1
  • 10KE Qi-fa,KANADE T. Robust L1 norm factorization in the psence of outliers and missing data by alternative convex programming[ C]// Proc of IEEE Computer Society Conference on Computer Vision and Pattern Recognition. 2004:739-746. 被引量:1

引证文献4

二级引证文献10

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部