一种新的基于最大边缘准则的监督流形学习方法

计算机科学 ›› 2014, Vol. 41 ›› Issue (4): 273-279.

一种新的基于最大边缘准则的监督流形学习方法

袁暋,杨瑞国,原媛,雷迎科

合肥学院网络与智能信息处理重点实验室合肥230022;电子工程学院合肥 230037;电子工程学院合肥 230037;电子工程学院合肥 230037

出版日期:2018-11-14 发布日期:2018-11-14
基金资助:
本文受国家自然科学基金(61272333,61273302,61005010),安徽省自然科学基金(1208085MF94,1208085MF98,1308085MF84)资助

New Supervised Manifold Learning Method Based on MMC

YUAN Min,YANG Rui-guo,YUAN Yuan and LEI Ying-ke

Online:2018-11-14 Published:2018-11-14

摘要/Abstract

摘要： 在深入研究局部样条嵌入算法(LSE)的基础上,引入明确的线性映射关系,构建平移缩放模型和正交化特征子空间,提出了一种正交局部样条判别投影算法(O-LSDP),有效解决了原始LSE算法存在的两个主要问题:样本外点学习问题和无监督模式学习问题。该算法能够应用于模式分类问题并显著改善算法的分类识别能力。在标准人脸数据库上进行的实验比较分析验证了该算法的有效性与可行性。

关键词: 特征提取,子空间学习,局部样条嵌入,最大边缘准则,流形学习

Abstract: Based on the analysis of local spline embedding (LSE) method,we proposed an efficient feature extraction algorithm called orthogonal local spline discriminant projection (O-LSDP)．By introducing an explicit linear mapping,constructing different translation and rescaling models for different classes as well as orthogonalizing feature subspace,O-LSDP can effectively circumvent the two major shortcomings of the original LSE algorithm,i.e.,out-of-sample and unsupervised learning．O-LSDP not only inherits the advantages of LSE which uses local tangent space as a representation of the local geometry so as to preserve the local structure,but also makes full use of class information and orthogonal subspace to significantly improve discriminant power．Extensive experiments on standard face databases and plant leaf data set verify the feasibility and effectiveness of the proposed algorithm.

Key words: Feature extraction,Subspace learning,Local spline embedding,Maximum margin criterion,Manifold learning

袁暋,杨瑞国,原媛,雷迎科. 一种新的基于最大边缘准则的监督流形学习方法[J]. 计算机科学, 2014, 41(4): 273-279. https://doi.org/

YUAN Min,YANG Rui-guo,YUAN Yuan and LEI Ying-ke. New Supervised Manifold Learning Method Based on MMC[J]. Computer Science, 2014, 41(4): 273-279. https://doi.org/

参考文献

[1] Tenenbaum J B,de Silva V,Langford J C．A global geometric framework for nonlinear dimensionality reduction [J]．Science,2000,290(5500):2319-2323
[2] Roweis S T,Saul L K．Nonlinear dimensionality reduction by locally linear embedding [J]．Science,2000,290(5500):2323-2326
[3] Saul L K,Roweis S T．Think globally,fit locally:unsupervised learning of low dimensional manifolds [J]．The Journal of Machine Learning Research,2003,4:119-155
[4] Belkin M,Niyogi P．Laplacian eigenmaps and spectral techniques for embedding and clustering [C]∥Proceedings of Advances in Neural Information Processing Systems．2002,14:585-591
[5] Belkin M,Niyogi P．Laplacian eigenmaps for dimensionality re-duction and data representation [J]．Neural Computation,2003,15(6):1373-1396
[6] Donoho D L,Grimes C．Hessian eigenmaps:Locally linear embedding techniques for high-dimensional data [C]∥Proceedings of the National Academy of Sciences of the United States of America．2003:5591-5596
[7] Weinberger K Q,Saul L K．Unsupervised learning of imagemanifolds by semidefinite programming [C]∥Proceedings of CVPR-04．2004:988-995
[8] Weinberger K Q,Saul L K．An introduction to nonlinear dimensionality reduction by maximum variance unfolding [C]∥Proceedings of AAA06．2006:1683-1686
[9] Zhang Z,Zha H．Principal manifolds and nonlinear dimension reduction via local tangent space alignment [J]．SIAM J．Scientific Computing,2005,26(1):313-338
[10] Lin T,Zha H．Riemannian manifold learning [J]．IEEE Transactions on Pattern Analysis and Machine Intelligence,2007,30(5):796-809
[11] Lin T,Zha H,Lee S．Riemannian manifold learning for nonlinear dimensionality reduction [C]∥Proceedings of ECCV2006．2006:44-55
[12] Xiang S,Nie F,Zhang C．Spline embedding for nonlinear dimensionality reduction．Machine Learning [C]∥Proceedings of ECML2006．2006:825-832
[13] Xiang S,Nie F,Zhang C．Nonlinear dimensionality reductionwith local spline embedding [J]．IEEE Transactions on Know-ledge and Data Engineering,2008,21(9):1285-1298
[14] Yan S,Xu D,Zhang B,et al．Graph embedding and extensions:A general framework for dimensionality reduction [J]．IEEE Transactions on Pattern Analysis and Machine Intelligence,2007,5(1):40-51
[15] He X,Yan S,Hu Y,et al．Face recognition using laplacianfaces [J]．IEEE Transactions on Pattern Analysis and Machine Intelligence,2005,27(3):328-340
[16] De Ridder D,Kouropteva O,Okun O,et al．Supervised locally linear embedding [C]∥Proceedings of the 2003Joint International Conference on Artificial Neural Networks and Neural Information Processing．2003:333-341
[17] Pan Y,Ge S S,Mamun A．Weighted locally linear embedding for dimension reduction [J]．Pattern Recognition,2009,42(5):798-811
[18] Li H,Jiang T,Zhang K．Efficient and robust feature extraction by maximum margin criterion [J]．IEEE Transactions on Neural Networks,2006,17(1-3):157-165
[19] Duchon J．Splines minimizing rotation-invariant semi-norms inSobolev spaces [C]∥Constructive theory of functions of several variables．1977:85-100
[20] Meinguet J．Multivariate interpolation at arbitrary points made simple [J]．Journal of Applied Mathematics and Physics,1979,30:292-304
[21] Wahba G．Spline models for observational data [M]．SIAMPress,1990
[22] Sderkvist O．Computer vision classification of leaves fromSwedish trees [D]．Master’s Thesis,Linkoping University,2001
[23] Ye J．Characterization of a family of algorithms for generalized discriminant analysis on undersampled problems [J]．Journal of Machine Learning Research,2006,6:483-502
[24] Belhumeur P N,Hespanha J P,Kriegman D J．Eigenfaces vs．fisherfaces:Recognition using class specific linear projection [J]．IEEE Transactions on Pattern Analysis and Machine Intelligence,1997,19(7):711-720
[25] Cai D,He X,Han J．Using graph model for face analysis [R]．Univ．Illinois Urbana-Champaign,Urbana,IL,Department of Computer Science,Technical Report,2005
[26] Zhang T,Yang J,Zhao D,et al．Linear local tangent space alignment and application to face recognition [J]．Neurocomputing,2007,70(7-9):1547-1553
[27] De Silva V,Tenenbaum J B．Global versus local methods in nonlinear dimensionality reduction [C]∥Proceedings of Advances in Neural Information Processing Systems．2003,15:721-728

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed