File Download
 
Links for fulltext
(May Require Subscription)
 
Supplementary

Article: Separable linear discriminant analysis
  • Basic View
  • Metadata View
  • XML View
TitleSeparable linear discriminant analysis
 
AuthorsZhao, J2
Yu, PLH1
Shi, L2
Li, S2
 
KeywordsFace Recognition
Linear Discriminant Analysis
Separable
Two-Dimensional Data
 
Issue Date2012
 
PublisherElsevier BV. The Journal's web site is located at http://www.elsevier.com/locate/csda
 
CitationComputational Statistics And Data Analysis, 2012, v. 56 n. 12, p. 4290-4300 [How to Cite?]
DOI: http://dx.doi.org/10.1016/j.csda.2012.04.003
 
AbstractLinear discriminant analysis (LDA) is a popular technique for supervised dimension reduction. Due to the curse of dimensionality usually suffered by LDA when applied to 2D data, several two-dimensional LDA (2DLDA) methods have been proposed in recent years. Among which, the Y2DLDA method, introduced by Ye et al. (2005), is an important development. The idea is to utilize the underlying 2D data structure to seek for an optimal bilinear transformation. However, it is found that the proposed algorithm does not guarantee convergence. In this paper, we show that the utilization of a bilinear transformation for 2D data is equivalent to modeling the covariance matrix of 2D data as separable covariance matrix. Based on this result, we propose a novel 2DLDA method called separable LDA (SLDA). The main contributions of SLDA include (1) it provides interesting theoretical relationships between LDA and some 2DLDA methods; (2) SLDA provides a building block for mixture extension; (3) unlike Y2DLDA, a neatly analytical solution can be obtained as that in LDA. Empirical results show that our proposed SLDA achieves better recognition performance than Y2DLDA while being computationally much more efficient. © 2012 Elsevier B.V. All rights reserved.
 
ISSN0167-9473
2013 Impact Factor: 1.151
2013 SCImago Journal Rankings: 1.399
 
DOIhttp://dx.doi.org/10.1016/j.csda.2012.04.003
 
ReferencesReferences in Scopus
 
DC FieldValue
dc.contributor.authorZhao, J
 
dc.contributor.authorYu, PLH
 
dc.contributor.authorShi, L
 
dc.contributor.authorLi, S
 
dc.date.accessioned2012-10-30T06:22:50Z
 
dc.date.available2012-10-30T06:22:50Z
 
dc.date.issued2012
 
dc.description.abstractLinear discriminant analysis (LDA) is a popular technique for supervised dimension reduction. Due to the curse of dimensionality usually suffered by LDA when applied to 2D data, several two-dimensional LDA (2DLDA) methods have been proposed in recent years. Among which, the Y2DLDA method, introduced by Ye et al. (2005), is an important development. The idea is to utilize the underlying 2D data structure to seek for an optimal bilinear transformation. However, it is found that the proposed algorithm does not guarantee convergence. In this paper, we show that the utilization of a bilinear transformation for 2D data is equivalent to modeling the covariance matrix of 2D data as separable covariance matrix. Based on this result, we propose a novel 2DLDA method called separable LDA (SLDA). The main contributions of SLDA include (1) it provides interesting theoretical relationships between LDA and some 2DLDA methods; (2) SLDA provides a building block for mixture extension; (3) unlike Y2DLDA, a neatly analytical solution can be obtained as that in LDA. Empirical results show that our proposed SLDA achieves better recognition performance than Y2DLDA while being computationally much more efficient. © 2012 Elsevier B.V. All rights reserved.
 
dc.description.natureLink_to_subscribed_fulltext
 
dc.identifier.citationComputational Statistics And Data Analysis, 2012, v. 56 n. 12, p. 4290-4300 [How to Cite?]
DOI: http://dx.doi.org/10.1016/j.csda.2012.04.003
 
dc.identifier.doihttp://dx.doi.org/10.1016/j.csda.2012.04.003
 
dc.identifier.epage4300
 
dc.identifier.hkuros210593
 
dc.identifier.issn0167-9473
2013 Impact Factor: 1.151
2013 SCImago Journal Rankings: 1.399
 
dc.identifier.issue12
 
dc.identifier.scopuseid_2-s2.0-84864148086
 
dc.identifier.spage4290
 
dc.identifier.urihttp://hdl.handle.net/10722/172504
 
dc.identifier.volume56
 
dc.languageeng
 
dc.publisherElsevier BV. The Journal's web site is located at http://www.elsevier.com/locate/csda
 
dc.publisher.placeNetherlands
 
dc.relation.ispartofComputational Statistics and Data Analysis
 
dc.relation.referencesReferences in Scopus
 
dc.subjectFace Recognition
 
dc.subjectLinear Discriminant Analysis
 
dc.subjectSeparable
 
dc.subjectTwo-Dimensional Data
 
dc.titleSeparable linear discriminant analysis
 
dc.typeArticle
 
<?xml encoding="utf-8" version="1.0"?>
<item><contributor.author>Zhao, J</contributor.author>
<contributor.author>Yu, PLH</contributor.author>
<contributor.author>Shi, L</contributor.author>
<contributor.author>Li, S</contributor.author>
<date.accessioned>2012-10-30T06:22:50Z</date.accessioned>
<date.available>2012-10-30T06:22:50Z</date.available>
<date.issued>2012</date.issued>
<identifier.citation>Computational Statistics And Data Analysis, 2012, v. 56 n. 12, p. 4290-4300</identifier.citation>
<identifier.issn>0167-9473</identifier.issn>
<identifier.uri>http://hdl.handle.net/10722/172504</identifier.uri>
<description.abstract>Linear discriminant analysis (LDA) is a popular technique for supervised dimension reduction. Due to the curse of dimensionality usually suffered by LDA when applied to 2D data, several two-dimensional LDA (2DLDA) methods have been proposed in recent years. Among which, the Y2DLDA method, introduced by Ye et al. (2005), is an important development. The idea is to utilize the underlying 2D data structure to seek for an optimal bilinear transformation. However, it is found that the proposed algorithm does not guarantee convergence. In this paper, we show that the utilization of a bilinear transformation for 2D data is equivalent to modeling the covariance matrix of 2D data as separable covariance matrix. Based on this result, we propose a novel 2DLDA method called separable LDA (SLDA). The main contributions of SLDA include (1) it provides interesting theoretical relationships between LDA and some 2DLDA methods; (2) SLDA provides a building block for mixture extension; (3) unlike Y2DLDA, a neatly analytical solution can be obtained as that in LDA. Empirical results show that our proposed SLDA achieves better recognition performance than Y2DLDA while being computationally much more efficient. &#169; 2012 Elsevier B.V. All rights reserved.</description.abstract>
<language>eng</language>
<publisher>Elsevier BV. The Journal&apos;s web site is located at http://www.elsevier.com/locate/csda</publisher>
<relation.ispartof>Computational Statistics and Data Analysis</relation.ispartof>
<subject>Face Recognition</subject>
<subject>Linear Discriminant Analysis</subject>
<subject>Separable</subject>
<subject>Two-Dimensional Data</subject>
<title>Separable linear discriminant analysis</title>
<type>Article</type>
<description.nature>Link_to_subscribed_fulltext</description.nature>
<identifier.doi>10.1016/j.csda.2012.04.003</identifier.doi>
<identifier.scopus>eid_2-s2.0-84864148086</identifier.scopus>
<identifier.hkuros>210593</identifier.hkuros>
<relation.references>http://www.scopus.com/mlt/select.url?eid=2-s2.0-84864148086&amp;selection=ref&amp;src=s&amp;origin=recordpage</relation.references>
<identifier.volume>56</identifier.volume>
<identifier.issue>12</identifier.issue>
<identifier.spage>4290</identifier.spage>
<identifier.epage>4300</identifier.epage>
<publisher.place>Netherlands</publisher.place>
</item>
Author Affiliations
  1. The University of Hong Kong
  2. Yunnan University