File Download
  Links for fulltext
     (May Require Subscription)
Supplementary

Article: Evolutionary cross-domain discriminative Hessian Eigenmaps

TitleEvolutionary cross-domain discriminative Hessian Eigenmaps
Authors
KeywordsCross-domain learning
Dimension reduction
Evolutionary search
Face recognition
Manifold learning
Web image annotation
Issue Date2010
PublisherIEEE.
Citation
Ieee Transactions On Image Processing, 2010, v. 19 n. 4, p. 1075-1086 How to Cite?
AbstractIs it possible to train a learning model to separate tigers from elks when we have 1) labeled samples of leopard and zebra and 2) unlabelled samples of tiger and elk at hand? Cross-domain learning algorithms can be used to solve the above problem. However, existing cross-domain algorithms cannot be applied for dimension reduction, which plays a key role in computer vision tasks, e.g., face recognition and web image annotation. This paper envisions the cross-domain discriminative dimension reduction to provide an effective solution for cross-domain dimension reduction. In particular, we propose the cross-domain discriminative Hessian Eigenmaps or CDHE for short. CDHE connects training and test samples by minimizing the quadratic distance between the distribution of the training set and that of the test set. Therefore, a common subspace for data representation can be well preserved. Furthermore, we basically expect the discriminative information used to separate leopards and zebra can be shared to separate tigers and elks, and thus we have a chance to duly address the above question. Margin maximization principle is adopted in CDHE so the discriminative information for separating different classes (e.g., leopard and zebra here) can be well preserved. Finally, CDHE encodes the local geometry of each training class (e.g., leopard and zebra here) in the local tangent space which is locally isometric to the data manifold and thus CDHE preserves the intraclass local geometry. The objective function of CDHE is not convex, so the gradient descent strategy can only find a local optimal solution. In this paper, we carefully design an evolutionary search strategy to find a better solution of CDHE. Experimental evidence on both synthetic and real word image datasets demonstrates the effectiveness of CDHE for cross-domain web image annotation and face recognition. © 2006 IEEE.
Persistent Identifierhttp://hdl.handle.net/10722/127357
ISSN
2015 Impact Factor: 3.735
2015 SCImago Journal Rankings: 2.727
ISI Accession Number ID
Funding AgencyGrant Number
HKU-SPF10400016
Nanyang SUGM58020010
Microsoft Operations PTE LTD-NTU Joint RDM48020065
K. C. Wong Education Foundation
Funding Information:

This work was supported in part by the HKU-SPF Grant (under project number 10400016), Nanyang SUG Grant (under project numberM58020010), in part by Microsoft Operations PTE LTD-NTU Joint R&D (under project number M48020065), and in part by the K. C. Wong Education Foundation Award.

References

 

DC FieldValueLanguage
dc.contributor.authorSi, Sen_HK
dc.contributor.authorTao, Den_HK
dc.contributor.authorChan, KPen_HK
dc.date.accessioned2010-10-31T13:20:53Z-
dc.date.available2010-10-31T13:20:53Z-
dc.date.issued2010en_HK
dc.identifier.citationIeee Transactions On Image Processing, 2010, v. 19 n. 4, p. 1075-1086en_HK
dc.identifier.issn1057-7149en_HK
dc.identifier.urihttp://hdl.handle.net/10722/127357-
dc.description.abstractIs it possible to train a learning model to separate tigers from elks when we have 1) labeled samples of leopard and zebra and 2) unlabelled samples of tiger and elk at hand? Cross-domain learning algorithms can be used to solve the above problem. However, existing cross-domain algorithms cannot be applied for dimension reduction, which plays a key role in computer vision tasks, e.g., face recognition and web image annotation. This paper envisions the cross-domain discriminative dimension reduction to provide an effective solution for cross-domain dimension reduction. In particular, we propose the cross-domain discriminative Hessian Eigenmaps or CDHE for short. CDHE connects training and test samples by minimizing the quadratic distance between the distribution of the training set and that of the test set. Therefore, a common subspace for data representation can be well preserved. Furthermore, we basically expect the discriminative information used to separate leopards and zebra can be shared to separate tigers and elks, and thus we have a chance to duly address the above question. Margin maximization principle is adopted in CDHE so the discriminative information for separating different classes (e.g., leopard and zebra here) can be well preserved. Finally, CDHE encodes the local geometry of each training class (e.g., leopard and zebra here) in the local tangent space which is locally isometric to the data manifold and thus CDHE preserves the intraclass local geometry. The objective function of CDHE is not convex, so the gradient descent strategy can only find a local optimal solution. In this paper, we carefully design an evolutionary search strategy to find a better solution of CDHE. Experimental evidence on both synthetic and real word image datasets demonstrates the effectiveness of CDHE for cross-domain web image annotation and face recognition. © 2006 IEEE.en_HK
dc.languageengen_HK
dc.publisherIEEE.en_HK
dc.relation.ispartofIEEE Transactions on Image Processingen_HK
dc.rightsCreative Commons: Attribution 3.0 Hong Kong License-
dc.rightsIEEE Transactions on Image Processing. Copyright © IEEE.-
dc.rights©2010 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.-
dc.subjectCross-domain learningen_HK
dc.subjectDimension reductionen_HK
dc.subjectEvolutionary searchen_HK
dc.subjectFace recognitionen_HK
dc.subjectManifold learningen_HK
dc.subjectWeb image annotationen_HK
dc.titleEvolutionary cross-domain discriminative Hessian Eigenmapsen_HK
dc.typeArticleen_HK
dc.identifier.openurlhttp://library.hku.hk:4550/resserv?sid=HKU:IR&issn=1057-7149&volume=19&issue=4&spage=1075&epage=1086&date=2010&atitle=Evolutionary+cross-domain+discriminative+Hessian+Eigenmaps-
dc.identifier.emailChan, KP:kpchan@cs.hku.hken_HK
dc.identifier.authorityChan, KP=rp00092en_HK
dc.description.naturepublished_or_final_version-
dc.identifier.doi10.1109/TIP.2009.2035867en_HK
dc.identifier.scopuseid_2-s2.0-77949743293en_HK
dc.identifier.hkuros176307en_HK
dc.relation.referenceshttp://www.scopus.com/mlt/select.url?eid=2-s2.0-77949743293&selection=ref&src=s&origin=recordpageen_HK
dc.identifier.volume19en_HK
dc.identifier.issue4en_HK
dc.identifier.spage1075en_HK
dc.identifier.epage1086en_HK
dc.identifier.eissn1941-0042-
dc.identifier.isiWOS:000275662900019-
dc.publisher.placeUnited Statesen_HK
dc.identifier.scopusauthoridSi, S=35422764200en_HK
dc.identifier.scopusauthoridTao, D=7102600334en_HK
dc.identifier.scopusauthoridChan, KP=7406032820en_HK

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats