A convergent solution to tensor subspace learning

Huan Wang, Shuicheng Yan, Thomas S Huang, Xiaoou Tang

Research output: Contribution to journalConference article

Abstract

Recently, substantial efforts have been devoted to the subspace learning techniques based on tensor representation, such as 2DLDA [Ye et al., 2004], DATER [Yan et al., 2005] and Tensor Subspace Analysis (TSA) [He et al., 2005]. In this context, a vital yet unsolved problem is that the computational convergency of these iterative algorithms is not guaranteed. In this work, we present a novel solution procedure for general tensor-based subspace learning, followed by a detailed convergency proof of the solution projection matrices and the objective function value. Extensive experiments on real-world databases verify the high convergence speed of the proposed procedure, as well as its superiority in classification capability over traditional solution procedures.

Original languageEnglish (US)
Pages (from-to)629-634
Number of pages6
JournalIJCAI International Joint Conference on Artificial Intelligence
StatePublished - Dec 1 2007
Event20th International Joint Conference on Artificial Intelligence, IJCAI 2007 - Hyderabad, India
Duration: Jan 6 2007Jan 12 2007

Fingerprint

Tensors
Experiments

ASJC Scopus subject areas

  • Artificial Intelligence

Cite this

A convergent solution to tensor subspace learning. / Wang, Huan; Yan, Shuicheng; Huang, Thomas S; Tang, Xiaoou.

In: IJCAI International Joint Conference on Artificial Intelligence, 01.12.2007, p. 629-634.

Research output: Contribution to journalConference article

@article{9b9ca9255242426b9cfdb9089f6e9ed1,
title = "A convergent solution to tensor subspace learning",
abstract = "Recently, substantial efforts have been devoted to the subspace learning techniques based on tensor representation, such as 2DLDA [Ye et al., 2004], DATER [Yan et al., 2005] and Tensor Subspace Analysis (TSA) [He et al., 2005]. In this context, a vital yet unsolved problem is that the computational convergency of these iterative algorithms is not guaranteed. In this work, we present a novel solution procedure for general tensor-based subspace learning, followed by a detailed convergency proof of the solution projection matrices and the objective function value. Extensive experiments on real-world databases verify the high convergence speed of the proposed procedure, as well as its superiority in classification capability over traditional solution procedures.",
author = "Huan Wang and Shuicheng Yan and Huang, {Thomas S} and Xiaoou Tang",
year = "2007",
month = "12",
day = "1",
language = "English (US)",
pages = "629--634",
journal = "IJCAI International Joint Conference on Artificial Intelligence",
issn = "1045-0823",

}

TY - JOUR

T1 - A convergent solution to tensor subspace learning

AU - Wang, Huan

AU - Yan, Shuicheng

AU - Huang, Thomas S

AU - Tang, Xiaoou

PY - 2007/12/1

Y1 - 2007/12/1

N2 - Recently, substantial efforts have been devoted to the subspace learning techniques based on tensor representation, such as 2DLDA [Ye et al., 2004], DATER [Yan et al., 2005] and Tensor Subspace Analysis (TSA) [He et al., 2005]. In this context, a vital yet unsolved problem is that the computational convergency of these iterative algorithms is not guaranteed. In this work, we present a novel solution procedure for general tensor-based subspace learning, followed by a detailed convergency proof of the solution projection matrices and the objective function value. Extensive experiments on real-world databases verify the high convergence speed of the proposed procedure, as well as its superiority in classification capability over traditional solution procedures.

AB - Recently, substantial efforts have been devoted to the subspace learning techniques based on tensor representation, such as 2DLDA [Ye et al., 2004], DATER [Yan et al., 2005] and Tensor Subspace Analysis (TSA) [He et al., 2005]. In this context, a vital yet unsolved problem is that the computational convergency of these iterative algorithms is not guaranteed. In this work, we present a novel solution procedure for general tensor-based subspace learning, followed by a detailed convergency proof of the solution projection matrices and the objective function value. Extensive experiments on real-world databases verify the high convergence speed of the proposed procedure, as well as its superiority in classification capability over traditional solution procedures.

UR - http://www.scopus.com/inward/record.url?scp=84880903188&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84880903188&partnerID=8YFLogxK

M3 - Conference article

AN - SCOPUS:84880903188

SP - 629

EP - 634

JO - IJCAI International Joint Conference on Artificial Intelligence

JF - IJCAI International Joint Conference on Artificial Intelligence

SN - 1045-0823

ER -