Multiple kernel sparse representations for supervised and unsupervised learning

Jayaraman J. Thiagarajan, Karthikeyan Natesan Ramamurthy, Andreas Spanias

Research output: Contribution to journalArticle

52 Citations (Scopus)

Abstract

In complex visual recognition tasks, it is typical to adopt multiple descriptors, which describe different aspects of the images, for obtaining an improved recognition performance. Descriptors that have diverse forms can be fused into a unified feature space in a principled manner using kernel methods. Sparse models that generalize well to the test data can be learned in the unified kernel space, and appropriate constraints can be incorporated for application in supervised and unsupervised learning. In this paper, we propose to perform sparse coding and dictionary learning in the multiple kernel space, where the weights of the ensemble kernel are tuned based on graph-embedding principles such that class discrimination is maximized. In our proposed algorithm, dictionaries are inferred using multiple levels of 1D subspace clustering in the kernel space, and the sparse codes are obtained using a simple levelwise pursuit scheme. Empirical results for object recognition and image clustering show that our algorithm outperforms existing sparse coding based approaches, and compares favorably to other state-of-the-art methods.

Original languageEnglish (US)
Article number6813695
Pages (from-to)2905-2915
Number of pages11
JournalIEEE Transactions on Image Processing
Volume23
Issue number7
DOIs
StatePublished - 2014

Fingerprint

Unsupervised learning
Supervised learning
Glossaries
Object recognition

Keywords

  • clustering
  • dictionary learning
  • multiple kernel learning
  • object recognition
  • Sparse coding

ASJC Scopus subject areas

  • Computer Graphics and Computer-Aided Design
  • Software

Cite this

Multiple kernel sparse representations for supervised and unsupervised learning. / Thiagarajan, Jayaraman J.; Ramamurthy, Karthikeyan Natesan; Spanias, Andreas.

In: IEEE Transactions on Image Processing, Vol. 23, No. 7, 6813695, 2014, p. 2905-2915.

Research output: Contribution to journalArticle

Thiagarajan, Jayaraman J. ; Ramamurthy, Karthikeyan Natesan ; Spanias, Andreas. / Multiple kernel sparse representations for supervised and unsupervised learning. In: IEEE Transactions on Image Processing. 2014 ; Vol. 23, No. 7. pp. 2905-2915.
@article{88b809e30b574f8593eb9e9adeadf0ac,
title = "Multiple kernel sparse representations for supervised and unsupervised learning",
abstract = "In complex visual recognition tasks, it is typical to adopt multiple descriptors, which describe different aspects of the images, for obtaining an improved recognition performance. Descriptors that have diverse forms can be fused into a unified feature space in a principled manner using kernel methods. Sparse models that generalize well to the test data can be learned in the unified kernel space, and appropriate constraints can be incorporated for application in supervised and unsupervised learning. In this paper, we propose to perform sparse coding and dictionary learning in the multiple kernel space, where the weights of the ensemble kernel are tuned based on graph-embedding principles such that class discrimination is maximized. In our proposed algorithm, dictionaries are inferred using multiple levels of 1D subspace clustering in the kernel space, and the sparse codes are obtained using a simple levelwise pursuit scheme. Empirical results for object recognition and image clustering show that our algorithm outperforms existing sparse coding based approaches, and compares favorably to other state-of-the-art methods.",
keywords = "clustering, dictionary learning, multiple kernel learning, object recognition, Sparse coding",
author = "Thiagarajan, {Jayaraman J.} and Ramamurthy, {Karthikeyan Natesan} and Andreas Spanias",
year = "2014",
doi = "10.1109/TIP.2014.2322938",
language = "English (US)",
volume = "23",
pages = "2905--2915",
journal = "IEEE Transactions on Image Processing",
issn = "1057-7149",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "7",

}

TY - JOUR

T1 - Multiple kernel sparse representations for supervised and unsupervised learning

AU - Thiagarajan, Jayaraman J.

AU - Ramamurthy, Karthikeyan Natesan

AU - Spanias, Andreas

PY - 2014

Y1 - 2014

N2 - In complex visual recognition tasks, it is typical to adopt multiple descriptors, which describe different aspects of the images, for obtaining an improved recognition performance. Descriptors that have diverse forms can be fused into a unified feature space in a principled manner using kernel methods. Sparse models that generalize well to the test data can be learned in the unified kernel space, and appropriate constraints can be incorporated for application in supervised and unsupervised learning. In this paper, we propose to perform sparse coding and dictionary learning in the multiple kernel space, where the weights of the ensemble kernel are tuned based on graph-embedding principles such that class discrimination is maximized. In our proposed algorithm, dictionaries are inferred using multiple levels of 1D subspace clustering in the kernel space, and the sparse codes are obtained using a simple levelwise pursuit scheme. Empirical results for object recognition and image clustering show that our algorithm outperforms existing sparse coding based approaches, and compares favorably to other state-of-the-art methods.

AB - In complex visual recognition tasks, it is typical to adopt multiple descriptors, which describe different aspects of the images, for obtaining an improved recognition performance. Descriptors that have diverse forms can be fused into a unified feature space in a principled manner using kernel methods. Sparse models that generalize well to the test data can be learned in the unified kernel space, and appropriate constraints can be incorporated for application in supervised and unsupervised learning. In this paper, we propose to perform sparse coding and dictionary learning in the multiple kernel space, where the weights of the ensemble kernel are tuned based on graph-embedding principles such that class discrimination is maximized. In our proposed algorithm, dictionaries are inferred using multiple levels of 1D subspace clustering in the kernel space, and the sparse codes are obtained using a simple levelwise pursuit scheme. Empirical results for object recognition and image clustering show that our algorithm outperforms existing sparse coding based approaches, and compares favorably to other state-of-the-art methods.

KW - clustering

KW - dictionary learning

KW - multiple kernel learning

KW - object recognition

KW - Sparse coding

UR - http://www.scopus.com/inward/record.url?scp=84901992572&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84901992572&partnerID=8YFLogxK

U2 - 10.1109/TIP.2014.2322938

DO - 10.1109/TIP.2014.2322938

M3 - Article

C2 - 24833593

AN - SCOPUS:84901992572

VL - 23

SP - 2905

EP - 2915

JO - IEEE Transactions on Image Processing

JF - IEEE Transactions on Image Processing

SN - 1057-7149

IS - 7

M1 - 6813695

ER -