An ensemble of deep support vector machines for image categorization

Azizi Abdullah, Remco C. Veltkamp, Marco A. Wiering

Research output: Chapter in Book/Report/Conference proceedingConference contribution

20 Citations (Scopus)

Abstract

This paper presents the deep support vector machine (D-SVM) inspired by the increasing popularity of deep belief networks for image recognition. Our deep SVM trains an SVM in the standard way and then uses the kernel activations of support vectors as inputs for training another SVM at the next layer. In this way, instead of the normal linear combination of kernel activations, we can create non-linear combinations of kernel activations on prototype examples. Furthermore, we combine different descriptors in an ensemble of deep SVMs where the product rule is used for combining probability estimates of the different classifiers. We have performed experiments on 20 classes from the Caltech object database and 10 classes from the Corel dataset. The results show that our ensemble of deep SVMs signi.cantly outperforms the naive approach that combines all descriptors directly in a very large single input vector for an SVM. Furthermore, our ensemble of D-SVMs achieves an accuracy of 95.2% on the Corel dataset with 10 classes, which is the best performance reported in literature until now.

Original languageEnglish
Title of host publicationSoCPaR 2009 - Soft Computing and Pattern Recognition
Pages301-306
Number of pages6
DOIs
Publication statusPublished - 2009
Externally publishedYes
EventInternational Conference on Soft Computing and Pattern Recognition, SoCPaR 2009 - Malacca
Duration: 4 Dec 20097 Dec 2009

Other

OtherInternational Conference on Soft Computing and Pattern Recognition, SoCPaR 2009
CityMalacca
Period4/12/097/12/09

Fingerprint

Support vector machines
Chemical activation
Image recognition
Bayesian networks
Classifiers
Experiments
Object-oriented databases

Keywords

  • Deep architectures
  • Ensemble methods
  • Image categorization
  • Product rule
  • Support vector machines

ASJC Scopus subject areas

  • Computational Theory and Mathematics
  • Computer Vision and Pattern Recognition
  • Software

Cite this

Abdullah, A., Veltkamp, R. C., & Wiering, M. A. (2009). An ensemble of deep support vector machines for image categorization. In SoCPaR 2009 - Soft Computing and Pattern Recognition (pp. 301-306). [5370984] https://doi.org/10.1109/SoCPaR.2009.67

An ensemble of deep support vector machines for image categorization. / Abdullah, Azizi; Veltkamp, Remco C.; Wiering, Marco A.

SoCPaR 2009 - Soft Computing and Pattern Recognition. 2009. p. 301-306 5370984.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abdullah, A, Veltkamp, RC & Wiering, MA 2009, An ensemble of deep support vector machines for image categorization. in SoCPaR 2009 - Soft Computing and Pattern Recognition., 5370984, pp. 301-306, International Conference on Soft Computing and Pattern Recognition, SoCPaR 2009, Malacca, 4/12/09. https://doi.org/10.1109/SoCPaR.2009.67
Abdullah A, Veltkamp RC, Wiering MA. An ensemble of deep support vector machines for image categorization. In SoCPaR 2009 - Soft Computing and Pattern Recognition. 2009. p. 301-306. 5370984 https://doi.org/10.1109/SoCPaR.2009.67
Abdullah, Azizi ; Veltkamp, Remco C. ; Wiering, Marco A. / An ensemble of deep support vector machines for image categorization. SoCPaR 2009 - Soft Computing and Pattern Recognition. 2009. pp. 301-306
@inproceedings{12737937213040c7ad2ddc7c46f520fd,
title = "An ensemble of deep support vector machines for image categorization",
abstract = "This paper presents the deep support vector machine (D-SVM) inspired by the increasing popularity of deep belief networks for image recognition. Our deep SVM trains an SVM in the standard way and then uses the kernel activations of support vectors as inputs for training another SVM at the next layer. In this way, instead of the normal linear combination of kernel activations, we can create non-linear combinations of kernel activations on prototype examples. Furthermore, we combine different descriptors in an ensemble of deep SVMs where the product rule is used for combining probability estimates of the different classifiers. We have performed experiments on 20 classes from the Caltech object database and 10 classes from the Corel dataset. The results show that our ensemble of deep SVMs signi.cantly outperforms the naive approach that combines all descriptors directly in a very large single input vector for an SVM. Furthermore, our ensemble of D-SVMs achieves an accuracy of 95.2{\%} on the Corel dataset with 10 classes, which is the best performance reported in literature until now.",
keywords = "Deep architectures, Ensemble methods, Image categorization, Product rule, Support vector machines",
author = "Azizi Abdullah and Veltkamp, {Remco C.} and Wiering, {Marco A.}",
year = "2009",
doi = "10.1109/SoCPaR.2009.67",
language = "English",
isbn = "9780769538792",
pages = "301--306",
booktitle = "SoCPaR 2009 - Soft Computing and Pattern Recognition",

}

TY - GEN

T1 - An ensemble of deep support vector machines for image categorization

AU - Abdullah, Azizi

AU - Veltkamp, Remco C.

AU - Wiering, Marco A.

PY - 2009

Y1 - 2009

N2 - This paper presents the deep support vector machine (D-SVM) inspired by the increasing popularity of deep belief networks for image recognition. Our deep SVM trains an SVM in the standard way and then uses the kernel activations of support vectors as inputs for training another SVM at the next layer. In this way, instead of the normal linear combination of kernel activations, we can create non-linear combinations of kernel activations on prototype examples. Furthermore, we combine different descriptors in an ensemble of deep SVMs where the product rule is used for combining probability estimates of the different classifiers. We have performed experiments on 20 classes from the Caltech object database and 10 classes from the Corel dataset. The results show that our ensemble of deep SVMs signi.cantly outperforms the naive approach that combines all descriptors directly in a very large single input vector for an SVM. Furthermore, our ensemble of D-SVMs achieves an accuracy of 95.2% on the Corel dataset with 10 classes, which is the best performance reported in literature until now.

AB - This paper presents the deep support vector machine (D-SVM) inspired by the increasing popularity of deep belief networks for image recognition. Our deep SVM trains an SVM in the standard way and then uses the kernel activations of support vectors as inputs for training another SVM at the next layer. In this way, instead of the normal linear combination of kernel activations, we can create non-linear combinations of kernel activations on prototype examples. Furthermore, we combine different descriptors in an ensemble of deep SVMs where the product rule is used for combining probability estimates of the different classifiers. We have performed experiments on 20 classes from the Caltech object database and 10 classes from the Corel dataset. The results show that our ensemble of deep SVMs signi.cantly outperforms the naive approach that combines all descriptors directly in a very large single input vector for an SVM. Furthermore, our ensemble of D-SVMs achieves an accuracy of 95.2% on the Corel dataset with 10 classes, which is the best performance reported in literature until now.

KW - Deep architectures

KW - Ensemble methods

KW - Image categorization

KW - Product rule

KW - Support vector machines

UR - http://www.scopus.com/inward/record.url?scp=77649297423&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77649297423&partnerID=8YFLogxK

U2 - 10.1109/SoCPaR.2009.67

DO - 10.1109/SoCPaR.2009.67

M3 - Conference contribution

SN - 9780769538792

SP - 301

EP - 306

BT - SoCPaR 2009 - Soft Computing and Pattern Recognition

ER -