ε-Sparse representations: Generalized sparse approximation and the equivalent family of SVM tasks

Zoltán Szabó, A. Lőrincz

Research output: Contribution to journalArticle

Abstract

Relation between a family of generalized Support Vector Machine (SVM) problems and the novel ε-sparse representation is provided. In defining ε-sparse representations, we use a natural generalization of the classical εinsensitive cost function for vectors. The insensitive parameter of the SVM problem is transformed into component-wise insensitivity and thus overall sparsification is replaced by component-wise sparsification. The connection between these two problems is built through the generalized Moore-Penrose inverse of the Gram matrix associated to the kernel.

Original languageEnglish
Pages (from-to)605-614
Number of pages10
JournalActa Cybernetica
Volume17
Issue number3
Publication statusPublished - 2006

Fingerprint

Sparse Approximation
Sparse Representation
Support vector machines
Support Vector Machine
Moore-Penrose Generalized Inverse
Cost functions
Gram Matrix
Insensitivity
Cost Function
kernel
Family
Support vector machine
Approximation
Penrose
Kernel
Cost function

ASJC Scopus subject areas

  • Hardware and Architecture
  • Software
  • Computational Theory and Mathematics
  • Theoretical Computer Science

Cite this

ε-Sparse representations : Generalized sparse approximation and the equivalent family of SVM tasks. / Szabó, Zoltán; Lőrincz, A.

In: Acta Cybernetica, Vol. 17, No. 3, 2006, p. 605-614.

Research output: Contribution to journalArticle

@article{7d3c3104840049259e69808c3374a509,
title = "ε-Sparse representations: Generalized sparse approximation and the equivalent family of SVM tasks",
abstract = "Relation between a family of generalized Support Vector Machine (SVM) problems and the novel ε-sparse representation is provided. In defining ε-sparse representations, we use a natural generalization of the classical εinsensitive cost function for vectors. The insensitive parameter of the SVM problem is transformed into component-wise insensitivity and thus overall sparsification is replaced by component-wise sparsification. The connection between these two problems is built through the generalized Moore-Penrose inverse of the Gram matrix associated to the kernel.",
author = "Zolt{\'a}n Szab{\'o} and A. Lőrincz",
year = "2006",
language = "English",
volume = "17",
pages = "605--614",
journal = "Acta Cybernetica",
issn = "0324-721X",
publisher = "University of Szeged",
number = "3",

}

TY - JOUR

T1 - ε-Sparse representations

T2 - Generalized sparse approximation and the equivalent family of SVM tasks

AU - Szabó, Zoltán

AU - Lőrincz, A.

PY - 2006

Y1 - 2006

N2 - Relation between a family of generalized Support Vector Machine (SVM) problems and the novel ε-sparse representation is provided. In defining ε-sparse representations, we use a natural generalization of the classical εinsensitive cost function for vectors. The insensitive parameter of the SVM problem is transformed into component-wise insensitivity and thus overall sparsification is replaced by component-wise sparsification. The connection between these two problems is built through the generalized Moore-Penrose inverse of the Gram matrix associated to the kernel.

AB - Relation between a family of generalized Support Vector Machine (SVM) problems and the novel ε-sparse representation is provided. In defining ε-sparse representations, we use a natural generalization of the classical εinsensitive cost function for vectors. The insensitive parameter of the SVM problem is transformed into component-wise insensitivity and thus overall sparsification is replaced by component-wise sparsification. The connection between these two problems is built through the generalized Moore-Penrose inverse of the Gram matrix associated to the kernel.

UR - http://www.scopus.com/inward/record.url?scp=33746041491&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33746041491&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:33746041491

VL - 17

SP - 605

EP - 614

JO - Acta Cybernetica

JF - Acta Cybernetica

SN - 0324-721X

IS - 3

ER -