A new perspective on convex relaxations of sparse SVM

Noam Goldberg; Sven Leyffer; Todd Munsonz

doi:10.1137/1.9781611972832.50

A new perspective on convex relaxations of sparse SVM

Noam Goldberg, Sven Leyffer, Todd Munsonz

Argonne National Laboratory

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

1 Scopus citations

Abstract

This paper proposes a convex relaxation of a sparse support vector machine (SVM) based on the perspective relaxation of mixed-integer nonlinear programs. We seek to minimize the zero-norm of the hyperplane normal vector with a standard SVM hinge-loss penalty and extend our approach to a zero-one loss penalty. The relaxation that we propose is a second-order cone formulation that can be efficiently solved by standard conic optimization solvers. We compare the optimization properties and classification performance of the second-order cone formulation with previous sparse SVM formulations suggested in the literature.

Original language	English
Title of host publication	Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013
Editors	Joydeep Ghosh, Zoran Obradovic, Jennifer Dy, Zhi-Hua Zhou, Chandrika Kamath, Srinivasan Parthasarathy
Publisher	Siam Society
Pages	450-457
Number of pages	8
ISBN (Electronic)	9781611972627
DOIs	https://doi.org/10.1137/1.9781611972832.50
State	Published - 2013
Externally published	Yes
Event	SIAM International Conference on Data Mining, SDM 2013 - Austin, United States Duration: 2 May 2013 → 4 May 2013

Publication series

Name	Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013

Conference

Conference	SIAM International Conference on Data Mining, SDM 2013
Country/Territory	United States
City	Austin
Period	2/05/13 → 4/05/13

Bibliographical note

Publisher Copyright:
Copyright © SIAM.

Access to Document

10.1137/1.9781611972832.50

Cite this

Goldberg, N., Leyffer, S., & Munsonz, T. (2013). A new perspective on convex relaxations of sparse SVM. In J. Ghosh, Z. Obradovic, J. Dy, Z.-H. Zhou, C. Kamath, & S. Parthasarathy (Eds.), Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013 (pp. 450-457). (Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013). Siam Society. https://doi.org/10.1137/1.9781611972832.50

Goldberg, Noam ; Leyffer, Sven ; Munsonz, Todd. / A new perspective on convex relaxations of sparse SVM. Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013. editor / Joydeep Ghosh ; Zoran Obradovic ; Jennifer Dy ; Zhi-Hua Zhou ; Chandrika Kamath ; Srinivasan Parthasarathy. Siam Society, 2013. pp. 450-457 (Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013).

@inproceedings{922a928965c948a6ae4ba0fadffa55bd,

title = "A new perspective on convex relaxations of sparse SVM",

abstract = "This paper proposes a convex relaxation of a sparse support vector machine (SVM) based on the perspective relaxation of mixed-integer nonlinear programs. We seek to minimize the zero-norm of the hyperplane normal vector with a standard SVM hinge-loss penalty and extend our approach to a zero-one loss penalty. The relaxation that we propose is a second-order cone formulation that can be efficiently solved by standard conic optimization solvers. We compare the optimization properties and classification performance of the second-order cone formulation with previous sparse SVM formulations suggested in the literature.",

author = "Noam Goldberg and Sven Leyffer and Todd Munsonz",

note = "Publisher Copyright: Copyright {\textcopyright} SIAM.; SIAM International Conference on Data Mining, SDM 2013 ; Conference date: 02-05-2013 Through 04-05-2013",

year = "2013",

doi = "10.1137/1.9781611972832.50",

language = "אנגלית",

series = "Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013",

publisher = "Siam Society",

pages = "450--457",

editor = "Joydeep Ghosh and Zoran Obradovic and Jennifer Dy and Zhi-Hua Zhou and Chandrika Kamath and Srinivasan Parthasarathy",

booktitle = "Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013",

}

Goldberg, N, Leyffer, S & Munsonz, T 2013, A new perspective on convex relaxations of sparse SVM. in J Ghosh, Z Obradovic, J Dy, Z-H Zhou, C Kamath & S Parthasarathy (eds), Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013. Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013, Siam Society, pp. 450-457, SIAM International Conference on Data Mining, SDM 2013, Austin, United States, 2/05/13. https://doi.org/10.1137/1.9781611972832.50

A new perspective on convex relaxations of sparse SVM. / Goldberg, Noam; Leyffer, Sven; Munsonz, Todd.
Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013. ed. / Joydeep Ghosh; Zoran Obradovic; Jennifer Dy; Zhi-Hua Zhou; Chandrika Kamath; Srinivasan Parthasarathy. Siam Society, 2013. p. 450-457 (Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - A new perspective on convex relaxations of sparse SVM

AU - Goldberg, Noam

AU - Leyffer, Sven

AU - Munsonz, Todd

PY - 2013

Y1 - 2013

N2 - This paper proposes a convex relaxation of a sparse support vector machine (SVM) based on the perspective relaxation of mixed-integer nonlinear programs. We seek to minimize the zero-norm of the hyperplane normal vector with a standard SVM hinge-loss penalty and extend our approach to a zero-one loss penalty. The relaxation that we propose is a second-order cone formulation that can be efficiently solved by standard conic optimization solvers. We compare the optimization properties and classification performance of the second-order cone formulation with previous sparse SVM formulations suggested in the literature.

AB - This paper proposes a convex relaxation of a sparse support vector machine (SVM) based on the perspective relaxation of mixed-integer nonlinear programs. We seek to minimize the zero-norm of the hyperplane normal vector with a standard SVM hinge-loss penalty and extend our approach to a zero-one loss penalty. The relaxation that we propose is a second-order cone formulation that can be efficiently solved by standard conic optimization solvers. We compare the optimization properties and classification performance of the second-order cone formulation with previous sparse SVM formulations suggested in the literature.

UR - http://www.scopus.com/inward/record.url?scp=84960078213&partnerID=8YFLogxK

U2 - 10.1137/1.9781611972832.50

DO - 10.1137/1.9781611972832.50

M3 - ???researchoutput.researchoutputtypes.contributiontobookanthology.conference???

AN - SCOPUS:84960078213

T3 - Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013

SP - 450

EP - 457

BT - Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013

A2 - Ghosh, Joydeep

A2 - Obradovic, Zoran

A2 - Dy, Jennifer

A2 - Zhou, Zhi-Hua

A2 - Kamath, Chandrika

A2 - Parthasarathy, Srinivasan

PB - Siam Society

T2 - SIAM International Conference on Data Mining, SDM 2013

Y2 - 2 May 2013 through 4 May 2013

ER -

Goldberg N, Leyffer S, Munsonz T. A new perspective on convex relaxations of sparse SVM. In Ghosh J, Obradovic Z, Dy J, Zhou ZH, Kamath C, Parthasarathy S, editors, Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013. Siam Society. 2013. p. 450-457. (Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013). doi: 10.1137/1.9781611972832.50

A new perspective on convex relaxations of sparse SVM

Abstract

Publication series

Conference

Bibliographical note

Access to Document

Other files and links

Fingerprint

Cite this