A Maximal Margin Hypersphere SVM

Rui Malha; Paula Amaral

doi:10.1007/978-3-030-86976-2_21

A Maximal Margin Hypersphere SVM

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

Abstract

In this work we propose a generalization of the Support Vector Machine (SVM) method in which the separator is a curve, but the concept of margin and maximization of the margin is still present. The idea of using different functions for the separation has been explored in particular in the scope of hyperspheres. However, most of these proposals use two spheres, using concepts different from maximal margin or one sphere but with a poor performance when data from the classes have a linear shape. In this paper we present a formulation of the linear SVM that generalizes it to a spherical separation shape, but still maximizing the margin. A linear relaxation of this quadratic formulation is also presented. The performance of these two formulations for classification purpose is tested and the results are encouraging.

Original language	English
Title of host publication	Computational Science and Its Applications – ICCSA 2021: 21st International Conference, Cagliari, Italy, September 13–16, 2021, Proceedings, Part V
Editors	Osvaldo Gervasi, Beniamino Murgante, Sanjay Misra, Chiara Garau, Ivan Blečić, David Taniar, Bernady O. Apduhan, Ana Maria Rocha, Eufemia Tarantino, Carmelo Maria Torre
Place of Publication	Cham
Publisher	Springer
Pages	304-319
Number of pages	16
ISBN (Electronic)	978-3-030-86976-2
ISBN (Print)	978-3-030-86975-5
DOIs	https://doi.org/10.1007/978-3-030-86976-2_21
Publication status	Published - 2021
Event	21st International Conference on Computational Science and Its Applications, ICCSA 2021 - Virtual, Online Duration: 13 Sept 2021 → 16 Sept 2021

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Publisher	Springer
Volume	12953 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	21st International Conference on Computational Science and Its Applications, ICCSA 2021
City	Virtual, Online
Period	13/09/21 → 16/09/21

Keywords

Automatic classification
Non-linear SVM
SVM

Access to Document

10.1007/978-3-030-86976-2_21

Cite this

Malha, R., & Amaral, P. (2021). A Maximal Margin Hypersphere SVM. In O. Gervasi, B. Murgante, S. Misra, C. Garau, I. Blečić, D. Taniar, B. O. Apduhan, A. M. Rocha, E. Tarantino, & C. M. Torre (Eds.), Computational Science and Its Applications – ICCSA 2021: 21st International Conference, Cagliari, Italy, September 13–16, 2021, Proceedings, Part V (pp. 304-319). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12953 LNCS). Springer. https://doi.org/10.1007/978-3-030-86976-2_21

Malha, Rui ; Amaral, Paula. / A Maximal Margin Hypersphere SVM. Computational Science and Its Applications – ICCSA 2021: 21st International Conference, Cagliari, Italy, September 13–16, 2021, Proceedings, Part V. editor / Osvaldo Gervasi ; Beniamino Murgante ; Sanjay Misra ; Chiara Garau ; Ivan Blečić ; David Taniar ; Bernady O. Apduhan ; Ana Maria Rocha ; Eufemia Tarantino ; Carmelo Maria Torre. Cham : Springer, 2021. pp. 304-319 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{c620e2d02e6d45b69830ba98bc22b2d1,

title = "A Maximal Margin Hypersphere SVM",

abstract = "In this work we propose a generalization of the Support Vector Machine (SVM) method in which the separator is a curve, but the concept of margin and maximization of the margin is still present. The idea of using different functions for the separation has been explored in particular in the scope of hyperspheres. However, most of these proposals use two spheres, using concepts different from maximal margin or one sphere but with a poor performance when data from the classes have a linear shape. In this paper we present a formulation of the linear SVM that generalizes it to a spherical separation shape, but still maximizing the margin. A linear relaxation of this quadratic formulation is also presented. The performance of these two formulations for classification purpose is tested and the results are encouraging.",

keywords = "Automatic classification, Non-linear SVM, SVM",

author = "Rui Malha and Paula Amaral",

note = "Funding Information: info:eu-repo/grantAgreement/FCT/6817 - DCRRNI ID/UIDB%2F00297%2F2020/PT# Publisher Copyright: {\textcopyright} 2021, Springer Nature Switzerland AG.; 21st International Conference on Computational Science and Its Applications, ICCSA 2021 ; Conference date: 13-09-2021 Through 16-09-2021",

year = "2021",

doi = "10.1007/978-3-030-86976-2_21",

language = "English",

isbn = "978-3-030-86975-5",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer",

pages = "304--319",

editor = "Osvaldo Gervasi and Beniamino Murgante and Sanjay Misra and Chiara Garau and Ivan Ble{\v c}i{\'c} and David Taniar and Apduhan, {Bernady O.} and Rocha, {Ana Maria} and Eufemia Tarantino and Torre, {Carmelo Maria}",

booktitle = "Computational Science and Its Applications – ICCSA 2021: 21st International Conference, Cagliari, Italy, September 13–16, 2021, Proceedings, Part V",

address = "Netherlands",

}

Malha, R & Amaral, P 2021, A Maximal Margin Hypersphere SVM. in O Gervasi, B Murgante, S Misra, C Garau, I Blečić, D Taniar, BO Apduhan, AM Rocha, E Tarantino & CM Torre (eds), Computational Science and Its Applications – ICCSA 2021: 21st International Conference, Cagliari, Italy, September 13–16, 2021, Proceedings, Part V. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 12953 LNCS, Springer, Cham, pp. 304-319, 21st International Conference on Computational Science and Its Applications, ICCSA 2021, Virtual, Online, 13/09/21. https://doi.org/10.1007/978-3-030-86976-2_21

A Maximal Margin Hypersphere SVM. / Malha, Rui ; Amaral, Paula.
Computational Science and Its Applications – ICCSA 2021: 21st International Conference, Cagliari, Italy, September 13–16, 2021, Proceedings, Part V. ed. / Osvaldo Gervasi; Beniamino Murgante; Sanjay Misra; Chiara Garau; Ivan Blečić; David Taniar; Bernady O. Apduhan; Ana Maria Rocha; Eufemia Tarantino; Carmelo Maria Torre. Cham: Springer, 2021. p. 304-319 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12953 LNCS).

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

TY - GEN

T1 - A Maximal Margin Hypersphere SVM

AU - Malha, Rui

AU - Amaral, Paula

PY - 2021

Y1 - 2021

N2 - In this work we propose a generalization of the Support Vector Machine (SVM) method in which the separator is a curve, but the concept of margin and maximization of the margin is still present. The idea of using different functions for the separation has been explored in particular in the scope of hyperspheres. However, most of these proposals use two spheres, using concepts different from maximal margin or one sphere but with a poor performance when data from the classes have a linear shape. In this paper we present a formulation of the linear SVM that generalizes it to a spherical separation shape, but still maximizing the margin. A linear relaxation of this quadratic formulation is also presented. The performance of these two formulations for classification purpose is tested and the results are encouraging.

AB - In this work we propose a generalization of the Support Vector Machine (SVM) method in which the separator is a curve, but the concept of margin and maximization of the margin is still present. The idea of using different functions for the separation has been explored in particular in the scope of hyperspheres. However, most of these proposals use two spheres, using concepts different from maximal margin or one sphere but with a poor performance when data from the classes have a linear shape. In this paper we present a formulation of the linear SVM that generalizes it to a spherical separation shape, but still maximizing the margin. A linear relaxation of this quadratic formulation is also presented. The performance of these two formulations for classification purpose is tested and the results are encouraging.

KW - Automatic classification

KW - Non-linear SVM

KW - SVM

UR - http://www.scopus.com/inward/record.url?scp=85115691667&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-86976-2_21

DO - 10.1007/978-3-030-86976-2_21

M3 - Conference contribution

AN - SCOPUS:85115691667

SN - 978-3-030-86975-5

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 304

EP - 319

BT - Computational Science and Its Applications – ICCSA 2021: 21st International Conference, Cagliari, Italy, September 13–16, 2021, Proceedings, Part V

A2 - Gervasi, Osvaldo

A2 - Murgante, Beniamino

A2 - Misra, Sanjay

A2 - Garau, Chiara

A2 - Blečić, Ivan

A2 - Taniar, David

A2 - Apduhan, Bernady O.

A2 - Rocha, Ana Maria

A2 - Tarantino, Eufemia

A2 - Torre, Carmelo Maria

PB - Springer

CY - Cham

T2 - 21st International Conference on Computational Science and Its Applications, ICCSA 2021

Y2 - 13 September 2021 through 16 September 2021

ER -

Malha R , Amaral P. A Maximal Margin Hypersphere SVM. In Gervasi O, Murgante B, Misra S, Garau C, Blečić I, Taniar D, Apduhan BO, Rocha AM, Tarantino E, Torre CM, editors, Computational Science and Its Applications – ICCSA 2021: 21st International Conference, Cagliari, Italy, September 13–16, 2021, Proceedings, Part V. Cham: Springer. 2021. p. 304-319. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-86976-2_21

A Maximal Margin Hypersphere SVM

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this