Trust region Levenberg-Marquardt method for linear SVM

Vinod Kumar Chauhan; Kalpana Dahiya; Anuj Sharma

doi:10.1109/ICAPR.2017.8593090

Trust region Levenberg-Marquardt method for linear SVM

Vinod Kumar Chauhan, Kalpana Dahiya, Anuj Sharma

Computer And Information Sciences

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution book

2 Citations (Scopus)

Abstract

Support Vector Machine is an optimal margin based classification technique in Machine Learning. In this paper, we have proposed Trust Region Levenberg-Marquardt (TRLM) method as a novel problem solver for L2 regularized L2 loss (L2RL2) primal SVM classification problem. Levenberg-Marquardt (LM) method is an extension of Gauss-Newton method for solving least squares non-linear optimization problems and Trust-Region (TR) method is used to find the step size. In LM method, LM parameter λ is changed by an arbitrary factor but in TRLM instead of changing λ, the trust region radius Δ is changed. The proposed solver for L2RL2 primal SVM, performs well with medium sized problems. Experimental results establish TRLM as a solver for linear SVM as it performs at par and better in selective cases than existing state of the art solvers in terms of test accuracy.

Original language	English
Title of host publication	2017 Ninth International Conference on Advances in Pattern Recognition (ICAPR)
Place of Publication	Piscatawy, N.J.
Publisher	IEEE
Pages	1-6
Number of pages	6
ISBN (Electronic)	9781538622414
DOIs	https://doi.org/10.1109/ICAPR.2017.8593090
Publication status	Published - 30 Dec 2018
Event	2017 Ninth International Conference on Advances in Pattern Recognition (ICAPR) - Bangalore Duration: 27 Dec 2017 → 30 Dec 2017

Conference

Conference	2017 Ninth International Conference on Advances in Pattern Recognition (ICAPR)
Period	27/12/17 → 30/12/17

Keywords

support vector machine
machine learning
linear support vector machine
trust region
optimization problem
Kernel
Newton method

Access to Document

10.1109/ICAPR.2017.8593090

Cite this

@inproceedings{a79f40cc297f4b6bae1b7cccc13c6ef0,

title = "Trust region Levenberg-Marquardt method for linear SVM",

abstract = "Support Vector Machine is an optimal margin based classification technique in Machine Learning. In this paper, we have proposed Trust Region Levenberg-Marquardt (TRLM) method as a novel problem solver for L2 regularized L2 loss (L2RL2) primal SVM classification problem. Levenberg-Marquardt (LM) method is an extension of Gauss-Newton method for solving least squares non-linear optimization problems and Trust-Region (TR) method is used to find the step size. In LM method, LM parameter λ is changed by an arbitrary factor but in TRLM instead of changing λ, the trust region radius Δ is changed. The proposed solver for L2RL2 primal SVM, performs well with medium sized problems. Experimental results establish TRLM as a solver for linear SVM as it performs at par and better in selective cases than existing state of the art solvers in terms of test accuracy.",

keywords = "support vector machine, machine learning, linear support vector machine, trust region, optimization problem, Kernel, Newton method",

author = "Chauhan, \{Vinod Kumar\} and Kalpana Dahiya and Anuj Sharma",

year = "2018",

month = dec,

day = "30",

doi = "10.1109/ICAPR.2017.8593090",

language = "English",

pages = "1--6",

booktitle = "2017 Ninth International Conference on Advances in Pattern Recognition (ICAPR)",

publisher = "IEEE",

note = "2017 Ninth International Conference on Advances in Pattern Recognition (ICAPR) ; Conference date: 27-12-2017 Through 30-12-2017",

}

TY - GEN

T1 - Trust region Levenberg-Marquardt method for linear SVM

AU - Chauhan, Vinod Kumar

AU - Dahiya, Kalpana

AU - Sharma, Anuj

PY - 2018/12/30

Y1 - 2018/12/30

N2 - Support Vector Machine is an optimal margin based classification technique in Machine Learning. In this paper, we have proposed Trust Region Levenberg-Marquardt (TRLM) method as a novel problem solver for L2 regularized L2 loss (L2RL2) primal SVM classification problem. Levenberg-Marquardt (LM) method is an extension of Gauss-Newton method for solving least squares non-linear optimization problems and Trust-Region (TR) method is used to find the step size. In LM method, LM parameter λ is changed by an arbitrary factor but in TRLM instead of changing λ, the trust region radius Δ is changed. The proposed solver for L2RL2 primal SVM, performs well with medium sized problems. Experimental results establish TRLM as a solver for linear SVM as it performs at par and better in selective cases than existing state of the art solvers in terms of test accuracy.

AB - Support Vector Machine is an optimal margin based classification technique in Machine Learning. In this paper, we have proposed Trust Region Levenberg-Marquardt (TRLM) method as a novel problem solver for L2 regularized L2 loss (L2RL2) primal SVM classification problem. Levenberg-Marquardt (LM) method is an extension of Gauss-Newton method for solving least squares non-linear optimization problems and Trust-Region (TR) method is used to find the step size. In LM method, LM parameter λ is changed by an arbitrary factor but in TRLM instead of changing λ, the trust region radius Δ is changed. The proposed solver for L2RL2 primal SVM, performs well with medium sized problems. Experimental results establish TRLM as a solver for linear SVM as it performs at par and better in selective cases than existing state of the art solvers in terms of test accuracy.

KW - support vector machine

KW - machine learning

KW - linear support vector machine

KW - trust region

KW - optimization problem

KW - Kernel

KW - Newton method

U2 - 10.1109/ICAPR.2017.8593090

DO - 10.1109/ICAPR.2017.8593090

M3 - Conference contribution book

SP - 1

EP - 6

BT - 2017 Ninth International Conference on Advances in Pattern Recognition (ICAPR)

PB - IEEE

CY - Piscatawy, N.J.

T2 - 2017 Ninth International Conference on Advances in Pattern Recognition (ICAPR)

Y2 - 27 December 2017 through 30 December 2017

ER -

Trust region Levenberg-Marquardt method for linear SVM

Abstract

Conference

Keywords

Access to Document

Fingerprint

Cite this