Sparse Proximal Support Vector Machine with a Specialized Interior-Point Method

Support vector machine (SVM) is a widely used method for classification. Proximal support vector machine (PSVM) is an extension of SVM and a promising method to lead to a fast and simple algorithm for generating a classifier. Motivated by the fast computational efforts of PSVM and the properties of...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Journal of the Operations Research Society of China (Internet) Jg. 3; H. 1; S. 1 - 15
Hauptverfasser: Bai, Yan-Qin, Zhu, Zhao-Ying, Yan, Wen-Li
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Heidelberg Operations Research Society of China 01.03.2015
Springer Nature B.V
Schlagworte:
Algorithms
Classification
Classifiers
Eigenvalues
Feature selection
Machine learning
Management Science
Mathematics
Mathematics and Statistics
Operations Research
Optimization
Original Paper
Repositories
Support vector machines
90C20
90C10
49M20
65K05
Classification accuracy
Proximal support vector machine
Preconditioned conjugate gradients algorithm
Interior-point methods
ISSN:2194-668X, 2194-6698
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Support vector machine (SVM) is a widely used method for classification. Proximal support vector machine (PSVM) is an extension of SVM and a promising method to lead to a fast and simple algorithm for generating a classifier. Motivated by the fast computational efforts of PSVM and the properties of sparse solution yielded by ℓ 1 -norm, in this paper, we first propose a PSVM with a cardinality constraint which is eventually relaxed by ℓ 1 -norm and leads to a trade-off ℓ 1 - ℓ 2 regularized sparse PSVM. Next we convert this ℓ 1 - ℓ 2 regularized sparse PSVM into an equivalent form of ℓ 1 regularized least squares (LS) and solve it by a specialized interior-point method proposed by Kim et al. (J Sel Top Signal Process 12:1932–4553, 2007 ). Finally, ℓ 1 - ℓ 2 regularized sparse PSVM is illustrated by means of a real-world dataset taken from the University of California, Irvine Machine Learning Repository (UCI Repository). Moreover, we compare the numerical results with the existing models such as generalized eigenvalue proximal SVM (GEPSVM), PSVM, and SVM-Light. The numerical results show that the ℓ 1 - ℓ 2 regularized sparse PSVM achieves not only better accuracy rate of classification than those of GEPSVM, PSVM, and SVM-Light, but also a sparser classifier compared with the ℓ 1 -PSVM.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:2194-668X
2194-6698
DOI:10.1007/s40305-014-0068-5