Sparse group lasso and high dimensional multinomial classification

Martin Vincent; N.R. Hansen

doi:10.1016/j.csda.2013.06.004

Sparse group lasso and high dimensional multinomial classification

Martin Vincent, N.R. Hansen

Department of Mathematical Sciences

65 Citations (Scopus)

211 Downloads (Pure)

Abstract

The sparse group lasso optimization problem is solved using a coordinate gradient descent algorithm. The algorithm is applicable to a broad class of convex loss functions. Convergence of the algorithm is established, and the algorithm is used to investigate the performance of the multinomial sparse group lasso classifier. On three different real data examples the multinomial group lasso clearly outperforms multinomial lasso in terms of achieved classification error rate and in terms of including fewer features for the classification. An implementation of the multinomial sparse group lasso algorithm is available in the R package msgl. Its performance scales well with the problem size as illustrated by one of the examples considered - a 50 class classification problem with 10 k features, which amounts to estimating 500 k parameters.

Original language	English
Journal	Computational Statistics & Data Analysis
Volume	71
Pages (from-to)	771-786
Number of pages	16
ISSN	0167-9473
DOIs	https://doi.org/10.1016/j.csda.2013.06.004
Publication status	Published - 1 Mar 2014

Access to Document

10.1016/j.csda.2013.06.004

1205.1245v2Submitted manuscript, 472 KBLicence: CC BY-ND

Cite this

@article{ae7391a10a0f499185a71828b03f7642,

title = "Sparse group lasso and high dimensional multinomial classification",

abstract = "The sparse group lasso optimization problem is solved using a coordinate gradient descent algorithm. The algorithm is applicable to a broad class of convex loss functions. Convergence of the algorithm is established, and the algorithm is used to investigate the performance of the multinomial sparse group lasso classifier. On three different real data examples the multinomial group lasso clearly outperforms multinomial lasso in terms of achieved classification error rate and in terms of including fewer features for the classification. An implementation of the multinomial sparse group lasso algorithm is available in the R package msgl. Its performance scales well with the problem size as illustrated by one of the examples considered - a 50 class classification problem with 10 k features, which amounts to estimating 500 k parameters.",

author = "Martin Vincent and N.R. Hansen",

year = "2014",

month = mar,

day = "1",

doi = "10.1016/j.csda.2013.06.004",

language = "English",

volume = "71",

pages = "771--786",

journal = "Computational Statistics and Data Analysis",

issn = "0167-9473",

publisher = "Elsevier",

}

TY - JOUR

T1 - Sparse group lasso and high dimensional multinomial classification

AU - Vincent, Martin

AU - Hansen, N.R.

PY - 2014/3/1

Y1 - 2014/3/1

N2 - The sparse group lasso optimization problem is solved using a coordinate gradient descent algorithm. The algorithm is applicable to a broad class of convex loss functions. Convergence of the algorithm is established, and the algorithm is used to investigate the performance of the multinomial sparse group lasso classifier. On three different real data examples the multinomial group lasso clearly outperforms multinomial lasso in terms of achieved classification error rate and in terms of including fewer features for the classification. An implementation of the multinomial sparse group lasso algorithm is available in the R package msgl. Its performance scales well with the problem size as illustrated by one of the examples considered - a 50 class classification problem with 10 k features, which amounts to estimating 500 k parameters.

AB - The sparse group lasso optimization problem is solved using a coordinate gradient descent algorithm. The algorithm is applicable to a broad class of convex loss functions. Convergence of the algorithm is established, and the algorithm is used to investigate the performance of the multinomial sparse group lasso classifier. On three different real data examples the multinomial group lasso clearly outperforms multinomial lasso in terms of achieved classification error rate and in terms of including fewer features for the classification. An implementation of the multinomial sparse group lasso algorithm is available in the R package msgl. Its performance scales well with the problem size as illustrated by one of the examples considered - a 50 class classification problem with 10 k features, which amounts to estimating 500 k parameters.

UR - http://www.scopus.com/inward/record.url?scp=84889083914&partnerID=8YFLogxK

U2 - 10.1016/j.csda.2013.06.004

DO - 10.1016/j.csda.2013.06.004

M3 - Journal article

AN - SCOPUS:84889083914

SN - 0167-9473

VL - 71

SP - 771

EP - 786

JO - Computational Statistics and Data Analysis

JF - Computational Statistics and Data Analysis

ER -

Sparse group lasso and high dimensional multinomial classification

Abstract

Access to Document

Other files and links

Fingerprint

Cite this