Feature elimination in kernel machines in moderately high dimensions

Sayan Dasgupta; Yair Goldberg; Michael R. Kosorok

doi:10.1214/18-AOS1696

Feature elimination in kernel machines in moderately high dimensions

Sayan Dasgupta, Yair Goldberg, Michael R. Kosorok

Department of Statistics

Research output: Contribution to journal › Article › peer-review

Abstract

We develop an approach for feature elimination in statistical learning with kernel machines, based on recursive elimination of features. We present theoretical properties of this method and show that it is uniformly consistent in finding the correct feature space under certain generalized assumptions. We present a few case studies to show that the assumptions are met in most practical situations and present simulation results to demonstrate performance of the proposed approach.

Original language	English
Pages (from-to)	497-526
Number of pages	30
Journal	Annals of Statistics
Volume	47
Issue number	1
DOIs	https://doi.org/10.1214/18-AOS1696
State	Published - Feb 2019

Bibliographical note

Funding Information:
Received December 2015; revised November 2017. 1Supported in part by Israel Science Foundation Grant 1308/12 and supported in part by United States NCI Grant PO1 CA142538. MSC2010 subject classifications. 62G20. Key words and phrases. Kernel machines, support vector machines, variable selection, recursive feature elimination.

Publisher Copyright:
© Institute of Mathematical Statistics, 2019

Keywords

Kernel machines
Recursive feature elimination
Support vector machines
Variable selection

ASJC Scopus subject areas

Statistics and Probability
Statistics, Probability and Uncertainty

Access to Document

10.1214/18-AOS1696

Cite this

@article{b2a9c0e768f74bc9b3e396053279bb61,

title = "Feature elimination in kernel machines in moderately high dimensions",

abstract = "We develop an approach for feature elimination in statistical learning with kernel machines, based on recursive elimination of features. We present theoretical properties of this method and show that it is uniformly consistent in finding the correct feature space under certain generalized assumptions. We present a few case studies to show that the assumptions are met in most practical situations and present simulation results to demonstrate performance of the proposed approach.",

keywords = "Kernel machines, Recursive feature elimination, Support vector machines, Variable selection",

author = "Sayan Dasgupta and Yair Goldberg and Kosorok, {Michael R.}",

note = "Funding Information: Received December 2015; revised November 2017. 1Supported in part by Israel Science Foundation Grant 1308/12 and supported in part by United States NCI Grant PO1 CA142538. MSC2010 subject classifications. 62G20. Key words and phrases. Kernel machines, support vector machines, variable selection, recursive feature elimination. Publisher Copyright: {\textcopyright} Institute of Mathematical Statistics, 2019",

year = "2019",

month = feb,

doi = "10.1214/18-AOS1696",

language = "English",

volume = "47",

pages = "497--526",

journal = "Annals of Statistics",

issn = "0090-5364",

publisher = "Institute of Mathematical Statistics",

number = "1",

}

TY - JOUR

T1 - Feature elimination in kernel machines in moderately high dimensions

AU - Dasgupta, Sayan

AU - Goldberg, Yair

AU - Kosorok, Michael R.

N1 - Funding Information: Received December 2015; revised November 2017. 1Supported in part by Israel Science Foundation Grant 1308/12 and supported in part by United States NCI Grant PO1 CA142538. MSC2010 subject classifications. 62G20. Key words and phrases. Kernel machines, support vector machines, variable selection, recursive feature elimination. Publisher Copyright: © Institute of Mathematical Statistics, 2019

PY - 2019/2

Y1 - 2019/2

N2 - We develop an approach for feature elimination in statistical learning with kernel machines, based on recursive elimination of features. We present theoretical properties of this method and show that it is uniformly consistent in finding the correct feature space under certain generalized assumptions. We present a few case studies to show that the assumptions are met in most practical situations and present simulation results to demonstrate performance of the proposed approach.

AB - We develop an approach for feature elimination in statistical learning with kernel machines, based on recursive elimination of features. We present theoretical properties of this method and show that it is uniformly consistent in finding the correct feature space under certain generalized assumptions. We present a few case studies to show that the assumptions are met in most practical situations and present simulation results to demonstrate performance of the proposed approach.

KW - Kernel machines

KW - Recursive feature elimination

KW - Support vector machines

KW - Variable selection

UR - http://www.scopus.com/inward/record.url?scp=85057823648&partnerID=8YFLogxK

U2 - 10.1214/18-AOS1696

DO - 10.1214/18-AOS1696

M3 - Article

AN - SCOPUS:85057823648

SN - 0090-5364

VL - 47

SP - 497

EP - 526

JO - Annals of Statistics

JF - Annals of Statistics

IS - 1

ER -

Feature elimination in kernel machines in moderately high dimensions

Abstract

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this