Regularization in kernel learning

Shahar Mendelson; Joseph Neeman

doi:10.1214/09-AOS728

Regularization in kernel learning

Shahar Mendelson^*, Joseph Neeman

^*Corresponding author for this work

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

81 Citations (Scopus)

Abstract

Under mild assumptions on the kernel, we obtain the best known error rates in a regularized learning scenario taking place in the corresponding reproducing kernel Hilbert space (RKHS). The main novelty in the analysis is a proof that one can use a regularization term that grows significantly slower than the standard quadratic growth in the RKHS norm.

Original language	English
Pages (from-to)	526-565
Number of pages	40
Journal	Annals of Statistics
Volume	38
Issue number	1
DOIs	https://doi.org/10.1214/09-AOS728
Publication status	Published - Feb 2010

Access to Document

10.1214/09-AOS728

Cite this

@article{0fc31d2402e14696a3f60073ed3a630a,

title = "Regularization in kernel learning",

abstract = "Under mild assumptions on the kernel, we obtain the best known error rates in a regularized learning scenario taking place in the corresponding reproducing kernel Hilbert space (RKHS). The main novelty in the analysis is a proof that one can use a regularization term that grows significantly slower than the standard quadratic growth in the RKHS norm.",

keywords = "Least-squares, Model selection, Regression, Regulation, Reproducing kernel Hilbert space",

author = "Shahar Mendelson and Joseph Neeman",

year = "2010",

month = feb,

doi = "10.1214/09-AOS728",

language = "English",

volume = "38",

pages = "526--565",

journal = "Annals of Statistics",

issn = "0090-5364",

publisher = "Institute of Mathematical Statistics",

number = "1",

}

TY - JOUR

T1 - Regularization in kernel learning

AU - Mendelson, Shahar

AU - Neeman, Joseph

PY - 2010/2

Y1 - 2010/2

N2 - Under mild assumptions on the kernel, we obtain the best known error rates in a regularized learning scenario taking place in the corresponding reproducing kernel Hilbert space (RKHS). The main novelty in the analysis is a proof that one can use a regularization term that grows significantly slower than the standard quadratic growth in the RKHS norm.

AB - Under mild assumptions on the kernel, we obtain the best known error rates in a regularized learning scenario taking place in the corresponding reproducing kernel Hilbert space (RKHS). The main novelty in the analysis is a proof that one can use a regularization term that grows significantly slower than the standard quadratic growth in the RKHS norm.

KW - Least-squares

KW - Model selection

KW - Regression

KW - Regulation

KW - Reproducing kernel Hilbert space

UR - http://www.scopus.com/inward/record.url?scp=73949116964&partnerID=8YFLogxK

U2 - 10.1214/09-AOS728

DO - 10.1214/09-AOS728

M3 - Article

SN - 0090-5364

VL - 38

SP - 526

EP - 565

JO - Annals of Statistics

JF - Annals of Statistics

IS - 1

ER -

Regularization in kernel learning

Abstract

Access to Document

Other files and links

Fingerprint

Cite this