Online regression with varying Gaussians and non-identical distributions

Ting Hu

doi:10.1142/S0219530511001923

Online regression with varying Gaussians and non-identical distributions

Ting Hu

Wuhan University

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

18 Scopus citations

Abstract

We consider a fully online regression algorithm associated with a general convex loss function and Gaussian kernels with changing variances. Error analysis is conducted in a setting with samples drawn from a non-identical sequence of probability measures. When a fixed Gaussian is used, it was known that the learning ability of induced algorithms is weak. By allowing varying Gaussians, we show that the achieved learning rates can be of polynomial decays.

Original language	English
Pages (from-to)	395-408
Number of pages	14
Journal	Analysis and Applications
Volume	9
Issue number	4
DOIs	https://doi.org/10.1142/S0219530511001923
State	Published - Oct 2011
Externally published	Yes

Keywords

Gaussian kernel
convex loss function
online learning
regression algorithm
reproducing kernel Hilbert space
variance of Gaussian

Access to Document

10.1142/S0219530511001923

Cite this

@article{1575241b93504f86be75004be364b4e7,

title = "Online regression with varying Gaussians and non-identical distributions",

abstract = "We consider a fully online regression algorithm associated with a general convex loss function and Gaussian kernels with changing variances. Error analysis is conducted in a setting with samples drawn from a non-identical sequence of probability measures. When a fixed Gaussian is used, it was known that the learning ability of induced algorithms is weak. By allowing varying Gaussians, we show that the achieved learning rates can be of polynomial decays.",

keywords = "Gaussian kernel, convex loss function, online learning, regression algorithm, reproducing kernel Hilbert space, variance of Gaussian",

author = "Ting Hu",

year = "2011",

month = oct,

doi = "10.1142/S0219530511001923",

language = "英语",

volume = "9",

pages = "395--408",

journal = "Analysis and Applications",

issn = "0219-5305",

publisher = "World Scientific",

number = "4",

}

TY - JOUR

T1 - Online regression with varying Gaussians and non-identical distributions

AU - Hu, Ting

PY - 2011/10

Y1 - 2011/10

N2 - We consider a fully online regression algorithm associated with a general convex loss function and Gaussian kernels with changing variances. Error analysis is conducted in a setting with samples drawn from a non-identical sequence of probability measures. When a fixed Gaussian is used, it was known that the learning ability of induced algorithms is weak. By allowing varying Gaussians, we show that the achieved learning rates can be of polynomial decays.

AB - We consider a fully online regression algorithm associated with a general convex loss function and Gaussian kernels with changing variances. Error analysis is conducted in a setting with samples drawn from a non-identical sequence of probability measures. When a fixed Gaussian is used, it was known that the learning ability of induced algorithms is weak. By allowing varying Gaussians, we show that the achieved learning rates can be of polynomial decays.

KW - Gaussian kernel

KW - convex loss function

KW - online learning

KW - regression algorithm

KW - reproducing kernel Hilbert space

KW - variance of Gaussian

UR - https://www.scopus.com/pages/publications/79960409363

U2 - 10.1142/S0219530511001923

DO - 10.1142/S0219530511001923

M3 - 文章

AN - SCOPUS:79960409363

SN - 0219-5305

VL - 9

SP - 395

EP - 408

JO - Analysis and Applications

JF - Analysis and Applications

IS - 4

ER -

Online regression with varying Gaussians and non-identical distributions

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this