Maximal width learning of binary functions

Martin Anthony; Joel Ratsaby

doi:10.1016/j.tcs.2009.09.020

Maximal width learning of binary functions

Martin Anthony, Joel Ratsaby

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

12 Scopus citations

Abstract

This paper concerns learning binary-valued functions defined on R, and investigates how a particular type of 'regularity' of hypotheses can be used to obtain better generalization error bounds. We derive error bounds that depend on the sample width (a notion analogous to that of sample margin for real-valued functions). This motivates learning algorithms that seek to maximize sample width.

Original language	English
Pages (from-to)	138-147
Number of pages	10
Journal	Theoretical Computer Science
Volume	411
Issue number	1
DOIs	https://doi.org/10.1016/j.tcs.2009.09.020
State	Published - 1 Jan 2010
Externally published	Yes

Keywords

Binary function classes
Learning algorithms

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Access to Document

10.1016/j.tcs.2009.09.020

Cite this

@article{b8e9c888602b4e038bbe2d11e400bcbc,

title = "Maximal width learning of binary functions",

abstract = "This paper concerns learning binary-valued functions defined on R, and investigates how a particular type of 'regularity' of hypotheses can be used to obtain better generalization error bounds. We derive error bounds that depend on the sample width (a notion analogous to that of sample margin for real-valued functions). This motivates learning algorithms that seek to maximize sample width.",

keywords = "Binary function classes, Learning algorithms",

author = "Martin Anthony and Joel Ratsaby",

year = "2010",

month = jan,

day = "1",

doi = "10.1016/j.tcs.2009.09.020",

language = "English",

volume = "411",

pages = "138--147",

journal = "Theoretical Computer Science",

issn = "0304-3975",

publisher = "Elsevier B.V.",

number = "1",

}

TY - JOUR

T1 - Maximal width learning of binary functions

AU - Anthony, Martin

AU - Ratsaby, Joel

PY - 2010/1/1

Y1 - 2010/1/1

N2 - This paper concerns learning binary-valued functions defined on R, and investigates how a particular type of 'regularity' of hypotheses can be used to obtain better generalization error bounds. We derive error bounds that depend on the sample width (a notion analogous to that of sample margin for real-valued functions). This motivates learning algorithms that seek to maximize sample width.

AB - This paper concerns learning binary-valued functions defined on R, and investigates how a particular type of 'regularity' of hypotheses can be used to obtain better generalization error bounds. We derive error bounds that depend on the sample width (a notion analogous to that of sample margin for real-valued functions). This motivates learning algorithms that seek to maximize sample width.

KW - Binary function classes

KW - Learning algorithms

UR - http://www.scopus.com/inward/record.url?scp=71749119513&partnerID=8YFLogxK

U2 - 10.1016/j.tcs.2009.09.020

DO - 10.1016/j.tcs.2009.09.020

M3 - Article

AN - SCOPUS:71749119513

SN - 0304-3975

VL - 411

SP - 138

EP - 147

JO - Theoretical Computer Science

JF - Theoretical Computer Science

IS - 1

ER -

Maximal width learning of binary functions

Abstract

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this