A note on extending generalization bounds for binary large-margin classifiers to multiple classes

Ürün Dogan; Tobias Glasmachers; Christian Igel

doi:10.1007/978-3-642-33460-3_13

A note on extending generalization bounds for binary large-margin classifiers to multiple classes

Ürün Dogan, Tobias Glasmachers, Christian Igel

Department of Computer Science

2 Citations (Scopus)

Abstract

A generic way to extend generalization bounds for binary large-margin classifiers to large-margin multi-category classifiers is presented. The simple proceeding leads to surprisingly tight bounds showing the same Õ(d ²) scaling in the number d of classes as state-of-the-art results. The approach is exemplified by extending a textbook bound based on Rademacher complexity, which leads to a multi-class bound depending on the sum of the margin violations of the classifier.

Original language	English
Title of host publication	Machine learning and knowledge discovery in databases : European Conference, ECML PKDD 2012, Bristol, UK, September 24-28, 2012. Proceedings, Part I
Editors	Peter A. Flach, Tijl De Bie, Nello Cristianini
Number of pages	8
Publisher	Springer
Publication date	2012
Pages	122-129
ISBN (Print)	978-3-642-33459-7
ISBN (Electronic)	978-3-642-33460-3
DOIs	https://doi.org/10.1007/978-3-642-33460-3_13
Publication status	Published - 2012

Series	Lecture notes in computer science
Volume	7523
ISSN	0302-9743

Access to Document

10.1007/978-3-642-33460-3_13

https://link.springer.com/content/pdf/10.1007%2F978-3-642-33460-3_13.pdf

Cite this

Dogan, Ü., Glasmachers, T., & Igel, C. (2012). A note on extending generalization bounds for binary large-margin classifiers to multiple classes. In P. A. Flach, T. D. Bie, & N. Cristianini (Eds.), Machine learning and knowledge discovery in databases: European Conference, ECML PKDD 2012, Bristol, UK, September 24-28, 2012. Proceedings, Part I (pp. 122-129). Springer. Lecture notes in computer science Vol. 7523 https://doi.org/10.1007/978-3-642-33460-3_13

A note on extending generalization bounds for binary large-margin classifiers to multiple classes. / Dogan, Ürün; Glasmachers, Tobias; Igel, Christian.

Machine learning and knowledge discovery in databases: European Conference, ECML PKDD 2012, Bristol, UK, September 24-28, 2012. Proceedings, Part I. ed. / Peter A. Flach; Tijl De Bie; Nello Cristianini. Springer, 2012. p. 122-129 (Lecture notes in computer science, Vol. 7523).

Research output: Chapter in Book/Report/Conference proceeding › Book chapter › Research › peer-review

Dogan, Ü, Glasmachers, T & Igel, C 2012, A note on extending generalization bounds for binary large-margin classifiers to multiple classes. in PA Flach, TD Bie & N Cristianini (eds), Machine learning and knowledge discovery in databases: European Conference, ECML PKDD 2012, Bristol, UK, September 24-28, 2012. Proceedings, Part I. Springer, Lecture notes in computer science, vol. 7523, pp. 122-129. https://doi.org/10.1007/978-3-642-33460-3_13

Dogan Ü, Glasmachers T, Igel C. A note on extending generalization bounds for binary large-margin classifiers to multiple classes. In Flach PA, Bie TD, Cristianini N, editors, Machine learning and knowledge discovery in databases: European Conference, ECML PKDD 2012, Bristol, UK, September 24-28, 2012. Proceedings, Part I. Springer. 2012. p. 122-129. (Lecture notes in computer science, Vol. 7523). doi: 10.1007/978-3-642-33460-3_13

Dogan, Ürün ; Glasmachers, Tobias ; Igel, Christian. / A note on extending generalization bounds for binary large-margin classifiers to multiple classes. Machine learning and knowledge discovery in databases: European Conference, ECML PKDD 2012, Bristol, UK, September 24-28, 2012. Proceedings, Part I. editor / Peter A. Flach ; Tijl De Bie ; Nello Cristianini. Springer, 2012. pp. 122-129 (Lecture notes in computer science, Vol. 7523).

@inbook{d295d6adeb3045c0a62c471494eca29e,

title = "A note on extending generalization bounds for binary large-margin classifiers to multiple classes",

abstract = "A generic way to extend generalization bounds for binary large-margin classifiers to large-margin multi-category classifiers is presented. The simple proceeding leads to surprisingly tight bounds showing the same {\~O}(d 2) scaling in the number d of classes as state-of-the-art results. The approach is exemplified by extending a textbook bound based on Rademacher complexity, which leads to a multi-class bound depending on the sum of the margin violations of the classifier.",

author = "{\"U}r{\"u}n Dogan and Tobias Glasmachers and Christian Igel",

year = "2012",

doi = "10.1007/978-3-642-33460-3_13",

language = "English",

isbn = "978-3-642-33459-7",

series = "Lecture notes in computer science",

publisher = "Springer",

pages = "122--129",

editor = "Flach, {Peter A.} and Bie, {Tijl De} and Nello Cristianini",

booktitle = "Machine learning and knowledge discovery in databases",

}

TY - CHAP

T1 - A note on extending generalization bounds for binary large-margin classifiers to multiple classes

AU - Dogan, Ürün

AU - Glasmachers, Tobias

AU - Igel, Christian

PY - 2012

Y1 - 2012

N2 - A generic way to extend generalization bounds for binary large-margin classifiers to large-margin multi-category classifiers is presented. The simple proceeding leads to surprisingly tight bounds showing the same Õ(d 2) scaling in the number d of classes as state-of-the-art results. The approach is exemplified by extending a textbook bound based on Rademacher complexity, which leads to a multi-class bound depending on the sum of the margin violations of the classifier.

AB - A generic way to extend generalization bounds for binary large-margin classifiers to large-margin multi-category classifiers is presented. The simple proceeding leads to surprisingly tight bounds showing the same Õ(d 2) scaling in the number d of classes as state-of-the-art results. The approach is exemplified by extending a textbook bound based on Rademacher complexity, which leads to a multi-class bound depending on the sum of the margin violations of the classifier.

U2 - 10.1007/978-3-642-33460-3_13

DO - 10.1007/978-3-642-33460-3_13

M3 - Book chapter

SN - 978-3-642-33459-7

T3 - Lecture notes in computer science

SP - 122

EP - 129

BT - Machine learning and knowledge discovery in databases

A2 - Flach, Peter A.

A2 - Bie, Tijl De

A2 - Cristianini, Nello

PB - Springer

ER -

A note on extending generalization bounds for binary large-margin classifiers to multiple classes

Abstract

Access to Document

Fingerprint

Cite this