Cost-sensitive learning based on bregman divergences

Raúl Santos-Rodríguez; Alicia Guerrero-Curieses; Rocío Alaiz-Rodríguez; Jesús Cid-Sueiro

doi:10.1007/s10994-009-5132-8

Cost-sensitive learning based on bregman divergences

Raúl Santos-Rodríguez
, Alicia Guerrero-Curieses
, Rocío Alaiz-Rodríguez
, Jesús Cid-Sueiro^*

^*Corresponding author for this work

Research output: Contribution to journal › Article (Academic Journal) › peer-review

16 Citations (Scopus)

Abstract

This paper analyzes the application of a particular class of Bregman divergences to design cost-sensitive classifiers for multiclass problems. We show that these divergence measures can be used to estimate posterior probabilities with maximal accuracy for the probability values that are close to the decision boundaries. Asymptotically, the proposed divergence measures provide classifiers minimizing the sum of decision costs in non-separable problems, and maximizing a margin in separable MAP problems.

Original language	English
Pages (from-to)	271-285
Number of pages	15
Journal	Machine Learning
Volume	76
Issue number	2-3
DOIs	https://doi.org/10.1007/s10994-009-5132-8
Publication status	Published - Sept 2009

Bibliographical note

Funding Information:
Acknowledgements This work was partially funded by project TEC2008-01348 from the Spanish Ministry of Science and Innovation.

Copyright:
Copyright 2009 Elsevier B.V., All rights reserved.

Keywords

Bregman divergence
Cost sensitive learning
Maximum margin
Posterior class probabilities

Access to Document

10.1007/s10994-009-5132-8

Handle.net

Persistent link

Cite this

@article{d2b47ca5c41646caa14cd90d89b570be,

title = "Cost-sensitive learning based on bregman divergences",

abstract = "This paper analyzes the application of a particular class of Bregman divergences to design cost-sensitive classifiers for multiclass problems. We show that these divergence measures can be used to estimate posterior probabilities with maximal accuracy for the probability values that are close to the decision boundaries. Asymptotically, the proposed divergence measures provide classifiers minimizing the sum of decision costs in non-separable problems, and maximizing a margin in separable MAP problems.",

keywords = "Bregman divergence, Cost sensitive learning, Maximum margin, Posterior class probabilities",

author = "Ra{\'u}l Santos-Rodr{\'i}guez and Alicia Guerrero-Curieses and Roc{\'i}o Alaiz-Rodr{\'i}guez and Jes{\'u}s Cid-Sueiro",

note = "Funding Information: Acknowledgements This work was partially funded by project TEC2008-01348 from the Spanish Ministry of Science and Innovation. Copyright: Copyright 2009 Elsevier B.V., All rights reserved.",

year = "2009",

month = sep,

doi = "10.1007/s10994-009-5132-8",

language = "English",

volume = "76",

pages = "271--285",

journal = "Machine Learning",

issn = "0885-6125",

publisher = "Springer Netherlands",

number = "2-3",

}

TY - JOUR

T1 - Cost-sensitive learning based on bregman divergences

AU - Santos-Rodríguez, Raúl

AU - Guerrero-Curieses, Alicia

AU - Alaiz-Rodríguez, Rocío

AU - Cid-Sueiro, Jesús

N1 - Funding Information: Acknowledgements This work was partially funded by project TEC2008-01348 from the Spanish Ministry of Science and Innovation. Copyright: Copyright 2009 Elsevier B.V., All rights reserved.

PY - 2009/9

Y1 - 2009/9

N2 - This paper analyzes the application of a particular class of Bregman divergences to design cost-sensitive classifiers for multiclass problems. We show that these divergence measures can be used to estimate posterior probabilities with maximal accuracy for the probability values that are close to the decision boundaries. Asymptotically, the proposed divergence measures provide classifiers minimizing the sum of decision costs in non-separable problems, and maximizing a margin in separable MAP problems.

AB - This paper analyzes the application of a particular class of Bregman divergences to design cost-sensitive classifiers for multiclass problems. We show that these divergence measures can be used to estimate posterior probabilities with maximal accuracy for the probability values that are close to the decision boundaries. Asymptotically, the proposed divergence measures provide classifiers minimizing the sum of decision costs in non-separable problems, and maximizing a margin in separable MAP problems.

KW - Bregman divergence

KW - Cost sensitive learning

KW - Maximum margin

KW - Posterior class probabilities

UR - https://www.scopus.com/pages/publications/68949096588

U2 - 10.1007/s10994-009-5132-8

DO - 10.1007/s10994-009-5132-8

M3 - Article (Academic Journal)

AN - SCOPUS:68949096588

SN - 0885-6125

VL - 76

SP - 271

EP - 285

JO - Machine Learning

JF - Machine Learning

IS - 2-3

ER -

Cost-sensitive learning based on bregman divergences

Abstract

Bibliographical note

Keywords

Access to Document

Handle.net

Other files and links

Fingerprint

Cite this