A boosting method with asymmetric mislabeling probabilities which depend on covariates

Research output: Contribution to journalArticle

3 Citations (Scopus)

Abstract

A new boosting method for a kind of noisy data is developed, where the probability of mislabeling depends on the label of a case. The mechanism of the model is based on a simple idea and gives natural interpretation as a mislabel model. The boosting algorithm is derived from an extension of the exponential loss function, which provides the AdaBoost algorithm. A connection between the proposed method and an asymmetric mislabel model is shown. It is also shown that the loss function proposed constructs a classifier which attains the minimum error rate for a true label. Numerical experiments illustrate how well the proposed method performs in comparison to existing methods.

Original languageEnglish
Pages (from-to)203-218
Number of pages16
JournalComputational Statistics
Volume27
Issue number2
DOIs
Publication statusPublished - 2012 Jun 1
Externally publishedYes

Fingerprint

Boosting
Covariates
Labels
Loss Function
Adaptive boosting
AdaBoost
Noisy Data
Classifiers
Error Rate
Classifier
Numerical Experiment
Model
Experiments
Loss function

Keywords

  • Asymmetric mislabeling mechanism
  • Bayes error rate
  • Boosting method
  • Robustness

ASJC Scopus subject areas

  • Statistics and Probability
  • Statistics, Probability and Uncertainty
  • Computational Mathematics

Cite this

A boosting method with asymmetric mislabeling probabilities which depend on covariates. / Hayashi, Kenichi.

In: Computational Statistics, Vol. 27, No. 2, 01.06.2012, p. 203-218.

Research output: Contribution to journalArticle

@article{72e9332518f54f36b7fa7d1122af2d12,
title = "A boosting method with asymmetric mislabeling probabilities which depend on covariates",
abstract = "A new boosting method for a kind of noisy data is developed, where the probability of mislabeling depends on the label of a case. The mechanism of the model is based on a simple idea and gives natural interpretation as a mislabel model. The boosting algorithm is derived from an extension of the exponential loss function, which provides the AdaBoost algorithm. A connection between the proposed method and an asymmetric mislabel model is shown. It is also shown that the loss function proposed constructs a classifier which attains the minimum error rate for a true label. Numerical experiments illustrate how well the proposed method performs in comparison to existing methods.",
keywords = "Asymmetric mislabeling mechanism, Bayes error rate, Boosting method, Robustness",
author = "Kenichi Hayashi",
year = "2012",
month = "6",
day = "1",
doi = "10.1007/s00180-011-0250-8",
language = "English",
volume = "27",
pages = "203--218",
journal = "Computational Statistics",
issn = "0943-4062",
publisher = "Springer Verlag",
number = "2",

}

TY - JOUR

T1 - A boosting method with asymmetric mislabeling probabilities which depend on covariates

AU - Hayashi, Kenichi

PY - 2012/6/1

Y1 - 2012/6/1

N2 - A new boosting method for a kind of noisy data is developed, where the probability of mislabeling depends on the label of a case. The mechanism of the model is based on a simple idea and gives natural interpretation as a mislabel model. The boosting algorithm is derived from an extension of the exponential loss function, which provides the AdaBoost algorithm. A connection between the proposed method and an asymmetric mislabel model is shown. It is also shown that the loss function proposed constructs a classifier which attains the minimum error rate for a true label. Numerical experiments illustrate how well the proposed method performs in comparison to existing methods.

AB - A new boosting method for a kind of noisy data is developed, where the probability of mislabeling depends on the label of a case. The mechanism of the model is based on a simple idea and gives natural interpretation as a mislabel model. The boosting algorithm is derived from an extension of the exponential loss function, which provides the AdaBoost algorithm. A connection between the proposed method and an asymmetric mislabel model is shown. It is also shown that the loss function proposed constructs a classifier which attains the minimum error rate for a true label. Numerical experiments illustrate how well the proposed method performs in comparison to existing methods.

KW - Asymmetric mislabeling mechanism

KW - Bayes error rate

KW - Boosting method

KW - Robustness

UR - http://www.scopus.com/inward/record.url?scp=84860364936&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84860364936&partnerID=8YFLogxK

U2 - 10.1007/s00180-011-0250-8

DO - 10.1007/s00180-011-0250-8

M3 - Article

AN - SCOPUS:84860364936

VL - 27

SP - 203

EP - 218

JO - Computational Statistics

JF - Computational Statistics

SN - 0943-4062

IS - 2

ER -