Robust active learning for linear regression via density power divergence

Yasuhiro Sogawa, Tsuyoshi Ueno, Yoshinobu Kawahara, Takashi Washio

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

The performance of active learning (AL) is crucially influenced by the existence of outliers in input samples. In this paper, we propose a robust pool-based AL measure based on the density power divergence. It is known that the density power divergence can be accurately estimated even under the existence of outliers within data. We further derive an AL scheme based on an asymptotic statistical analysis on the M-estimator. The performance of the proposed framework is investigated empirically using artificial and real-world data.

Original languageEnglish
Title of host publicationNeural Information Processing - 19th International Conference, ICONIP 2012, Proceedings
Pages594-602
Number of pages9
EditionPART 3
DOIs
Publication statusPublished - Nov 19 2012
Externally publishedYes
Event19th International Conference on Neural Information Processing, ICONIP 2012 - Doha, Qatar
Duration: Nov 12 2012Nov 15 2012

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 3
Volume7665 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other19th International Conference on Neural Information Processing, ICONIP 2012
CountryQatar
CityDoha
Period11/12/1211/15/12

Fingerprint

Power Divergence
Active Learning
Linear regression
Outlier
M-estimator
Asymptotic Analysis
Statistical Analysis
Statistical methods
Problem-Based Learning

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Sogawa, Y., Ueno, T., Kawahara, Y., & Washio, T. (2012). Robust active learning for linear regression via density power divergence. In Neural Information Processing - 19th International Conference, ICONIP 2012, Proceedings (PART 3 ed., pp. 594-602). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7665 LNCS, No. PART 3). https://doi.org/10.1007/978-3-642-34487-9_72

Robust active learning for linear regression via density power divergence. / Sogawa, Yasuhiro; Ueno, Tsuyoshi; Kawahara, Yoshinobu; Washio, Takashi.

Neural Information Processing - 19th International Conference, ICONIP 2012, Proceedings. PART 3. ed. 2012. p. 594-602 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7665 LNCS, No. PART 3).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Sogawa, Y, Ueno, T, Kawahara, Y & Washio, T 2012, Robust active learning for linear regression via density power divergence. in Neural Information Processing - 19th International Conference, ICONIP 2012, Proceedings. PART 3 edn, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 3, vol. 7665 LNCS, pp. 594-602, 19th International Conference on Neural Information Processing, ICONIP 2012, Doha, Qatar, 11/12/12. https://doi.org/10.1007/978-3-642-34487-9_72
Sogawa Y, Ueno T, Kawahara Y, Washio T. Robust active learning for linear regression via density power divergence. In Neural Information Processing - 19th International Conference, ICONIP 2012, Proceedings. PART 3 ed. 2012. p. 594-602. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 3). https://doi.org/10.1007/978-3-642-34487-9_72
Sogawa, Yasuhiro ; Ueno, Tsuyoshi ; Kawahara, Yoshinobu ; Washio, Takashi. / Robust active learning for linear regression via density power divergence. Neural Information Processing - 19th International Conference, ICONIP 2012, Proceedings. PART 3. ed. 2012. pp. 594-602 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 3).
@inproceedings{1070048d34b74b879ea3d43be41d2e0f,
title = "Robust active learning for linear regression via density power divergence",
abstract = "The performance of active learning (AL) is crucially influenced by the existence of outliers in input samples. In this paper, we propose a robust pool-based AL measure based on the density power divergence. It is known that the density power divergence can be accurately estimated even under the existence of outliers within data. We further derive an AL scheme based on an asymptotic statistical analysis on the M-estimator. The performance of the proposed framework is investigated empirically using artificial and real-world data.",
author = "Yasuhiro Sogawa and Tsuyoshi Ueno and Yoshinobu Kawahara and Takashi Washio",
year = "2012",
month = "11",
day = "19",
doi = "10.1007/978-3-642-34487-9_72",
language = "English",
isbn = "9783642344862",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
number = "PART 3",
pages = "594--602",
booktitle = "Neural Information Processing - 19th International Conference, ICONIP 2012, Proceedings",
edition = "PART 3",

}

TY - GEN

T1 - Robust active learning for linear regression via density power divergence

AU - Sogawa, Yasuhiro

AU - Ueno, Tsuyoshi

AU - Kawahara, Yoshinobu

AU - Washio, Takashi

PY - 2012/11/19

Y1 - 2012/11/19

N2 - The performance of active learning (AL) is crucially influenced by the existence of outliers in input samples. In this paper, we propose a robust pool-based AL measure based on the density power divergence. It is known that the density power divergence can be accurately estimated even under the existence of outliers within data. We further derive an AL scheme based on an asymptotic statistical analysis on the M-estimator. The performance of the proposed framework is investigated empirically using artificial and real-world data.

AB - The performance of active learning (AL) is crucially influenced by the existence of outliers in input samples. In this paper, we propose a robust pool-based AL measure based on the density power divergence. It is known that the density power divergence can be accurately estimated even under the existence of outliers within data. We further derive an AL scheme based on an asymptotic statistical analysis on the M-estimator. The performance of the proposed framework is investigated empirically using artificial and real-world data.

UR - http://www.scopus.com/inward/record.url?scp=84869053228&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84869053228&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-34487-9_72

DO - 10.1007/978-3-642-34487-9_72

M3 - Conference contribution

AN - SCOPUS:84869053228

SN - 9783642344862

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 594

EP - 602

BT - Neural Information Processing - 19th International Conference, ICONIP 2012, Proceedings

ER -