Some results on weakly accurate base learners for boosting regression and classification

Wenxin Jiang

doi:10.1007/3-540-45014-9_8

Some results on weakly accurate base learners for boosting regression and classification

Wenxin Jiang^*

^*Corresponding author for this work

Statistics and Data Science

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

4 Scopus citations

Abstract

One basic property of the boosting algorithmis its ability to reduce the training error, subject to the critical assumption that the base learners generate 'weak' (or more appropriately, 'weakly accurate') hypotheses that are better that randomguessing. We exploit analogies between regression and classification to give a characterization on what base learners generate weak hypotheses, by introducing a geometric concept called the angular span for the base hypothesis space. The exponential convergence rates of boosting algorithms are shown to be bounded below by essentially the angular spans. Sufficient conditions for nonzero angular span are also given and validated for a wide class of regression and classification systems.

Original language	English (US)
Title of host publication	Multiple Classifier Systems - First International Workshop, MCS 2000, Proceedings
Editors	Josef Kittler, Fabio Roli
Publisher	Springer Verlag
Pages	87-96
Number of pages	10
ISBN (Print)	3540677046, 9783540677048
DOIs	https://doi.org/10.1007/3-540-45014-9_8
State	Published - 2000
Event	1st International Workshop on Multiple Classifier Systems, MCS 2000 - Cagliari, Italy Duration: Jun 21 2000 → Jun 23 2000

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	1857 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	1st International Workshop on Multiple Classifier Systems, MCS 2000
Country/Territory	Italy
City	Cagliari
Period	6/21/00 → 6/23/00

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Access to Document

10.1007/3-540-45014-9_8

Cite this

Jiang, W. (2000). Some results on weakly accurate base learners for boosting regression and classification. In J. Kittler, & F. Roli (Eds.), Multiple Classifier Systems - First International Workshop, MCS 2000, Proceedings (pp. 87-96). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1857 LNCS). Springer Verlag. https://doi.org/10.1007/3-540-45014-9_8

Jiang, Wenxin. / Some results on weakly accurate base learners for boosting regression and classification. Multiple Classifier Systems - First International Workshop, MCS 2000, Proceedings. editor / Josef Kittler ; Fabio Roli. Springer Verlag, 2000. pp. 87-96 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{4a074598c521493bbe96ce44ad391420,

title = "Some results on weakly accurate base learners for boosting regression and classification",

abstract = "One basic property of the boosting algorithmis its ability to reduce the training error, subject to the critical assumption that the base learners generate 'weak' (or more appropriately, 'weakly accurate') hypotheses that are better that randomguessing. We exploit analogies between regression and classification to give a characterization on what base learners generate weak hypotheses, by introducing a geometric concept called the angular span for the base hypothesis space. The exponential convergence rates of boosting algorithms are shown to be bounded below by essentially the angular spans. Sufficient conditions for nonzero angular span are also given and validated for a wide class of regression and classification systems.",

author = "Wenxin Jiang",

year = "2000",

doi = "10.1007/3-540-45014-9_8",

language = "English (US)",

isbn = "3540677046",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "87--96",

editor = "Josef Kittler and Fabio Roli",

booktitle = "Multiple Classifier Systems - First International Workshop, MCS 2000, Proceedings",

note = "1st International Workshop on Multiple Classifier Systems, MCS 2000 ; Conference date: 21-06-2000 Through 23-06-2000",

}

Jiang, W 2000, Some results on weakly accurate base learners for boosting regression and classification. in J Kittler & F Roli (eds), Multiple Classifier Systems - First International Workshop, MCS 2000, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 1857 LNCS, Springer Verlag, pp. 87-96, 1st International Workshop on Multiple Classifier Systems, MCS 2000, Cagliari, Italy, 6/21/00. https://doi.org/10.1007/3-540-45014-9_8

Some results on weakly accurate base learners for boosting regression and classification. / Jiang, Wenxin.
Multiple Classifier Systems - First International Workshop, MCS 2000, Proceedings. ed. / Josef Kittler; Fabio Roli. Springer Verlag, 2000. p. 87-96 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1857 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Some results on weakly accurate base learners for boosting regression and classification

AU - Jiang, Wenxin

PY - 2000

Y1 - 2000

N2 - One basic property of the boosting algorithmis its ability to reduce the training error, subject to the critical assumption that the base learners generate 'weak' (or more appropriately, 'weakly accurate') hypotheses that are better that randomguessing. We exploit analogies between regression and classification to give a characterization on what base learners generate weak hypotheses, by introducing a geometric concept called the angular span for the base hypothesis space. The exponential convergence rates of boosting algorithms are shown to be bounded below by essentially the angular spans. Sufficient conditions for nonzero angular span are also given and validated for a wide class of regression and classification systems.

AB - One basic property of the boosting algorithmis its ability to reduce the training error, subject to the critical assumption that the base learners generate 'weak' (or more appropriately, 'weakly accurate') hypotheses that are better that randomguessing. We exploit analogies between regression and classification to give a characterization on what base learners generate weak hypotheses, by introducing a geometric concept called the angular span for the base hypothesis space. The exponential convergence rates of boosting algorithms are shown to be bounded below by essentially the angular spans. Sufficient conditions for nonzero angular span are also given and validated for a wide class of regression and classification systems.

UR - http://www.scopus.com/inward/record.url?scp=84867069706&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84867069706&partnerID=8YFLogxK

U2 - 10.1007/3-540-45014-9_8

DO - 10.1007/3-540-45014-9_8

M3 - Conference contribution

AN - SCOPUS:84867069706

SN - 3540677046

SN - 9783540677048

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

SP - 87

EP - 96

BT - Multiple Classifier Systems - First International Workshop, MCS 2000, Proceedings

A2 - Kittler, Josef

A2 - Roli, Fabio

PB - Springer Verlag

T2 - 1st International Workshop on Multiple Classifier Systems, MCS 2000

Y2 - 21 June 2000 through 23 June 2000

ER -

Jiang W. Some results on weakly accurate base learners for boosting regression and classification. In Kittler J, Roli F, editors, Multiple Classifier Systems - First International Workshop, MCS 2000, Proceedings. Springer Verlag. 2000. p. 87-96. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/3-540-45014-9_8

Some results on weakly accurate base learners for boosting regression and classification

Abstract

Publication series

Other

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this