Weight-based boosting model for cross-domain relevance ranking adaptation

Peng Cai, Wei Gao, Kam Fai Wong, Aoying Zhou

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

2 Citations (Scopus)

Abstract

Adaptation techniques based on importance weighting were shown effective for RankSVM and RankNet, viz., each training instance is assigned a target weight denoting its importance to the target domain and incorporated into loss functions. In this work, we extend RankBoost using importance weighting framework for ranking adaptation. We find it non-trivial to incorporate the target weight into the boosting-based ranking algorithms because it plays a contradictory role against the innate weight of boosting, namely source weight that focuses on adjusting source-domain ranking accuracy. Our experiments show that among three variants, the additive weight-based RankBoost, which dynamically balances the two types of weights, significantly and consistently outperforms the baseline trained directly on the source domain.

Original languageEnglish
Title of host publicationAdvances in Information Retrieval - 33rd European Conference on IR Research, ECIR 2011, Proceedings
PublisherSpringer Verlag
Pages562-567
Number of pages6
ISBN (Print)9783642201608
Publication statusPublished - 1 Jan 2011
Event33rd European Conference on Information Retrieval, ECIR 2011 - Dublin, Ireland
Duration: 18 Apr 201121 Apr 2011

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume6611 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other33rd European Conference on Information Retrieval, ECIR 2011
CountryIreland
CityDublin
Period18/4/1121/4/11

Fingerprint

Boosting
Ranking
Experiments
Target
Weighting
Model
Loss Function
Relevance
Baseline
Experiment

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Cai, P., Gao, W., Wong, K. F., & Zhou, A. (2011). Weight-based boosting model for cross-domain relevance ranking adaptation. In Advances in Information Retrieval - 33rd European Conference on IR Research, ECIR 2011, Proceedings (pp. 562-567). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6611 LNCS). Springer Verlag.

Weight-based boosting model for cross-domain relevance ranking adaptation. / Cai, Peng; Gao, Wei; Wong, Kam Fai; Zhou, Aoying.

Advances in Information Retrieval - 33rd European Conference on IR Research, ECIR 2011, Proceedings. Springer Verlag, 2011. p. 562-567 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6611 LNCS).

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

Cai, P, Gao, W, Wong, KF & Zhou, A 2011, Weight-based boosting model for cross-domain relevance ranking adaptation. in Advances in Information Retrieval - 33rd European Conference on IR Research, ECIR 2011, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 6611 LNCS, Springer Verlag, pp. 562-567, 33rd European Conference on Information Retrieval, ECIR 2011, Dublin, Ireland, 18/4/11.
Cai P, Gao W, Wong KF, Zhou A. Weight-based boosting model for cross-domain relevance ranking adaptation. In Advances in Information Retrieval - 33rd European Conference on IR Research, ECIR 2011, Proceedings. Springer Verlag. 2011. p. 562-567. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
Cai, Peng ; Gao, Wei ; Wong, Kam Fai ; Zhou, Aoying. / Weight-based boosting model for cross-domain relevance ranking adaptation. Advances in Information Retrieval - 33rd European Conference on IR Research, ECIR 2011, Proceedings. Springer Verlag, 2011. pp. 562-567 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{6f6bffb5e24e498f96bba8dface9743b,
title = "Weight-based boosting model for cross-domain relevance ranking adaptation",
abstract = "Adaptation techniques based on importance weighting were shown effective for RankSVM and RankNet, viz., each training instance is assigned a target weight denoting its importance to the target domain and incorporated into loss functions. In this work, we extend RankBoost using importance weighting framework for ranking adaptation. We find it non-trivial to incorporate the target weight into the boosting-based ranking algorithms because it plays a contradictory role against the innate weight of boosting, namely source weight that focuses on adjusting source-domain ranking accuracy. Our experiments show that among three variants, the additive weight-based RankBoost, which dynamically balances the two types of weights, significantly and consistently outperforms the baseline trained directly on the source domain.",
author = "Peng Cai and Wei Gao and Wong, {Kam Fai} and Aoying Zhou",
year = "2011",
month = "1",
day = "1",
language = "English",
isbn = "9783642201608",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "562--567",
booktitle = "Advances in Information Retrieval - 33rd European Conference on IR Research, ECIR 2011, Proceedings",

}

TY - GEN

T1 - Weight-based boosting model for cross-domain relevance ranking adaptation

AU - Cai, Peng

AU - Gao, Wei

AU - Wong, Kam Fai

AU - Zhou, Aoying

PY - 2011/1/1

Y1 - 2011/1/1

N2 - Adaptation techniques based on importance weighting were shown effective for RankSVM and RankNet, viz., each training instance is assigned a target weight denoting its importance to the target domain and incorporated into loss functions. In this work, we extend RankBoost using importance weighting framework for ranking adaptation. We find it non-trivial to incorporate the target weight into the boosting-based ranking algorithms because it plays a contradictory role against the innate weight of boosting, namely source weight that focuses on adjusting source-domain ranking accuracy. Our experiments show that among three variants, the additive weight-based RankBoost, which dynamically balances the two types of weights, significantly and consistently outperforms the baseline trained directly on the source domain.

AB - Adaptation techniques based on importance weighting were shown effective for RankSVM and RankNet, viz., each training instance is assigned a target weight denoting its importance to the target domain and incorporated into loss functions. In this work, we extend RankBoost using importance weighting framework for ranking adaptation. We find it non-trivial to incorporate the target weight into the boosting-based ranking algorithms because it plays a contradictory role against the innate weight of boosting, namely source weight that focuses on adjusting source-domain ranking accuracy. Our experiments show that among three variants, the additive weight-based RankBoost, which dynamically balances the two types of weights, significantly and consistently outperforms the baseline trained directly on the source domain.

UR - http://www.scopus.com/inward/record.url?scp=84996495121&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84996495121&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84996495121

SN - 9783642201608

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

SP - 562

EP - 567

BT - Advances in Information Retrieval - 33rd European Conference on IR Research, ECIR 2011, Proceedings

PB - Springer Verlag

ER -