Boosting information fusion

Costin Barbu, Jing Peng, Guna Seetharaman

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

1 Citation (Scopus)

Abstract

Ensemble methods provide a principled framework for building high performance classifiers and representing many types of data. As a result, these methods can be useful for making inferences in many domains such as classification and multi-modal biometrics. We introduce a novel ensemble method for combining multiple representations (or views). The method is a multiple view generalization of AdaBoost. Similar to AdaBoost, base classifiers are independently built from each representation. Unlike AdaBoost, however, all data types share the same sampling distribution as the view whose weighted training error is the smallest among all the views. As a result, the most consistent data type dominates over time, thereby significantly reducing sensitivity to noise. In addition, our proposal is provably better than AdaBoost trained on any single type of data. The proposed method is applied to the problems of facial and gender prediction based on biometric traits as well as of protein classification. Experimental results show that our method outperforms several competing techniques including kernel-based data fusion.

Original languageEnglish
Title of host publication13th Conference on Information Fusion, Fusion 2010
StatePublished - 1 Dec 2010
Event13th Conference on Information Fusion, Fusion 2010 - Edinburgh, United Kingdom
Duration: 26 Jul 201029 Jul 2010

Publication series

Name13th Conference on Information Fusion, Fusion 2010

Other

Other13th Conference on Information Fusion, Fusion 2010
CountryUnited Kingdom
CityEdinburgh
Period26/07/1029/07/10

Fingerprint

Adaptive boosting
Information fusion
Biometrics
Classifiers
Data fusion
Sampling
Proteins

Keywords

  • AdaBoost
  • Data fusion
  • Semi-definite programming
  • Stacking

Cite this

Barbu, C., Peng, J., & Seetharaman, G. (2010). Boosting information fusion. In 13th Conference on Information Fusion, Fusion 2010 [5711976] (13th Conference on Information Fusion, Fusion 2010).
Barbu, Costin ; Peng, Jing ; Seetharaman, Guna. / Boosting information fusion. 13th Conference on Information Fusion, Fusion 2010. 2010. (13th Conference on Information Fusion, Fusion 2010).
@inproceedings{f4535787cdbb41798da9f7f89932123c,
title = "Boosting information fusion",
abstract = "Ensemble methods provide a principled framework for building high performance classifiers and representing many types of data. As a result, these methods can be useful for making inferences in many domains such as classification and multi-modal biometrics. We introduce a novel ensemble method for combining multiple representations (or views). The method is a multiple view generalization of AdaBoost. Similar to AdaBoost, base classifiers are independently built from each representation. Unlike AdaBoost, however, all data types share the same sampling distribution as the view whose weighted training error is the smallest among all the views. As a result, the most consistent data type dominates over time, thereby significantly reducing sensitivity to noise. In addition, our proposal is provably better than AdaBoost trained on any single type of data. The proposed method is applied to the problems of facial and gender prediction based on biometric traits as well as of protein classification. Experimental results show that our method outperforms several competing techniques including kernel-based data fusion.",
keywords = "AdaBoost, Data fusion, Semi-definite programming, Stacking",
author = "Costin Barbu and Jing Peng and Guna Seetharaman",
year = "2010",
month = "12",
day = "1",
language = "English",
isbn = "9780982443811",
series = "13th Conference on Information Fusion, Fusion 2010",
booktitle = "13th Conference on Information Fusion, Fusion 2010",

}

Barbu, C, Peng, J & Seetharaman, G 2010, Boosting information fusion. in 13th Conference on Information Fusion, Fusion 2010., 5711976, 13th Conference on Information Fusion, Fusion 2010, 13th Conference on Information Fusion, Fusion 2010, Edinburgh, United Kingdom, 26/07/10.

Boosting information fusion. / Barbu, Costin; Peng, Jing; Seetharaman, Guna.

13th Conference on Information Fusion, Fusion 2010. 2010. 5711976 (13th Conference on Information Fusion, Fusion 2010).

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

TY - GEN

T1 - Boosting information fusion

AU - Barbu, Costin

AU - Peng, Jing

AU - Seetharaman, Guna

PY - 2010/12/1

Y1 - 2010/12/1

N2 - Ensemble methods provide a principled framework for building high performance classifiers and representing many types of data. As a result, these methods can be useful for making inferences in many domains such as classification and multi-modal biometrics. We introduce a novel ensemble method for combining multiple representations (or views). The method is a multiple view generalization of AdaBoost. Similar to AdaBoost, base classifiers are independently built from each representation. Unlike AdaBoost, however, all data types share the same sampling distribution as the view whose weighted training error is the smallest among all the views. As a result, the most consistent data type dominates over time, thereby significantly reducing sensitivity to noise. In addition, our proposal is provably better than AdaBoost trained on any single type of data. The proposed method is applied to the problems of facial and gender prediction based on biometric traits as well as of protein classification. Experimental results show that our method outperforms several competing techniques including kernel-based data fusion.

AB - Ensemble methods provide a principled framework for building high performance classifiers and representing many types of data. As a result, these methods can be useful for making inferences in many domains such as classification and multi-modal biometrics. We introduce a novel ensemble method for combining multiple representations (or views). The method is a multiple view generalization of AdaBoost. Similar to AdaBoost, base classifiers are independently built from each representation. Unlike AdaBoost, however, all data types share the same sampling distribution as the view whose weighted training error is the smallest among all the views. As a result, the most consistent data type dominates over time, thereby significantly reducing sensitivity to noise. In addition, our proposal is provably better than AdaBoost trained on any single type of data. The proposed method is applied to the problems of facial and gender prediction based on biometric traits as well as of protein classification. Experimental results show that our method outperforms several competing techniques including kernel-based data fusion.

KW - AdaBoost

KW - Data fusion

KW - Semi-definite programming

KW - Stacking

UR - http://www.scopus.com/inward/record.url?scp=79952426333&partnerID=8YFLogxK

M3 - Conference contribution

SN - 9780982443811

T3 - 13th Conference on Information Fusion, Fusion 2010

BT - 13th Conference on Information Fusion, Fusion 2010

ER -

Barbu C, Peng J, Seetharaman G. Boosting information fusion. In 13th Conference on Information Fusion, Fusion 2010. 2010. 5711976. (13th Conference on Information Fusion, Fusion 2010).