An ensemble approach to data fusion and its application to ATR

Costin Barbu, Jing Peng, Richard Sims

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

Abstract

Ensemble methods provide a principled framework in which to build high performance classifiers and represent many types of data. As a result, these methods can be useful for making inferences about biometrie and biological events. 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 represetation. Unlike AdaBoost, however, all data types share the same sampling distribution computed from the base classifier having the smallest error rate among input sources. As a result, the most consistent data type dominates over time, thereby significantly reducing sensitivity to noise. The method is applied to the problem of facial and gender prediction based on biometrie traits. The new method outperforms several competing techniques including kernel-based data fusion, and is provably better than AdaBoost trained on any single type of data.

Original languageEnglish
Title of host publicationAutomatic Target Recognition XVII
DOIs
StatePublished - 15 Nov 2007
EventAutomatic Target Recognition XVII - Orlando, FL, United States
Duration: 10 Apr 200712 Apr 2007

Publication series

NameProceedings of SPIE - The International Society for Optical Engineering
Volume6566
ISSN (Print)0277-786X

Other

OtherAutomatic Target Recognition XVII
CountryUnited States
CityOrlando, FL
Period10/04/0712/04/07

Fingerprint

Adaptive boosting
AdaBoost
multisensor fusion
Data Fusion
Data fusion
classifiers
Ensemble
Ensemble Methods
Classifiers
Classifier
inference
Sampling Distribution
sampling
Error Rate
High Performance
sensitivity
Sampling
kernel
predictions
Prediction

Keywords

  • ATR
  • Adaboost
  • Data fusion
  • Semi-definite programming
  • Stacking

Cite this

Barbu, C., Peng, J., & Sims, R. (2007). An ensemble approach to data fusion and its application to ATR. In Automatic Target Recognition XVII [65660O] (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 6566). https://doi.org/10.1117/12.720156
Barbu, Costin ; Peng, Jing ; Sims, Richard. / An ensemble approach to data fusion and its application to ATR. Automatic Target Recognition XVII. 2007. (Proceedings of SPIE - The International Society for Optical Engineering).
@inproceedings{9852b80907124527b27079c0f48f795f,
title = "An ensemble approach to data fusion and its application to ATR",
abstract = "Ensemble methods provide a principled framework in which to build high performance classifiers and represent many types of data. As a result, these methods can be useful for making inferences about biometrie and biological events. 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 represetation. Unlike AdaBoost, however, all data types share the same sampling distribution computed from the base classifier having the smallest error rate among input sources. As a result, the most consistent data type dominates over time, thereby significantly reducing sensitivity to noise. The method is applied to the problem of facial and gender prediction based on biometrie traits. The new method outperforms several competing techniques including kernel-based data fusion, and is provably better than AdaBoost trained on any single type of data.",
keywords = "ATR, Adaboost, Data fusion, Semi-definite programming, Stacking",
author = "Costin Barbu and Jing Peng and Richard Sims",
year = "2007",
month = "11",
day = "15",
doi = "10.1117/12.720156",
language = "English",
isbn = "0819466883",
series = "Proceedings of SPIE - The International Society for Optical Engineering",
booktitle = "Automatic Target Recognition XVII",

}

Barbu, C, Peng, J & Sims, R 2007, An ensemble approach to data fusion and its application to ATR. in Automatic Target Recognition XVII., 65660O, Proceedings of SPIE - The International Society for Optical Engineering, vol. 6566, Automatic Target Recognition XVII, Orlando, FL, United States, 10/04/07. https://doi.org/10.1117/12.720156

An ensemble approach to data fusion and its application to ATR. / Barbu, Costin; Peng, Jing; Sims, Richard.

Automatic Target Recognition XVII. 2007. 65660O (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 6566).

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

TY - GEN

T1 - An ensemble approach to data fusion and its application to ATR

AU - Barbu, Costin

AU - Peng, Jing

AU - Sims, Richard

PY - 2007/11/15

Y1 - 2007/11/15

N2 - Ensemble methods provide a principled framework in which to build high performance classifiers and represent many types of data. As a result, these methods can be useful for making inferences about biometrie and biological events. 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 represetation. Unlike AdaBoost, however, all data types share the same sampling distribution computed from the base classifier having the smallest error rate among input sources. As a result, the most consistent data type dominates over time, thereby significantly reducing sensitivity to noise. The method is applied to the problem of facial and gender prediction based on biometrie traits. The new method outperforms several competing techniques including kernel-based data fusion, and is provably better than AdaBoost trained on any single type of data.

AB - Ensemble methods provide a principled framework in which to build high performance classifiers and represent many types of data. As a result, these methods can be useful for making inferences about biometrie and biological events. 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 represetation. Unlike AdaBoost, however, all data types share the same sampling distribution computed from the base classifier having the smallest error rate among input sources. As a result, the most consistent data type dominates over time, thereby significantly reducing sensitivity to noise. The method is applied to the problem of facial and gender prediction based on biometrie traits. The new method outperforms several competing techniques including kernel-based data fusion, and is provably better than AdaBoost trained on any single type of data.

KW - ATR

KW - Adaboost

KW - Data fusion

KW - Semi-definite programming

KW - Stacking

UR - http://www.scopus.com/inward/record.url?scp=35948940513&partnerID=8YFLogxK

U2 - 10.1117/12.720156

DO - 10.1117/12.720156

M3 - Conference contribution

AN - SCOPUS:35948940513

SN - 0819466883

SN - 9780819466884

T3 - Proceedings of SPIE - The International Society for Optical Engineering

BT - Automatic Target Recognition XVII

ER -

Barbu C, Peng J, Sims R. An ensemble approach to data fusion and its application to ATR. In Automatic Target Recognition XVII. 2007. 65660O. (Proceedings of SPIE - The International Society for Optical Engineering). https://doi.org/10.1117/12.720156