Class-specific ensembles for active learning in digital imagery

Amit Mandvikar, Huan Liu

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

Abstract

In many real-world tasks of image classification, limited amounts of labeled data are available to train automatic classifiers. Consequently, extensive human expert involvement is required for instance labeling. Detecting Egeria densa in digital imagery is one such real-world classification task. It presents an additional challenge due to subtle spectral changes in Egeria, which makes it difficult to find a single accurate classifier. A novel solution is proposed to employ an ensemble of classifiers for each class (class-specific ensembles), combined with an active learning scheme. The class-specific ensembles are implicitly diverse. Diversity is required to increase the overall accuracy when combining predictions. The combined predictions of the ensembles can be used to reduce the uncertainty in detecting Egeria. Iterative active learning is then suggested to adapt the ensembles to the new images, unseen to the active learner. A novel solution to build compact ensembles is also presented, which are needed to expedite the re-training of the active learner. The combined results are accurate and compact ensembles, which require significantly less expert involvement for image region classification.

Original languageEnglish (US)
Title of host publicationSIAM Proceedings Series
EditorsM.W. Berry, U. Dayal, C. Kamath, D. Skillicorn
Pages412-421
Number of pages10
StatePublished - 2004
EventProceedings of the Fourth SIAM International Conference on Data Mining - Lake Buena Vista, FL, United States
Duration: Apr 22 2004Apr 24 2004

Other

OtherProceedings of the Fourth SIAM International Conference on Data Mining
CountryUnited States
CityLake Buena Vista, FL
Period4/22/044/24/04

Fingerprint

Active Learning
Ensemble
Classifier
Prediction
Image Classification
Class
Imagery
Labeling
Uncertainty

ASJC Scopus subject areas

  • Mathematics(all)

Cite this

Mandvikar, A., & Liu, H. (2004). Class-specific ensembles for active learning in digital imagery. In M. W. Berry, U. Dayal, C. Kamath, & D. Skillicorn (Eds.), SIAM Proceedings Series (pp. 412-421)

Class-specific ensembles for active learning in digital imagery. / Mandvikar, Amit; Liu, Huan.

SIAM Proceedings Series. ed. / M.W. Berry; U. Dayal; C. Kamath; D. Skillicorn. 2004. p. 412-421.

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

Mandvikar, A & Liu, H 2004, Class-specific ensembles for active learning in digital imagery. in MW Berry, U Dayal, C Kamath & D Skillicorn (eds), SIAM Proceedings Series. pp. 412-421, Proceedings of the Fourth SIAM International Conference on Data Mining, Lake Buena Vista, FL, United States, 4/22/04.
Mandvikar A, Liu H. Class-specific ensembles for active learning in digital imagery. In Berry MW, Dayal U, Kamath C, Skillicorn D, editors, SIAM Proceedings Series. 2004. p. 412-421
Mandvikar, Amit ; Liu, Huan. / Class-specific ensembles for active learning in digital imagery. SIAM Proceedings Series. editor / M.W. Berry ; U. Dayal ; C. Kamath ; D. Skillicorn. 2004. pp. 412-421
@inproceedings{966195d430b84bd1ab93913f02386ec4,
title = "Class-specific ensembles for active learning in digital imagery",
abstract = "In many real-world tasks of image classification, limited amounts of labeled data are available to train automatic classifiers. Consequently, extensive human expert involvement is required for instance labeling. Detecting Egeria densa in digital imagery is one such real-world classification task. It presents an additional challenge due to subtle spectral changes in Egeria, which makes it difficult to find a single accurate classifier. A novel solution is proposed to employ an ensemble of classifiers for each class (class-specific ensembles), combined with an active learning scheme. The class-specific ensembles are implicitly diverse. Diversity is required to increase the overall accuracy when combining predictions. The combined predictions of the ensembles can be used to reduce the uncertainty in detecting Egeria. Iterative active learning is then suggested to adapt the ensembles to the new images, unseen to the active learner. A novel solution to build compact ensembles is also presented, which are needed to expedite the re-training of the active learner. The combined results are accurate and compact ensembles, which require significantly less expert involvement for image region classification.",
author = "Amit Mandvikar and Huan Liu",
year = "2004",
language = "English (US)",
pages = "412--421",
editor = "M.W. Berry and U. Dayal and C. Kamath and D. Skillicorn",
booktitle = "SIAM Proceedings Series",

}

TY - GEN

T1 - Class-specific ensembles for active learning in digital imagery

AU - Mandvikar, Amit

AU - Liu, Huan

PY - 2004

Y1 - 2004

N2 - In many real-world tasks of image classification, limited amounts of labeled data are available to train automatic classifiers. Consequently, extensive human expert involvement is required for instance labeling. Detecting Egeria densa in digital imagery is one such real-world classification task. It presents an additional challenge due to subtle spectral changes in Egeria, which makes it difficult to find a single accurate classifier. A novel solution is proposed to employ an ensemble of classifiers for each class (class-specific ensembles), combined with an active learning scheme. The class-specific ensembles are implicitly diverse. Diversity is required to increase the overall accuracy when combining predictions. The combined predictions of the ensembles can be used to reduce the uncertainty in detecting Egeria. Iterative active learning is then suggested to adapt the ensembles to the new images, unseen to the active learner. A novel solution to build compact ensembles is also presented, which are needed to expedite the re-training of the active learner. The combined results are accurate and compact ensembles, which require significantly less expert involvement for image region classification.

AB - In many real-world tasks of image classification, limited amounts of labeled data are available to train automatic classifiers. Consequently, extensive human expert involvement is required for instance labeling. Detecting Egeria densa in digital imagery is one such real-world classification task. It presents an additional challenge due to subtle spectral changes in Egeria, which makes it difficult to find a single accurate classifier. A novel solution is proposed to employ an ensemble of classifiers for each class (class-specific ensembles), combined with an active learning scheme. The class-specific ensembles are implicitly diverse. Diversity is required to increase the overall accuracy when combining predictions. The combined predictions of the ensembles can be used to reduce the uncertainty in detecting Egeria. Iterative active learning is then suggested to adapt the ensembles to the new images, unseen to the active learner. A novel solution to build compact ensembles is also presented, which are needed to expedite the re-training of the active learner. The combined results are accurate and compact ensembles, which require significantly less expert involvement for image region classification.

UR - http://www.scopus.com/inward/record.url?scp=2942596261&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=2942596261&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:2942596261

SP - 412

EP - 421

BT - SIAM Proceedings Series

A2 - Berry, M.W.

A2 - Dayal, U.

A2 - Kamath, C.

A2 - Skillicorn, D.

ER -