Active learning with optimal instance subset selection

Yifan Fu, Xingquan Zhu, Ahmed Elmagarmid

Research output: Contribution to journalArticle

12 Citations (Scopus)

Abstract

Active learning (AL) traditionally relies on some instance-based utility measures (such as uncertainty) to assess individual instances and label the ones with the maximum values for training. In this paper, we argue that such approaches cannot produce good labeling subsets mainly because instances are evaluated independently without considering their interactions, and individuals with maximal ability do not necessarily form an optimal instance subset for learning. Alternatively, we propose to achieve AL with optimal subset selection (ALOSS), where the key is to find an instance subset with a maximum utility value. To achieve the goal, ALOSS simultaneously considers the following: 1) the importance of individual instances and 2) the disparity between instances, to build an instance-correlation matrix. As a result, AL is transformed to a semidefinite programming problem to select a k-instance subset with a maximum utility value. Experimental results demonstrate that ALOSS outperforms state-of-the-art approaches for AL.

Original languageEnglish
Pages (from-to)464-475
Number of pages12
JournalIEEE Transactions on Cybernetics
Volume43
Issue number2
DOIs
Publication statusPublished - 1 Apr 2013

Fingerprint

Labeling
Labels
Problem-Based Learning
Uncertainty

Keywords

  • Active learning
  • Instance subset selection
  • Machine learning

ASJC Scopus subject areas

  • Computer Science Applications
  • Human-Computer Interaction
  • Information Systems
  • Software
  • Control and Systems Engineering
  • Electrical and Electronic Engineering

Cite this

Active learning with optimal instance subset selection. / Fu, Yifan; Zhu, Xingquan; Elmagarmid, Ahmed.

In: IEEE Transactions on Cybernetics, Vol. 43, No. 2, 01.04.2013, p. 464-475.

Research output: Contribution to journalArticle

Fu, Yifan ; Zhu, Xingquan ; Elmagarmid, Ahmed. / Active learning with optimal instance subset selection. In: IEEE Transactions on Cybernetics. 2013 ; Vol. 43, No. 2. pp. 464-475.
@article{1a55f8e6a34049d3a5827b78a8be7fb5,
title = "Active learning with optimal instance subset selection",
abstract = "Active learning (AL) traditionally relies on some instance-based utility measures (such as uncertainty) to assess individual instances and label the ones with the maximum values for training. In this paper, we argue that such approaches cannot produce good labeling subsets mainly because instances are evaluated independently without considering their interactions, and individuals with maximal ability do not necessarily form an optimal instance subset for learning. Alternatively, we propose to achieve AL with optimal subset selection (ALOSS), where the key is to find an instance subset with a maximum utility value. To achieve the goal, ALOSS simultaneously considers the following: 1) the importance of individual instances and 2) the disparity between instances, to build an instance-correlation matrix. As a result, AL is transformed to a semidefinite programming problem to select a k-instance subset with a maximum utility value. Experimental results demonstrate that ALOSS outperforms state-of-the-art approaches for AL.",
keywords = "Active learning, Instance subset selection, Machine learning",
author = "Yifan Fu and Xingquan Zhu and Ahmed Elmagarmid",
year = "2013",
month = "4",
day = "1",
doi = "10.1109/TSMCB.2012.2209177",
language = "English",
volume = "43",
pages = "464--475",
journal = "IEEE Transactions on Cybernetics",
issn = "2168-2267",
publisher = "IEEE Advancing Technology for Humanity",
number = "2",

}

TY - JOUR

T1 - Active learning with optimal instance subset selection

AU - Fu, Yifan

AU - Zhu, Xingquan

AU - Elmagarmid, Ahmed

PY - 2013/4/1

Y1 - 2013/4/1

N2 - Active learning (AL) traditionally relies on some instance-based utility measures (such as uncertainty) to assess individual instances and label the ones with the maximum values for training. In this paper, we argue that such approaches cannot produce good labeling subsets mainly because instances are evaluated independently without considering their interactions, and individuals with maximal ability do not necessarily form an optimal instance subset for learning. Alternatively, we propose to achieve AL with optimal subset selection (ALOSS), where the key is to find an instance subset with a maximum utility value. To achieve the goal, ALOSS simultaneously considers the following: 1) the importance of individual instances and 2) the disparity between instances, to build an instance-correlation matrix. As a result, AL is transformed to a semidefinite programming problem to select a k-instance subset with a maximum utility value. Experimental results demonstrate that ALOSS outperforms state-of-the-art approaches for AL.

AB - Active learning (AL) traditionally relies on some instance-based utility measures (such as uncertainty) to assess individual instances and label the ones with the maximum values for training. In this paper, we argue that such approaches cannot produce good labeling subsets mainly because instances are evaluated independently without considering their interactions, and individuals with maximal ability do not necessarily form an optimal instance subset for learning. Alternatively, we propose to achieve AL with optimal subset selection (ALOSS), where the key is to find an instance subset with a maximum utility value. To achieve the goal, ALOSS simultaneously considers the following: 1) the importance of individual instances and 2) the disparity between instances, to build an instance-correlation matrix. As a result, AL is transformed to a semidefinite programming problem to select a k-instance subset with a maximum utility value. Experimental results demonstrate that ALOSS outperforms state-of-the-art approaches for AL.

KW - Active learning

KW - Instance subset selection

KW - Machine learning

UR - http://www.scopus.com/inward/record.url?scp=84890425291&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84890425291&partnerID=8YFLogxK

U2 - 10.1109/TSMCB.2012.2209177

DO - 10.1109/TSMCB.2012.2209177

M3 - Article

AN - SCOPUS:84890425291

VL - 43

SP - 464

EP - 475

JO - IEEE Transactions on Cybernetics

JF - IEEE Transactions on Cybernetics

SN - 2168-2267

IS - 2

ER -