To Actively Initialize Active Learning

Yazhou  Yang; Marco Loog

doi:10.1016/j.patcog.2022.108836

To Actively Initialize Active Learning

Yazhou Yang^*, Marco Loog

^*Corresponding author for this work

Pattern Recognition and Bioinformatics

Research output: Contribution to journal › Article › Scientific › peer-review

2 Citations (Scopus)

21 Downloads (Pure)

Abstract

Though much effort has been spent on designing new active learning algorithms, little attention has been paid to the initialization problem of active learning, i.e., how to find a set of labeled samples which contains at least one instance per category. This work identifies the initialization of active learning as a separate and novel research problem, reviews existing methods that can be adapted to be used for this task and, in addition, proposes a new active initialization criterion: the Nearest Neighbor Criterion. Experiments on 16 benchmark datasets verify that the novel method often finds an initialization set with fewer queried samples than other methods do.

Original language	English
Article number	108836
Number of pages	15
Journal	Pattern Recognition
Volume	131
DOIs	https://doi.org/10.1016/j.patcog.2022.108836
Publication status	Published - 2022

Bibliographical note

Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care
Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.

Keywords

active initialization
active learning
minimum nearest neighbor distance
nearest neighbor criterion

Access to Document

10.1016/j.patcog.2022.108836

1_s2.0_S003132032200317X_mainFinal published version, 2.24 MB

Cite this

@article{6e229bbfaf894ae7817bb7989676c94b,

title = "To Actively Initialize Active Learning",

abstract = "Though much effort has been spent on designing new active learning algorithms, little attention has been paid to the initialization problem of active learning, i.e., how to find a set of labeled samples which contains at least one instance per category. This work identifies the initialization of active learning as a separate and novel research problem, reviews existing methods that can be adapted to be used for this task and, in addition, proposes a new active initialization criterion: the Nearest Neighbor Criterion. Experiments on 16 benchmark datasets verify that the novel method often finds an initialization set with fewer queried samples than other methods do.",

keywords = "active initialization, active learning, minimum nearest neighbor distance, nearest neighbor criterion",

author = "Yazhou Yang and Marco Loog",

note = "Green Open Access added to TU Delft Institutional Repository {\textquoteleft}You share, we take care!{\textquoteright} – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public. ",

year = "2022",

doi = "10.1016/j.patcog.2022.108836",

language = "English",

volume = "131",

journal = "Pattern Recognition",

issn = "0031-3203",

publisher = "Elsevier",

}

TY - JOUR

T1 - To Actively Initialize Active Learning

AU - Yang, Yazhou

AU - Loog, Marco

N1 - Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.

PY - 2022

Y1 - 2022

N2 - Though much effort has been spent on designing new active learning algorithms, little attention has been paid to the initialization problem of active learning, i.e., how to find a set of labeled samples which contains at least one instance per category. This work identifies the initialization of active learning as a separate and novel research problem, reviews existing methods that can be adapted to be used for this task and, in addition, proposes a new active initialization criterion: the Nearest Neighbor Criterion. Experiments on 16 benchmark datasets verify that the novel method often finds an initialization set with fewer queried samples than other methods do.

AB - Though much effort has been spent on designing new active learning algorithms, little attention has been paid to the initialization problem of active learning, i.e., how to find a set of labeled samples which contains at least one instance per category. This work identifies the initialization of active learning as a separate and novel research problem, reviews existing methods that can be adapted to be used for this task and, in addition, proposes a new active initialization criterion: the Nearest Neighbor Criterion. Experiments on 16 benchmark datasets verify that the novel method often finds an initialization set with fewer queried samples than other methods do.

KW - active initialization

KW - active learning

KW - minimum nearest neighbor distance

KW - nearest neighbor criterion

UR - http://www.scopus.com/inward/record.url?scp=85132774973&partnerID=8YFLogxK

U2 - 10.1016/j.patcog.2022.108836

DO - 10.1016/j.patcog.2022.108836

M3 - Article

AN - SCOPUS:85132774973

SN - 0031-3203

VL - 131

JO - Pattern Recognition

JF - Pattern Recognition

M1 - 108836

ER -

To Actively Initialize Active Learning

Abstract

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this