Probabilistic Active Learning for Active Class Selection

Daniel Kottke; Georg Krempl; Marianne Stecklina; Cornelius Styp von Rekowski; Tim Sabsch; Tuan Pham Minh; Matthias Deliano; Myra Spiliopoulou; Bernhard Sick

doi:10.48550/arXiv.2108.03891

Probabilistic Active Learning for Active Class Selection

Daniel Kottke, Georg Krempl, Marianne Stecklina, Cornelius Styp von Rekowski, Tim Sabsch, Tuan Pham Minh, Matthias Deliano, Myra Spiliopoulou, Bernhard Sick

Sub Algorithmic Data Analysis

Research output: Working paper › Preprint › Academic

Abstract

In machine learning, active class selection (ACS) algorithms aim to actively select a class and ask the oracle to provide an instance for that class to optimize a classifier's performance while minimizing the number of requests. In this paper, we propose a new algorithm (PAL-ACS) that transforms the ACS problem into an active learning task by introducing pseudo instances. These are used to estimate the usefulness of an upcoming instance for each class using the performance gain model from probabilistic active learning. Our experimental evaluation (on synthetic and real data) shows the advantages of our algorithm compared to state-of-the-art algorithms. It effectively prefers the sampling of difficult classes and thereby improves the classification performance.

Original language	English
Publisher	arXiv
Pages	1-9
DOIs	https://doi.org/10.48550/arXiv.2108.03891
Publication status	Published - 9 Aug 2021

Bibliographical note

DBLP's bibliographic metadata records provided through http://dblp.org/search/publ/api are distributed under a Creative Commons CC0 1.0 Universal Public Domain Dedication. Although the bibliographic metadata records are provided consistent with CC0 1.0 Dedication, the content described by the metadata records is not. Content may be subject to copyright, rights of privacy, rights of publicity and other restrictions.

Access to Document

10.48550/arXiv.2108.03891Licence: CC BY

2108.03891Submitted manuscript, 593 KBLicence: CC BY

https://arxiv.org/abs/2108.03891Licence: CC BY

Cite this

@techreport{9e58942c913f483c8d60595da5a7c2db,

title = "Probabilistic Active Learning for Active Class Selection",

abstract = "In machine learning, active class selection (ACS) algorithms aim to actively select a class and ask the oracle to provide an instance for that class to optimize a classifier's performance while minimizing the number of requests. In this paper, we propose a new algorithm (PAL-ACS) that transforms the ACS problem into an active learning task by introducing pseudo instances. These are used to estimate the usefulness of an upcoming instance for each class using the performance gain model from probabilistic active learning. Our experimental evaluation (on synthetic and real data) shows the advantages of our algorithm compared to state-of-the-art algorithms. It effectively prefers the sampling of difficult classes and thereby improves the classification performance.",

author = "Daniel Kottke and Georg Krempl and Marianne Stecklina and Rekowski, {Cornelius Styp von} and Tim Sabsch and Minh, {Tuan Pham} and Matthias Deliano and Myra Spiliopoulou and Bernhard Sick",

note = "DBLP's bibliographic metadata records provided through http://dblp.org/search/publ/api are distributed under a Creative Commons CC0 1.0 Universal Public Domain Dedication. Although the bibliographic metadata records are provided consistent with CC0 1.0 Dedication, the content described by the metadata records is not. Content may be subject to copyright, rights of privacy, rights of publicity and other restrictions.",

year = "2021",

month = aug,

day = "9",

doi = "10.48550/arXiv.2108.03891",

language = "English",

pages = "1--9",

publisher = "arXiv",

type = "WorkingPaper",

institution = "arXiv",

}

TY - UNPB

T1 - Probabilistic Active Learning for Active Class Selection

AU - Kottke, Daniel

AU - Krempl, Georg

AU - Stecklina, Marianne

AU - Rekowski, Cornelius Styp von

AU - Sabsch, Tim

AU - Minh, Tuan Pham

AU - Deliano, Matthias

AU - Spiliopoulou, Myra

AU - Sick, Bernhard

N1 - DBLP's bibliographic metadata records provided through http://dblp.org/search/publ/api are distributed under a Creative Commons CC0 1.0 Universal Public Domain Dedication. Although the bibliographic metadata records are provided consistent with CC0 1.0 Dedication, the content described by the metadata records is not. Content may be subject to copyright, rights of privacy, rights of publicity and other restrictions.

PY - 2021/8/9

Y1 - 2021/8/9

N2 - In machine learning, active class selection (ACS) algorithms aim to actively select a class and ask the oracle to provide an instance for that class to optimize a classifier's performance while minimizing the number of requests. In this paper, we propose a new algorithm (PAL-ACS) that transforms the ACS problem into an active learning task by introducing pseudo instances. These are used to estimate the usefulness of an upcoming instance for each class using the performance gain model from probabilistic active learning. Our experimental evaluation (on synthetic and real data) shows the advantages of our algorithm compared to state-of-the-art algorithms. It effectively prefers the sampling of difficult classes and thereby improves the classification performance.

AB - In machine learning, active class selection (ACS) algorithms aim to actively select a class and ask the oracle to provide an instance for that class to optimize a classifier's performance while minimizing the number of requests. In this paper, we propose a new algorithm (PAL-ACS) that transforms the ACS problem into an active learning task by introducing pseudo instances. These are used to estimate the usefulness of an upcoming instance for each class using the performance gain model from probabilistic active learning. Our experimental evaluation (on synthetic and real data) shows the advantages of our algorithm compared to state-of-the-art algorithms. It effectively prefers the sampling of difficult classes and thereby improves the classification performance.

U2 - 10.48550/arXiv.2108.03891

DO - 10.48550/arXiv.2108.03891

M3 - Preprint

SP - 1

EP - 9

BT - Probabilistic Active Learning for Active Class Selection

PB - arXiv

ER -

Probabilistic Active Learning for Active Class Selection

Abstract

Bibliographical note

Access to Document

Fingerprint

Cite this