Algorithm selection on a meta level

Publikation: Beitrag in FachzeitschriftArtikelForschungPeer-Review

Autoren

Externe Organisationen

  • Universität Paderborn
  • Ludwig-Maximilians-Universität München (LMU)
Forschungs-netzwerk anzeigen

Details

OriginalspracheEnglisch
Seiten (von - bis)1253-1286
Seitenumfang34
FachzeitschriftMachine learning
Jahrgang112
Ausgabenummer4
Frühes Online-Datum18 Apr. 2022
PublikationsstatusVeröffentlicht - Apr. 2023
Extern publiziertJa

Abstract

The problem of selecting an algorithm that appears most suitable for a specific instance of an algorithmic problem class, such as the Boolean satisfiability problem, is called instance-specific algorithm selection. Over the past decade, the problem has received considerable attention, resulting in a number of different methods for algorithm selection. Although most of these methods are based on machine learning, surprisingly little work has been done on meta learning, that is, on taking advantage of the complementarity of existing algorithm selection methods in order to combine them into a single superior algorithm selector. In this paper, we introduce the problem of meta algorithm selection, which essentially asks for the best way to combine a given set of algorithm selectors. We present a general methodological framework for meta algorithm selection as well as several concrete learning methods as instantiations of this framework, essentially combining ideas of meta learning and ensemble learning. In an extensive experimental evaluation, we demonstrate that ensembles of algorithm selectors can significantly outperform single algorithm selectors and have the potential to form the new state of the art in algorithm selection.

ASJC Scopus Sachgebiete

Zitieren

Algorithm selection on a meta level. / Tornede, Alexander; Gehring, Lukas; Tornede, Tanja et al.
in: Machine learning, Jahrgang 112, Nr. 4, 04.2023, S. 1253-1286.

Publikation: Beitrag in FachzeitschriftArtikelForschungPeer-Review

Tornede, A, Gehring, L, Tornede, T, Wever, M & Hüllermeier, E 2023, 'Algorithm selection on a meta level', Machine learning, Jg. 112, Nr. 4, S. 1253-1286. https://doi.org/10.1007/s10994-022-06161-4
Tornede A, Gehring L, Tornede T, Wever M, Hüllermeier E. Algorithm selection on a meta level. Machine learning. 2023 Apr;112(4):1253-1286. Epub 2022 Apr 18. doi: 10.1007/s10994-022-06161-4
Tornede, Alexander ; Gehring, Lukas ; Tornede, Tanja et al. / Algorithm selection on a meta level. in: Machine learning. 2023 ; Jahrgang 112, Nr. 4. S. 1253-1286.
Download
@article{1a5193ef7e0e4168a1d7733b43c34a30,
title = "Algorithm selection on a meta level",
abstract = "The problem of selecting an algorithm that appears most suitable for a specific instance of an algorithmic problem class, such as the Boolean satisfiability problem, is called instance-specific algorithm selection. Over the past decade, the problem has received considerable attention, resulting in a number of different methods for algorithm selection. Although most of these methods are based on machine learning, surprisingly little work has been done on meta learning, that is, on taking advantage of the complementarity of existing algorithm selection methods in order to combine them into a single superior algorithm selector. In this paper, we introduce the problem of meta algorithm selection, which essentially asks for the best way to combine a given set of algorithm selectors. We present a general methodological framework for meta algorithm selection as well as several concrete learning methods as instantiations of this framework, essentially combining ideas of meta learning and ensemble learning. In an extensive experimental evaluation, we demonstrate that ensembles of algorithm selectors can significantly outperform single algorithm selectors and have the potential to form the new state of the art in algorithm selection.",
keywords = "Algorithm selection, Bagging, Boosting, Ensemble learning, Meta learning, Stacking",
author = "Alexander Tornede and Lukas Gehring and Tanja Tornede and Marcel Wever and Eyke H{\"u}llermeier",
note = "Publisher Copyright: {\textcopyright} 2022, The Author(s).",
year = "2023",
month = apr,
doi = "10.1007/s10994-022-06161-4",
language = "English",
volume = "112",
pages = "1253--1286",
journal = "Machine learning",
issn = "0885-6125",
publisher = "Springer Netherlands",
number = "4",

}

Download

TY - JOUR

T1 - Algorithm selection on a meta level

AU - Tornede, Alexander

AU - Gehring, Lukas

AU - Tornede, Tanja

AU - Wever, Marcel

AU - Hüllermeier, Eyke

N1 - Publisher Copyright: © 2022, The Author(s).

PY - 2023/4

Y1 - 2023/4

N2 - The problem of selecting an algorithm that appears most suitable for a specific instance of an algorithmic problem class, such as the Boolean satisfiability problem, is called instance-specific algorithm selection. Over the past decade, the problem has received considerable attention, resulting in a number of different methods for algorithm selection. Although most of these methods are based on machine learning, surprisingly little work has been done on meta learning, that is, on taking advantage of the complementarity of existing algorithm selection methods in order to combine them into a single superior algorithm selector. In this paper, we introduce the problem of meta algorithm selection, which essentially asks for the best way to combine a given set of algorithm selectors. We present a general methodological framework for meta algorithm selection as well as several concrete learning methods as instantiations of this framework, essentially combining ideas of meta learning and ensemble learning. In an extensive experimental evaluation, we demonstrate that ensembles of algorithm selectors can significantly outperform single algorithm selectors and have the potential to form the new state of the art in algorithm selection.

AB - The problem of selecting an algorithm that appears most suitable for a specific instance of an algorithmic problem class, such as the Boolean satisfiability problem, is called instance-specific algorithm selection. Over the past decade, the problem has received considerable attention, resulting in a number of different methods for algorithm selection. Although most of these methods are based on machine learning, surprisingly little work has been done on meta learning, that is, on taking advantage of the complementarity of existing algorithm selection methods in order to combine them into a single superior algorithm selector. In this paper, we introduce the problem of meta algorithm selection, which essentially asks for the best way to combine a given set of algorithm selectors. We present a general methodological framework for meta algorithm selection as well as several concrete learning methods as instantiations of this framework, essentially combining ideas of meta learning and ensemble learning. In an extensive experimental evaluation, we demonstrate that ensembles of algorithm selectors can significantly outperform single algorithm selectors and have the potential to form the new state of the art in algorithm selection.

KW - Algorithm selection

KW - Bagging

KW - Boosting

KW - Ensemble learning

KW - Meta learning

KW - Stacking

UR - http://www.scopus.com/inward/record.url?scp=85128259652&partnerID=8YFLogxK

U2 - 10.1007/s10994-022-06161-4

DO - 10.1007/s10994-022-06161-4

M3 - Article

VL - 112

SP - 1253

EP - 1286

JO - Machine learning

JF - Machine learning

SN - 0885-6125

IS - 4

ER -

Von denselben Autoren