Effect of transformations of numerical parameters in automatic algorithm configuration

Alberto Franzin; Leslie Pérez Cáceres; Thomas Stützle

doi:10.1007/s11590-018-1240-3

Effect of transformations of numerical parameters in automatic algorithm configuration

Alberto Franzin, Leslie Pérez Cáceres, Thomas Stützle

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

5 Scopus citations

Abstract

We study the impact of altering the sampling space of parameters in automatic algorithm configurators. We show that a proper transformation can strongly improve the convergence towards better configurations; at the same time, biases about good parameter values, possibly based on misleading prior knowledge, may lead to wrong choices in the transformations and be detrimental for the configuration process. To emphasize the impact of the transformations, we initially study their effect on configuration tasks with a single parameter in different experimental settings. We also propose a mechanism for how to adapt towards an appropriate transformation and give exemplary experimental results of that scheme.

Original language	English
Pages (from-to)	1741-1753
Number of pages	13
Journal	Optimization Letters
Volume	12
Issue number	8
DOIs	https://doi.org/10.1007/s11590-018-1240-3
State	Published - 1 Dec 2018
Externally published	Yes

Keywords

Automatic algorithm configuration
Numerical parameters
Parameter transformation
Parameter tuning

Access to Document

10.1007/s11590-018-1240-3

Cite this

@article{440f48aab9e14f93b31fd292a84e1b67,

title = "Effect of transformations of numerical parameters in automatic algorithm configuration",

abstract = "We study the impact of altering the sampling space of parameters in automatic algorithm configurators. We show that a proper transformation can strongly improve the convergence towards better configurations; at the same time, biases about good parameter values, possibly based on misleading prior knowledge, may lead to wrong choices in the transformations and be detrimental for the configuration process. To emphasize the impact of the transformations, we initially study their effect on configuration tasks with a single parameter in different experimental settings. We also propose a mechanism for how to adapt towards an appropriate transformation and give exemplary experimental results of that scheme.",

keywords = "Automatic algorithm configuration, Numerical parameters, Parameter transformation, Parameter tuning",

author = "Alberto Franzin and {P{\'e}rez C{\'a}ceres}, Leslie and Thomas St{\"u}tzle",

note = "Publisher Copyright: {\textcopyright} 2018, Springer-Verlag GmbH Germany, part of Springer Nature.",

year = "2018",

month = dec,

day = "1",

doi = "10.1007/s11590-018-1240-3",

language = "English",

volume = "12",

pages = "1741--1753",

journal = "Optimization Letters",

issn = "1862-4472",

publisher = "Springer Verlag",

number = "8",

}

TY - JOUR

T1 - Effect of transformations of numerical parameters in automatic algorithm configuration

AU - Franzin, Alberto

AU - Pérez Cáceres, Leslie

AU - Stützle, Thomas

PY - 2018/12/1

Y1 - 2018/12/1

N2 - We study the impact of altering the sampling space of parameters in automatic algorithm configurators. We show that a proper transformation can strongly improve the convergence towards better configurations; at the same time, biases about good parameter values, possibly based on misleading prior knowledge, may lead to wrong choices in the transformations and be detrimental for the configuration process. To emphasize the impact of the transformations, we initially study their effect on configuration tasks with a single parameter in different experimental settings. We also propose a mechanism for how to adapt towards an appropriate transformation and give exemplary experimental results of that scheme.

AB - We study the impact of altering the sampling space of parameters in automatic algorithm configurators. We show that a proper transformation can strongly improve the convergence towards better configurations; at the same time, biases about good parameter values, possibly based on misleading prior knowledge, may lead to wrong choices in the transformations and be detrimental for the configuration process. To emphasize the impact of the transformations, we initially study their effect on configuration tasks with a single parameter in different experimental settings. We also propose a mechanism for how to adapt towards an appropriate transformation and give exemplary experimental results of that scheme.

KW - Automatic algorithm configuration

KW - Numerical parameters

KW - Parameter transformation

KW - Parameter tuning

UR - http://www.scopus.com/inward/record.url?scp=85041908006&partnerID=8YFLogxK

U2 - 10.1007/s11590-018-1240-3

DO - 10.1007/s11590-018-1240-3

M3 - Article

AN - SCOPUS:85041908006

SN - 1862-4472

VL - 12

SP - 1741

EP - 1753

JO - Optimization Letters

JF - Optimization Letters

IS - 8

ER -

Effect of transformations of numerical parameters in automatic algorithm configuration

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this