Effective collaborative strategies to setup tuners

Elizabeth Montero, María Cristina Riff

Resultado de la investigación: Article

Resumen

Parameter setting problem has demonstrated being a relevant problem related to the use of metaheuristics. ParamILS and I-Race are sophisticated tuning methods that can provide valuable information for designers as well as manage conditional parameters. However, the quality of parameter configurations they can find strongly depends on a proper definition of parameter search space. Evoca is a recently proposed tuner which has demonstrated being less sensitive to the setup of parameters search space. In this paper, we propose an effective collaborative approach that combines Evoca and I-Race as well as Evoca and ParamILS. In both collaborative strategies, Evoca is used to define a proper parameter search space for each tuner. Results demonstrated that the collaborative approaches studied are able to find good parameter configurations reducing the effort required to properly define the parameter search space.

Idioma originalEnglish
PublicaciónSoft Computing
DOI
EstadoPublished - 1 ene 2019

Huella dactilar

Tuners
Search Space
Parameter Space
Tuning
Configuration
Metaheuristics
Strategy

ASJC Scopus subject areas

  • Software
  • Theoretical Computer Science
  • Geometry and Topology

Citar esto

@article{01958b72009a4c00987a64989d474e41,
title = "Effective collaborative strategies to setup tuners",
abstract = "Parameter setting problem has demonstrated being a relevant problem related to the use of metaheuristics. ParamILS and I-Race are sophisticated tuning methods that can provide valuable information for designers as well as manage conditional parameters. However, the quality of parameter configurations they can find strongly depends on a proper definition of parameter search space. Evoca is a recently proposed tuner which has demonstrated being less sensitive to the setup of parameters search space. In this paper, we propose an effective collaborative approach that combines Evoca and I-Race as well as Evoca and ParamILS. In both collaborative strategies, Evoca is used to define a proper parameter search space for each tuner. Results demonstrated that the collaborative approaches studied are able to find good parameter configurations reducing the effort required to properly define the parameter search space.",
keywords = "Collaborative approach, Evoca, I-Race, Parameter search space, ParamILS, Tuning methods",
author = "Elizabeth Montero and Riff, {Mar{\'i}a Cristina}",
year = "2019",
month = "1",
day = "1",
doi = "10.1007/s00500-019-04252-4",
language = "English",
journal = "Soft Computing",
issn = "1432-7643",
publisher = "Springer Verlag",

}

Effective collaborative strategies to setup tuners. / Montero, Elizabeth; Riff, María Cristina.

En: Soft Computing, 01.01.2019.

Resultado de la investigación: Article

TY - JOUR

T1 - Effective collaborative strategies to setup tuners

AU - Montero, Elizabeth

AU - Riff, María Cristina

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Parameter setting problem has demonstrated being a relevant problem related to the use of metaheuristics. ParamILS and I-Race are sophisticated tuning methods that can provide valuable information for designers as well as manage conditional parameters. However, the quality of parameter configurations they can find strongly depends on a proper definition of parameter search space. Evoca is a recently proposed tuner which has demonstrated being less sensitive to the setup of parameters search space. In this paper, we propose an effective collaborative approach that combines Evoca and I-Race as well as Evoca and ParamILS. In both collaborative strategies, Evoca is used to define a proper parameter search space for each tuner. Results demonstrated that the collaborative approaches studied are able to find good parameter configurations reducing the effort required to properly define the parameter search space.

AB - Parameter setting problem has demonstrated being a relevant problem related to the use of metaheuristics. ParamILS and I-Race are sophisticated tuning methods that can provide valuable information for designers as well as manage conditional parameters. However, the quality of parameter configurations they can find strongly depends on a proper definition of parameter search space. Evoca is a recently proposed tuner which has demonstrated being less sensitive to the setup of parameters search space. In this paper, we propose an effective collaborative approach that combines Evoca and I-Race as well as Evoca and ParamILS. In both collaborative strategies, Evoca is used to define a proper parameter search space for each tuner. Results demonstrated that the collaborative approaches studied are able to find good parameter configurations reducing the effort required to properly define the parameter search space.

KW - Collaborative approach

KW - Evoca

KW - I-Race

KW - Parameter search space

KW - ParamILS

KW - Tuning methods

UR - http://www.scopus.com/inward/record.url?scp=85070741956&partnerID=8YFLogxK

U2 - 10.1007/s00500-019-04252-4

DO - 10.1007/s00500-019-04252-4

M3 - Article

JO - Soft Computing

JF - Soft Computing

SN - 1432-7643

ER -