dc.contributorDelgado, Myriam Regattieri de Biase da Silva
dc.contributorhttps://orcid.org/0000-0002-2791-174X
dc.contributorhttp://lattes.cnpq.br/4166922845507601
dc.contributorKessaci, Marie Eleonore
dc.contributorPozo, Aurora Trinidad Ramirez
dc.contributorhttps://orcid.org/ 0000-0001-5808-3919
dc.contributorhttp://lattes.cnpq.br/2815946827655352
dc.contributorAlmeida, Carolina Paula de
dc.contributorhttps://orcid.org/0000-0003-4939-6432
dc.contributorhttp://lattes.cnpq.br/8586489892942437
dc.contributorDhaenens, Clarisse
dc.contributor.
dc.contributorDelgado, Myriam Regattieri de Biase da Silva
dc.contributorhttps://orcid.org/0000-0002-2791-174X
dc.contributorhttp://lattes.cnpq.br/4166922845507601
dc.contributorLuders, Ricardo
dc.contributorhttps://orcid.org/ 0000-0001-6483-4694
dc.contributorhttp://lattes.cnpq.br/5158617067991861
dc.creatorPavelski, Lucas Marcondes
dc.date.accessioned2022-04-08T13:52:33Z
dc.date.accessioned2022-12-06T14:19:49Z
dc.date.available2022-04-08T13:52:33Z
dc.date.available2022-12-06T14:19:49Z
dc.date.created2022-04-08T13:52:33Z
dc.date.issued2021-12-13
dc.identifierPAVELSKI, Lucas Marcondes. Per-instance algorithm configuration: from meta-learning to multi-objective decomposition. 2021. Tese (Doutorado em Engenharia Elétrica e Informática Industrial) - Universidade Tecnológica Federal do Paraná, Curitiba, 2021.
dc.identifierhttp://repositorio.utfpr.edu.br/jspui/handle/1/27906
dc.identifier.urihttps://repositorioslatinoamericanos.uchile.cl/handle/2250/5246592
dc.description.abstractThe search for the best algorithm and its configuration is a difficult task on most optimization scenarios, especially on NP-hard problems, since different proposed metaheuristics exist, and testing many parameters demands high computational costs. Moreover, the understanding of such parameters and their relation to problem instances is of great importance in the field of algorithm configuration. The literature on Automatic Algorithm Configuration (AAC) proposes several strategies to find out the best configuration, although the focus is usually less on explainability and more on the performance of the different configurations. Based on past experience obtained from data, Per-Instance Algorithm Configuration (PIAC) focuses on the mapping built to recommend the best configurations. This work aims at proposing and analyzing two PIAC approaches. The first, namely MetaL PIAC, is an extension of the algorithm selection problem and uses meta-learning to recommend metaheuristics and their configuration parameters. The other, namely MOAAC/D is based on a brand new multi-objective formulation of the AAC problem, that decomposes the problem space and uses a decomposition-based framework to provide generalist and specialist configurations at the same time. For each objective, there is a set of problems related to it, and a decomposition based multi-objective algorithm is proposed to find good trade-off configurations. As the main study case, the work addresses flowshop problems. Extensive experiments performed on more than 6000 instances, consider MetaL PIAC to tune the parameters of different metaheuristics, and MOAAC/D to tune iterated local search and iterated greedy configurations. The results show that both strategies outperform the generalist solution provided by irace – one of the best well-known AAC – with a slight advantage of MOAAC/D over MetaL PIAC.
dc.publisherUniversidade Tecnológica Federal do Paraná
dc.publisherCuritiba
dc.publisherBrasil
dc.publisherPrograma de Pós-Graduação em Engenharia Elétrica e Informática Industrial
dc.publisherUTFPR
dc.rightshttp://creativecommons.org/licenses/by/4.0/
dc.rightsopenAccess
dc.subjectAlgoritmos
dc.subjectOtimização matemática
dc.subjectHeurística
dc.subjectAprendizado de máquinas
dc.subjectAlgorítmos genéticos
dc.subjectOtimização combinatória
dc.subjectAlgorithms
dc.subjectMathematical optimization
dc.subjectHeuristic
dc.subjectMachine learning
dc.subjectGenetic algorithms
dc.subjectCombinatorial optimization
dc.titlePer-instance algorithm configuration: from meta-learning to multi-objective decomposition
dc.typedoctoralThesis


Este ítem pertenece a la siguiente institución