Skip to Main content Skip to Navigation
New interface
Conference papers

Time-Dependent Automatic Parameter Configuration of a Local Search Algorithm

Weerapan Sae-Dan 1 Marie-Eléonore Kessaci 1 Nadarajen Veerapen 1 Laetitia Jourdan 1 
1 ORKAD - Operational Research, Knowledge And Data
CRIStAL - Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189
Abstract : In combinatorial optimization, where problems are often NP-hard, metaheuristics and other approximation algorithms frequently have many parameters in order to adapt to a wide range of scenarios. Very often, obtaining good values for these parameters is a long and tedious manual task but automatic algorithm configuration has been shown to overcome this issue. At the same time, it may also be useful for parameter values to change during the search in order to fine-tune the search process. These parameters include low-level heuristic components. In this article, we propose to use automatic parameter configuration coupled with a control mechanism that switches between parameter configurations at specific times during the search, as an in-between classic parameter tuning and selection hyperheuristics. We test this idea on a local search algorithm, whose parameters allow for selecting different design components, and three combinatorial problems: the Permutation Flowshop Problem, the Traveling Salesman Problem, and the Quadratic Assignment Problem. In comparisons with traditional static automatic parameter configuration, the proposed approach time-dependent is shown to perform better. Additionally, better-performing local search component configurations are identified and discussed.
Complete list of metadata

Cited literature [16 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-02895548
Contributor : Nadarajen Veerapen Connect in order to contact the contributor
Submitted on : Thursday, July 9, 2020 - 6:36:18 PM
Last modification on : Tuesday, November 22, 2022 - 2:26:16 PM
Long-term archiving on: : Monday, November 30, 2020 - 6:27:55 PM

File

GECCO2020-ECADA-final.pdf
Files produced by the author(s)

Identifiers

Collections

Citation

Weerapan Sae-Dan, Marie-Eléonore Kessaci, Nadarajen Veerapen, Laetitia Jourdan. Time-Dependent Automatic Parameter Configuration of a Local Search Algorithm. GECCO '20: Genetic and Evolutionary Computation Conference, ACM, Jul 2020, Cancún, Mexico. ⟨10.1145/3377929.3398107⟩. ⟨hal-02895548⟩

Share

Metrics

Record views

48

Files downloads

92