When Simpler is Better: Automated Configuration of a University Timetabling Solver
Résumé
The Curriculum-Based Course Timetabling problem, which involves scheduling classes in a curriculum, is a prevalent issue in university timetabling.
A number of papers have been published, offering various solutions to the problem by utilizing its unique characteristics.
A current leading method is a hybrid approach combining different local search techniques.
Our study aims to deconstruct this method by breaking it down into smaller components that can be adjusted through the use of parameters.
In order to achieve this, we employ a configurator, more specifically irace, to identify the most optimal configurations.
Our findings reveal that better configurations exist, and notably, these configurations are simpler than the original method. This indicates that the state-of-the-art method can be outperformed by simpler methods that involve fewer algorithms.
This paper presents a comprehensive analysis of the new configurations, which sheds light on the least important components of the original method. Additionally, we conduct an ablation analysis to identify the most crucial parameters or sub-parts that contribute to the most effective solvers.
Domaines
Informatique [cs]
Fichier principal
CEC_Thomas_FEUTRIER_2023_Camera_ready_version.pdf (613.85 Ko)
Télécharger le fichier
Origine | Fichiers produits par l'(les) auteur(s) |
---|