Aggressive Bound Descent for Constraint Optimization - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2022

Aggressive Bound Descent for Constraint Optimization

Résumé

Backtrack search is a classical complete approach for exploring the search space of a constraint optimization problem. Each time a new solution is found during search, its associated bound is used to constrain more the problem, and so the remaining search. An extreme (bad) scenario is when solutions are found in sequence with very small differences between successive bounds. In this paper, we propose an aggressive bound descent (ABD) approach to remedy this problem: new bounds are modified exponentially as long as the searching algorithm is successful. We show that this approach can render the solver more robust, especially at the beginning of search. Our experiments confirm this behavior for constraint optimization
Fichier principal
Vignette du fichier
paper.pdf (664.91 Ko) Télécharger le fichier
Origine : Fichiers éditeurs autorisés sur une archive ouverte

Dates et versions

hal-03749141 , version 1 (10-08-2022)

Identifiants

  • HAL Id : hal-03749141 , version 1

Citer

Thibault Falque, Christophe Lecoutre, Bertrand Mazure, Hugues Wattez. Aggressive Bound Descent for Constraint Optimization. Doctoral Program of the 28th International Conference on Principles and Practice of Constraint Programming., Jul 2022, Haïfa, Israel. ⟨hal-03749141⟩
57 Consultations
53 Téléchargements

Partager

Gmail Facebook X LinkedIn More