Profound Degree: A Conservative Heuristic to Repair Dynamic CSPs - Inria - Institut national de recherche en sciences et technologies du numérique Accéder directement au contenu
Communication Dans Un Congrès Année : 2014

Profound Degree: A Conservative Heuristic to Repair Dynamic CSPs

Résumé

For a better treatment of Dynamic Constraint Satisfaction Problems (DCSPs), several techniques have been developed to be used in repair algorithms. We cite, for example, the variables/values ordering heuristics and local search techniques.We distinguish between static heuristics, which calculate their values once at the beginning of the search, and dynamic heuristics that use an expensive intelligence in terms of solving time.In this paper, we propose a new static variable ordering heuristic, Profound Degree (pdeg), based on deg heuristic, which calculates the degree of influence of a given variable, on the whole constraints network, relatively to its position in the network.We evaluate this heuristic on the Extended Partial-order Dynamic Backtracking (EPBD) approach, which is an approach to repair DCSPs solutions, and we compare it to the best-known variables ordering heuristics (VOHs) for repairing. The evaluation of performance is on random binary problems and meeting scheduling problems, with the criteria of computation time, number of constraints checks and Hamming distance between the former and the current solution.
Fichier principal
Vignette du fichier
978-3-662-44654-6_14_Chapter.pdf (515.03 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-01391305 , version 1 (03-11-2016)

Licence

Paternité

Identifiants

Citer

Yosra Acodad, Amine Benamrane, Imade Benelallam, El Houssine Bouyakhf. Profound Degree: A Conservative Heuristic to Repair Dynamic CSPs. 10th IFIP International Conference on Artificial Intelligence Applications and Innovations (AIAI), Sep 2014, Rhodes, Greece. pp.140-149, ⟨10.1007/978-3-662-44654-6_14⟩. ⟨hal-01391305⟩
49 Consultations
91 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More