Valued Constraint Satisfaction Problems

Problèmes de Satisfaction de Contraintes Valuées

Martin Cooper (1) , Simon de Givry (2) , Thomas Schiex (2)

1 IRIT-ADRIA - Argumentation, Décision, Raisonnement, Incertitude et Apprentissage
2 MIAT INRAE - Unité de Mathématiques et Informatique Appliquées de Toulouse

As an extension of constraint networks, valued constraint networks (or valued CSPs) define a unifying framework for modelling optimisation problems over finite domains in which the cost domain (also denoted as the valuation structure) can be symbolic or numeric but is totally ordered. As is the case for constraint networks, valued CSPs can be viewed as graphical models, defined by a set of variables with an associated finite domain and by a set of cost functions, each involving a subset of the variables. Before reading this chapter, we strongly advise the reader to finish reading chapter “Constraint Reasoning” of this Volume, dedicated to the Constraint Satisfaction Problem.

Domaines: Intelligence artificielle [cs.AI]

Martin Cooper, Simon de Givry, Thomas Schiex. Valued Constraint Satisfaction Problems. A Guided Tour of Artificial Intelligence Research, 2, Springer International Publishing, pp.185-207, 2020, AI Algorithms, 978-3-030-06167-8. ⟨10.1007/978-3-030-06167-8_7⟩. ⟨hal-02942366⟩