Filtering Decomposable Global Cost Functions

David Allouche (1, 2) , Christian Bessiere (3) , Patrice Boizumault (4) , Simon de Givry (1, 5) , Patricia Gutierrez (6) , Samir Loudni (4) , Jean-Philippe Metivier (4) , Thomas Schiex (1, 2)

1 LBBE - Laboratoire de Biométrie et Biologie Evolutive - UMR 5558
2 UBIA - Unité de Biométrie et Intelligence Artificielle (ancêtre de MIAT)
3 COCONUT - Agents, Apprentissage, Contraintes
4 Equipe CODAG - Laboratoire GREYC - UMR6072
5 INRA - Institut National de la Recherche Agronomique
6 IIIA / CSIC - Artificial Intelligence Research Institute / Spanish Scientific Research Council

As Lee and Leung have previously shown this year, weighted constraint satisfaction problems can benefit from the introduction of global cost functions, leading to a new Cost Function Programming paradigm. In this paper, we explore the possibility of decomposing global cost functions in such a way that enforcing soft local consistencies on the decomposition achieves the same level of consistency on the original global cost function. We give conditions under which directional and virtual arc consistency offer such guarantees. We conclude by experiments on decomposable cost functions showing that decompositions may be very useful to easily integrate efficient global cost functions in solvers.

Domaines: Système multi-agents [cs.MA]; Intelligence artificielle [cs.AI]

David Allouche, Christian Bessiere, Patrice Boizumault, Simon de Givry, Patricia Gutierrez, et al.. Filtering Decomposable Global Cost Functions. AAAI Conference on Artificial Intelligence, Jul 2012, Toronto, ON, Canada. ⟨lirmm-00748187⟩