Possibilistic Networks: A New Setting for Modeling Preferences
Résumé
Possibilistic networks are the counterpart of Bayesian networks in the possibilistic setting. Possibilistic networks have only been studied and developed from a reasoning-under-uncertainty point of view until now. In this short note, for the first time, one advocates their interest in preference modeling. Beyond their graphical appeal, they can be shown to provide a natural encoding of preferences agreeing with the inclusion-based partial order applied to the subsets of preferences violated in the different situations. Moreover they do not encounter the limitations of CP-Nets in terms of representation capabilities. They also enjoy a logical counterpart that may be used for consistency checking. This short note provides a comparative discussion of the merits of possibilistic networks with respect to other existing preference modeling frameworks.
Domaines
Intelligence artificielle [cs.AI]
Origine : Fichiers éditeurs autorisés sur une archive ouverte