Expressivity of Possibilistic Preference Networks with Constraints
Résumé
Among several graphical models for preferences, CP-nets are often used for learning and representation purposes. They rely on a simple preference independence property known as the ceteris paribus independence. Our paper uses a recent symbolic graphical model, based on possibilistic networks, that induces a preference ordering on configurations consistent with the ordering induced by CP-nets. Ceteris paribus preferences in the latter can be retrieved by adding suitable constraints between products of symbolic weights. This connection between possibilistic networks and CP-nets allows for an extension of the expressive power of the latter while maintaining its qualitative nature. Elicitation complexity is thus kept stable, while the complexity of dominance and optimization queries is cut down.
Domaines
Modélisation et simulationOrigine | Fichiers produits par l'(les) auteur(s) |
---|
Loading...