Graphical Models for Preference Representation: An Overview

Nahla Ben Amor (1) , Didier Dubois (2, 3) , Héla Gouider (1) , Henri Prade (2, 3)

1 LARODEC - Laboratoire de Recherche Opérationnelle de Décision et de Contrôle de Processus
2 IRIT-ADRIA - Argumentation, Décision, Raisonnement, Incertitude et Apprentissage
3 CNRS - Centre National de la Recherche Scientifique

Representing preferences into a compact structure has become an important research topic. Graphical models are of special interest. Indeed, they facilitate elicitation, exhibit some form of independence, and serve as a basis for solving optimization and dominance queries about choices. The expressiveness of the representation setting and the complexity of answering queries are then central issues for each approach. This paper proposes an extensive overview of the main graphical models for preference representation and provides a comparative survey by emphasizing their main characteristics. We also indicate possible transformations between some of these models. We contrast qualitative models such as CP-nets and TCP-nets with quantitative ones such as GAI networks, UCP-nets, and Marginal utility nets, and advocate π-Pref nets, recently introduced by the authors, as an interesting compromise between the two types of models.

Mots clés: Intelligence artificielle

Domaines: Intelligence artificielle [cs.AI]; Apprentissage [cs.LG]; Logique en informatique [cs.LO]; Informatique et langage [cs.CL]

Nahla Ben Amor, Didier Dubois, Héla Gouider, Henri Prade. Graphical Models for Preference Representation: An Overview. 10th International Conference on Scalable Uncertainty Management (SUM 2016), Oct 2016, Nice, France. pp. 96-111. ⟨hal-01445229⟩