Réseaux possibilistes de préférences et arbres de préférences lexicographiques – Une comparaison

The paper compares two graphical approaches proposed for the qualitative modeling of preferences : ?-pref nets and LP-trees. The former uses the graphical setting of possibilistic networks for completing partial specifications of user preferences, while the latter, which is based on lexicographic ordering, appears to offer a convenient framework for learning preferences. The ?-pref network representation appears to be more flexible, even if the addition of very specific constraints allows us to recover the total order of the LP-trees.