Representing and exploiting conditional preferences within a possibilistic framework : Comparison with other approaches and extensions

Overview & Synopsis

π-pref nets, CP-nets and possibly LP-trees are able to encode specifications of the form “In the context of u, I prefer a to its negation” which is quite similar to the piece of knowledge “If u then generally a”. This rule can be encoded by possibility logic. In some works, researchers have been particularly interested in reasoning with default rules for representing some state of affairs in a possibilistic framework. Because of the similarity between a user preference and a default rule, this work has caught our attention and led us to question whether interpreting a collection of user preference statements as default rules and using some informational principles permit to construct the same ordering as induced by a given graphical preference representation. One of the main goals of this manuscript is also to compare the expressive power of CP-nets, LP-trees and π-pref nets. Using a possibilistic framework, specifications of a user may be also encoded as default rules on which several reasoning approaches are applied to therefore compare their induced orderings. The work is restricted to Boolean variables. An implemented toolbox that supports CP-nets and π-pref nets as graphical structures in addition to the default rule-based algorithms is given in the last chapter.