Relating and extending semantical approaches to possibilistic reasoning (Q1328146)

The two semantical approaches of Dubois-Prade and of Ruspini are put into relation. It is shown that the monotonic fragment of possibilistic logic (PL) can be semantically embedded into similarity logic (SL). Following the proposition of a semantical reasoning concept as an extension of possibilistic reasoning, a fuzzy truth-valued logic (FTL) is introduced. In FTL incompleteness modeled as a fuzzy set of possible `worlds'. This is the same as in PL, but FTL differs from PL in that FTL allows each world to be described by a many-valued interpretation of a set of logical formulas. The authors study the above-mentioned approaches from the point of view of FTL. Section 2 shows the basic concepts of possibilistic reasoning due to Dubois-Prade, and of Ruspini's semantical foundations of fuzzy logic. Specifying language, semantics, and inference, basic characteristics of PL are given, as well as Ruspini's unconditional, conditional possibility an necessity distributions, including a generalized modus ponens, are introduced. These main constructs of Ruspuni are then reformulated, both from a measure theoretic and from a fuzzy sets theory points of view. It is shown that both notions of fuzzy sets and of similarity relation can be combined. Both approaches differ on the inference level. In SL, inference is formulated through conditional measures, in PL conditional statements are modeled by means of material implication. Thus, both approaches are treated at the semantic level. Section 3 deals with mapping SL into PL, and mapping PL and SL. Theorem 3 shows how PL can be embedded in SL. Semantical relation ships between PL and SL, and further aspects of the relationship between their inference mechanisms are considered. Section 4 presents FTL as generalization of PL and SL. Section 5 gives the main conclusions, and in annex A the main results on T-similarity relations and its generators are summarized. Annex B gives a characterization for the subclass of SL models that corresponds to the class of PL models.