Gelfond-Zhang aggregates as propositional formulas

Pedro Cabalar (1) , Jorge Fandinno (1, 2) , Torsten Schaub (3, 4) , Sebastian Schellhorn (4)

1 Universidade da Coruña
2 IRIT-ADRIA - Argumentation, Décision, Raisonnement, Incertitude et Apprentissage
3 MORPHEME - Morphologie et Images
4 University of Potsdam = Universität Potsdam

Answer Set Programming (ASP) has become a popular and widespread paradigm for practical Knowledge Representation thanks to its expressiveness and the available enhancements of its input language. One of such enhancements is the use of aggregates, for which different semantic proposals have been made. In this paper, we show that any ASP aggregate interpreted under Gelfond and Zhang's (GZ) semantics can be replaced (under strong equivalence) by a propositional formula. Restricted to the original GZ syntax, the resulting formula is reducible to a disjunction of conjunctions of literals but the formulation is still applicable even when the syntax is extended to allow for arbitrary formulas (including nested aggregates) in the condition. Once GZ-aggregates are represented as formulas, we establish a formal comparison (in terms of the logic of Here-and-There) to Ferraris' (F) aggregates, which are defined by a different formula translation involving nested implications. In particular, we prove that if we replace an F-aggregate by a GZ-aggregate in a rule head, we do not lose answer sets (although more can be gained). This extends the previously known result that the opposite happens in rule bodies, i.e., replacing a GZ-aggregate by an F-aggregate in the body may yield more answer sets. Finally, we characterize a class of aggregates for which GZ- and F-semantics coincide.

Mots clés: Aggregates; Answer set programming

Domaines: Intelligence artificielle [cs.AI]

Pedro Cabalar, Jorge Fandinno, Torsten Schaub, Sebastian Schellhorn. Gelfond-Zhang aggregates as propositional formulas. Artificial Intelligence, 2019, 274, pp.26-43. ⟨10.1016/j.artint.2018.10.007⟩. ⟨hal-02452472⟩