Equivalence between answer-set programs under (partially) fixed input

• Published in: Annals of mathematics and artificial intelligence Vol. 83; no. 3-4; pp. 277 - 295
• Main Authors: Bliem, Bernhard, Woltran, Stefan
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.08.2018; Springer; Springer Nature B.V
• Subjects: Artificial Intelligence; Complex Systems; Computer Science; Context; Declarative programming; Equivalence; Investigations; Logic programs; Mathematical programming; Mathematics; Modularity; Problem solving; Equivalence; 68N17; Answer set programming
• ISSN: 1012-2443; 1573-7470; 1573-7470
• DOI: 10.1007/s10472-017-9567-5

Answer Set Programming has become an increasingly popular formalism for declarative problem solving. Among the huge body of theoretical results, investigations of different equivalence notions between logic programs play a fundamental role for understanding modularity and optimization. While strong equivalence between two programs holds if they can be faithfully replaced by each other in any context (facts and rules), uniform equivalence amounts to equivalent behavior of programs under any set of facts. Both notions (as well as several variants thereof) have been extensively studied. However, the somewhat reverse notion of equivalence which holds if two programs are equivalent under the addition of any set of proper rules (i.e., all rules except facts) has not been investigated yet. In this paper, we close this gap and give a thorough study of this notion, which we call rule equivalence , and its parameterized version where we allow facts over a given restricted alphabet to appear in the context. This notion of equivalence is thus a relationship between two programs whose input is (partially) fixed but where additional proper rules might still be added. Such a notion might be helpful in debugging of programs. We give full characterization results and a complexity analysis for the propositional case of rule equivalence and its relativized versions. Moreover, we address the problem of program simplification under rule equivalence. Finally, we show that rule equivalence is decidable in the non-ground case.