In this paper, we study issues on disjunctions of propositional Horn theories. In particular, we consider the problems of deciding whether a disjunction of Horn theories is Horn, and, if not, computing a Horn core (i....
详细信息
In this paper, we study issues on disjunctions of propositional Horn theories. In particular, we consider the problems of deciding whether a disjunction of Horn theories is Horn, and, if not, computing a Horn core (i.e., a maximal Horn theory included in this disjunction) and the Horn envelope (i.e., the minimum Horn theory including the disjunction), where a Horn core and the Horn envelope are important approximations of the original theory in artificialintelligence. The problems are investigated for two different representations of Horn theories, namely, for Horn conjunctive normal forms (CNFs) and characteristic models. While the problems are shown to be intractable in general, in the case of bounded disjunctions, we present polynomial time algorithms for testing the Horn property in both representations and for computing a Horn core in the CNF representation. Even in the case of bounded disjunction, no polynomial algorithm exists (unless P=NP) for computing a Horn core in the characteristic model representation. Computing the Horn envelope is polynomial in the characteristic model representation, while it is exponential in the CNF representation, even for bounded disjunction.
暂无评论