Optimizing fixpoint evaluation of logic programs with uncertainty

Abstract

We investigate efficient techniques for bottom-up, fixpoint evaluation of logic programs and deductive databases with uncertainty over the certainty domain [0,1], often assumed to be a complete lattice ordered by ≤ with min and max as the meet and join operators, respectively. Standard evaluation methods are inadequate in our context in particular when multiset is used as the semantics structure and when programs use aggregate functions other than the lattice join. We propose a semi-naïve method which adopts and extends relation partitioning and backtracking techniques used in standard case. We developed a running prototype of our method, called SNPB, and studied its performance. Our experimental results indicated a speed-up gain, over the semi-naïve method, ranging from 1.25 to 203, depending on the structures and sizes of the input data set and the programs. © 2008 Springer-Verlag.