Focused most probable world computations in probabilistic logic programs (Q454145)

The paper develops an approach based on action probabilistic logic programs. The first two sections give an introduction to the topic and present basic notions that are used further in the article. Section 3 is dedicated to ``a new linear program formulation for worlds of interest''. Starting from a basic ap-program a set of linear constraints is associated; the constraints are defined over the set of all possible worlds of interest. An algorithm to refine the constraints is presented and analyzed in terms of the running time. In the second part of this section, refinement algorithms based on the Monte Carlo technique are analyzed. The experimental results from Section 4 show that the system proposed performs well. Related work and conclusions finish the paper.