Learning the parameters of probabilistic logic programs from interpretations

ProbLog is a recently introduced probabilistic extension of the logic programming language Prolog, in which facts can be annotated with the probability that they hold. The advantage of a logic programming based host language is that one can naturally express generative processes using a declarative model. A novel parameter estimation algorithm for learning ProbLog programs from interpretations is introduced. Interpretations are relational state descriptions or possible worlds. The algorithm is essentially a soft-EM algorithm that computes binary decision diagrams for each interpretation allowing for a dynamic programming approach to be implemented. The resulting algorithm has been experimentally which justifies the approach and show its effectiveness.nrpages: 13status: publishe