Parameter estimation in ProbLog from annotated queries

We introduce the problem of learning the parameters of the probabilistic database ProbLog. Given the observed success probabilities of a set of queries, we compute the unobserved probabilities attached to facts that have a low approximation error on the training examples as well as on unseen examples. The objective function to be minimized is the squared-error between the measured and computed values of the queries. As we will show, our approach is able to learn both from queries and from proofs and even from both simultaneously. This makes it flexible and allows faster training in domains where proofs are available. Experiments on real world data show the usefulness and effectiveness of this least squares calibration of probabilistic databases.nrpages: 34status: publishe