Learning hierarchical probabilistic logic programs

Abstract Probabilistic logic programming (PLP) combines programs and probabilities. Due to its expressiveness simplicity, it has been considered as a powerful tool for learning reasoning in relational domains characterized by uncertainty. Still, the parameter structure of general PLP is computationally expensive due inference cost. We have recently proposed restriction language called hierarchical (HPLP) which clauses predicates are hierarchically organized. HPLPs can be converted into arithmetic circuits or deep neural networks much cheaper than PLP. In this paper we present algorithms both parameters from data. first an algorithm, probabilistic (PHIL) performs estimation using gradient descent expectation maximization. also propose (SLEAHP), that learns Experiments were performed comparing PHIL SLEAHP with Markov Logic Networks state-of-the art systems respectively. was compared EMBLEM, ProbLog2 Tuffy SLIPCOVER, PROBFOIL+, MLB-BC, MLN-BT RDN-B. The experiments on five well known datasets show our achieve similar often better accuracies but shorter time.