Bandit-based Monte-Carlo structure learning of probabilistic logic programs

Probabilistic logic programming allows to model domains with complex and uncertain relationships among entities. While the problem of learning the parameters of such programs has been considered by various authors, the problem of learning their structure is yet to be explored in depth. In this work we present an approximate search method based on a one-player game approach, called LEMUR. It relies on the Monte-Carlo tree search UCT algorithm that combines the precision of tree search with the generality of random sampling. LEMUR works by modifying the UCT algorithm in a similar fashion to FUSE, that considers a finite unknown horizon and deals with the problem of having a huge branching factor. The proposed system has been tested on the UW-CSE and Hepatitis datasets and has shown better performances than those of SLIPCASE and LSM.

• Type: Journal
• Year: 2015
• Where: ML
• Authors: Nicola Di Mauro, Elena Bellodi, Fabrizio Riguzzi
• Keywords: Machine Learning, ML 2015