PLP - Final Project

A Monte Carlo System for Probabilistic Logic Programming

Sampling of probablistic logic programs for SWI-Prolog, similar to Fabrizio Riguzzi's MCINTYRE (2013): http://mcs.unife.it/~friguzzi/Papers/Rig13-FI-IJ.pdf

For information regarding approximate inference via Gibbs sampling / Markov Chain Monte Carlo (MCMC) sampling, see An Analysis of Gibbs Sampling for Probabilistic Logic Programs or A Comparison of MCMC Sampling for Probabilistic Logic Programming (p. 19-29).

Further documentation: https://lukasrieger.github.io/plp-project/

Running the sampler

1. Start a SWI-Prolog session at the command line via swipl
2. Load the montecarlo module, e. g. via [montecarlo].
3. Invoke montecarlo('input.pl', Query, P)., where the first parameter denotes the path to a PLP object program and the second denotes the query to be sampled. After successful sampling, P will contain the sampled probability of Query being true under the given input.

Refer to the PlDoc to customize sampling settings.