This repository contains the code used in the experiments for the following paper:
R. Morel, A. Cropper and L. Ong. Typed meta-interpretive learning of logic programs.
data description
For each experiment, the experimental data are in the folders marked by the date they were run on. The folders suffixed '-results' contain the outputs of the system. The csv files prefixed by 'data-typed' contain the data collection for the typed experiments, likewise for the 'data-untyped' files for the untyped runs of the experiments. The '.table' files contain means and standard deviations. These files were generated from the csv files by the 'raw-data-to-plot-data.py' script.
Rerunning the experiments
Each subfolder in the experiments directory contains a 'run.bash' script which runs the respective experiment. The ASP experiments are in the sub-folders named 'asp'.
To run the Prolog experiments a Prolog interpreter is needed. As the experiments were run on 'SWI-Prolog', this is recommended interpreter. The ASP experiments require the hexlite ASP solver.
Contact rolf.morel@cs.ox.ac.uk with any queries.
Online appendix for HEXMILT encoding
The file appendix-HEXMIL_T-encoding.tx contains the full encoding for typed HEXMIL.