This page describes the ILP system Progol which can be accessed by FTP .
Progol combines Inverse Entailment with general-to-specific search through a refinement graph. Inverse Entailment is used with mode declarations to derive the most-specific clause within the mode language which entails a given example. This clause is used to guide a refinement-graph search. Unlike the searches of Shapiro's MIS and Quinlan's FOIL, Progol's search is efficient and has a provable guarantee of returning a solution having the maximum "compression" in the search-space. To do so it performs an admissible A*-like search, guided by compression, over clauses which subsume the most specific clause. Progol deals with noisy data by using the compression measure to trade-off the description of errors against the hypothesis description length. Progol allows arbitrary Prolog programs as background knowledge and arbitrary definite clauses as examples. Despite this, in bench-tests the efficiency of Progol compares favourably with FOIL.
For more details about Progol see the following papers.
Inverse
Entailment and Progol, S. Muggleton, "New Generation Computing
Journal", Vol. 13, pp. 245-286, 1995.
Learning
from positive data, S. Muggleton, "Proceedings of the Sixth
International Workshop on Inductive Logic
progrramming", Springer-Verlag, LNAI 1314, 1997.