The system FOCL learns Horn clause programs from examples and background
knowledge (optionally). It integrates an explanation-based learning component
with the inductive learning approach of FOIL.
FOCL is able to use
intensionally defined background knowledge and accepts as input a partial,
possibly incorrect rule as an approximation of the target predicate.
User-defined constraints which realise a
declarative language bias allow to restrict the search space.
References
M.J. Pazzani and D. Kibler.
The utility of knowledge in inductive learning.
Machine Learning, 9(1):57-94, 1992.