Repeat Learning Using Predicate Invention

Khalid Khan1,2, Stephen Muggleton1, and Rupert Parson2

1 Department of Computer Science, University of York
{kmk, stephen}@cs.york.ac.uk
2 Oxford University Computing Laboratory
rupert.parson@comlab.ox.ac.uk



Abstract. Most of machine learning is concerned with learning a single 
concept from a sequence of examples. In repeat learning the teacher
chooses a series of related concepts randomly and independently from a
distribution D A finite sequence of examples is provided for each concept 
in the series. The learner does not initially know V, but progressively
updates a posterior estimation of V as the series progresses. This paper
considers predicate invention within Inductive Logic Programming as a
mechanism for updating the learners estimation of V. A new predicate
invention mechanism implemented in Progol4.4 is used in repeat learning
experiments within a chess domain. The results indicate that significant
performance increases can be achieved. The paper develops a Bayesian
framework and demonstrates initial theoretical results for repeat learning.
References

[1]	J. Baxter. Theoretical models of learning to learn. In T. Mitchell and S. Thrun,
editors, Learning to Learn. Kiuwer, Boston, 1997.
[2]	J. Goodacre. Inductive learning of chess rules using progol. Masters thesis,
Oxford University, Oxford University Computing Laboratory, 1996.
[3]	C. Ling. Inventing necessary theoretical terms in scientific discovery and inductive
logic programming. Technical Report 302, Dept. of Comp. Sci., Univ. of Western
Ontario, 1991.
[4]	S. Muggleton. Predicate invention and utilisation. Journal of Experimental and
Theoretical Artificial Intelligence, 6(1):127130, 1994.
[5]	S. Muggleton. Inverse entailment and Progol. New Generation Computing,
13:245286, 1995.
[6]	S. Muggleton. Learning from positive data. In Proceedings of the Sixth Workshop
on Inductive Logic Programming, Stockholm, 1996.
[7]	S. Muggleton and W. Buntine. Machine invention of first-order predicates by inverting 
resolution. In Proceedings of the 5th International Conference on Machine
Learning, pages 339352. Kaufmann, 1988.
[8]	S. Muggleton and C. D. Page. A learnability model for Universal representations.
In S. Wrobel, editor, Proceedings of the 4th International Workshop on Inductive
Logic Programming, 1994.
[9]	I. Stahl. Constructive induction in inductive logic programming: an overview.
Technical report, Fakultat Informatik, Universitat Stuttgart, 1992.
[10]	I. Stahl. Predicate invention in inductive logic programming. In L. De Raedt, editor, 
Advances in Inductive Logic Programming, pages 3447. lOS Press, Ohmsha,
Amsterdam, 1996.
[11]	R. Wirth and P. ORorke. Constraints on predicate invention. In Proceedings of
the 8th International Workshop on Machine Learning, pages 457461. Kaufmann,
1991.
