A Constraint Satisfaction Approach to Tractable Theory Induction
A novel framework for combining logical constraints with theory induction in Inductive Logic Programming is presented. The constraints are solved using a boolean satisfiability solver (SAT solver) to obtain a candidate solution. This speeds up induction by avoiding generation of unnecessary candidates with respect to the constraints. Moreover, using a complete SAT solver, search space exhaustion is always detectable, leading to faster small clause/base case induction. We run benchmarks using two constraints: input-output specification and search space pruning. The benchmarks suggest our constraint satisfaction approach can speed up theory induction by four orders of magnitude or more, making certain intractable problems tractable.
KeywordsInductive Logic Programming Theory induction Constraint satisfaction
The work described in this paper was supported by a grant from the Research Grants Council of the Hong Kong Special Administrative Region, China [Project No. CityU 124409].
- 3.Blackburn, P., Bos, J., Striegnitz, K.: Learn Prolog Now!. College Publications, London (2006)Google Scholar
- 9.Sebag, M., Rouveirol, C.: Constraint inductive logic programming. In: De Raedt, L. (ed.) Advances in ILP. IOS Press, Amsterdam (1996)Google Scholar
- 13.Moskewicz, M.W., Madigan, C.F., Zhao, Y., Zhang, L., Malik, S.: Chaff: engineering an efficient SAT solver. In: Proceedings of the 38th Annual Design Automation Conference, DAC ’01, pp. 530–535. ACM, New York (2001)Google Scholar