Induction, abduction, and consequence-finding

Abstract

This paper shows a sound and complete method for inverse entailment in inductive logic programming. We show that inverse entailment can be computed with a resolution method for consequence-finding. In comparison with previous work, induction via consequence-finding is sound and complete for finding hypotheses from full clausal theories, and can be used for inducing not only definite clauses but also non-Horn clauses and integrity constraints. We also compare induction and abduction from the viewpoint of consequence-finding, and clarify the relationship and difference between the two.