Efficient induction of recursive prolog definitions

摘要

The ability to learn recursive defintions is a desirable characteristic of a learner. This paper presents Clam, a system that efficiently learns Prolog purely and left-recursive definitions from small data sets by using inverse implication. A learning curve for Clam shows that the accuracy grows with the increase of both positive and negative examples. We believe our system can be used as a preprocessor for a general-purpose system when few examples are at hand.