Learning Default Theories

Abstract

This paper proposes a method to learn from a set of examples a theory expressed in default logic, more precisely in Lukaszewicz’default logic. The main characteristic of our method is to deal with theories where the definitions of a predicate p and definitions for its negation: ¬p are explicitly and simultaneously learned. This method relies on classical generalization techniques proposed in the field of Inductive Logic Programming and on the notion of credulous/skeptical theorem in Default Logic.