Explanation-based generalization in a logic programming environment

This paper describes a domain-independent implementation of explanation-based generalization (EBG) within a logic-programming environment. Explanation is interleaved with generalization, so that as the training instance is proven to be a positive example of the goal concept, the generalization is simultaneously created. All aspects of the EBG task are viewed in logic, which provides a clear semantics for EBG, and allows its integration into the logic-programming system. In this light operationally becomes a property requiring explicit reasoning. Additionally, viewing EBG in logic clarifies the relation of learning search-control to EBG, and suggests solutions for dealing with imperfect domain theories.