Incorporating Explanation-Based Generalization with Analogical Reasoning
Eiju Hirowatari, Setsuo Arikawa · Bulletin of informatics and cybernetics · 1994
The EBG system builds an explanation and learns a concept definition as its generalization provided a domain theory is complete. It does not work when a domain theory is incomplete. Then we introduce a notion of generalizations by an analogy which makes it possible to construct rules necessary for domain theories. Furthermore, we develop EBG by analogical reasoning which copes with the incompleteness by using generalizations by an analogy. We first formulate EBG in a mathematical way, construct EBG by analogical reasoning in terms of our formulation, and realize EBG by analogical reasoning system as a Prolog program. 1 Introduction EBG(Explanation-Based Generalization) takes as inputs a domain theory, a training example, a goal concept and an operationality criterion. It constructs an explanation in terms of the domain theory that proves how the training example satisfies the goal concept definition. Then it determines a set of operationally sufficient conditions for the goal concept ...