The University of Sussex

An information-theoretic foundation for inductive generalisation

Christopher Thornton

Although the framework recently proposed by Valiant provides some foundations for empirical learning techniques in probability theory, they still lack a firm foundation in information theory. The paper attempts to make a contribution towards such a foundation by connecting Shannon's information-content measure to one of the main empirical, concept learning techniques; namely, the inductive generalisation mechanism known as mark-raising,


This paper is not available online