Finding Accurate Frontiers: A Knowledge-Intensive Approach to Relational Learning

learning (EBL) component. In this paper we provide a brief review of FOIL and FOCL, then discuss how operationalizing a domain theory can adversely affect the accuracy of a learned concept. We argue that instead of operationalizing a domain theory, an analytic learner should return the most general implication of the domain theory, provided this implication is not less accurate than any more specialized implication. We discuss the computational complexity of an algorithm that enumerates all such descriptions and then describe a greedy algorithm that efficiently addresses the problem. Finally, we present a variety of experiments that indicate replacing the operationalization algorithm of FOCL with the new analytic learning method results in more accurate learned concept descriptions. An approach to analytic learning is described that searches for accurate entailments of a Horn Clause domain theory. A hill-climbing search, guided by an information based evaluation function, is performed ...

• Type: Conference
• Year: 1993
• Where: AAAI
• Authors: Michael J. Pazzani, Clifford Brunk
• Keywords: AAAI 1993, Analytic Learning, Analytic Learning Method, Domain Theory, Intelligent Agents