Interest and Predictability: Deciding What to Learn, When to Learn

Lebowitz, Michael

Inductive learning, which involves largely structural comparisons of examples, and explanation-based learning, a knowledge-intensive method for analyzing examples to build generalized schemas, are two major learning techniques used in AI. In this paper, we show how a combination of the two methods - applying generalization-based techniques during the course of inductive learning - can achieve the power of explanation-based learning without some of the computational problems that arise in domains lacking detailed explanatory rules. We show how the ideas predictability and interest can be particulary valuable in this text.



More About This Work

Academic Units
Computer Science
Department of Computer Science, Columbia University
Columbia University Computer Science Technical Reports, CUCS-110-84
Published Here
February 17, 2012