next up previous
Next: Introduction

Learning How to Reason from Prior Experiences

Andrew Kinley, David Wilson, and David B. Leake
Computer Science Department, Indiana University
* Bloomington, IN 47405
* {akinley,davwils,leake}@cs.indiana.edu

Abstract:

Case-based reasoning (CBR) solves problems by retrieving the solutions to similar prior problems and adapting the prior solutions to new circumstances. Psychological research has provided support and motivation for CBR as a model of some types of human reasoning and learning. In artificial intelligence, studies on learning in CBR have traditionally focused on learning two types of knowledge: new cases, and new indexing criteria for case retrieval. Little attention, however, has been given to the processes by which criteria for applying retrieved cases may be learned. This paper describes ongoing research on how a case-based reasoner can learn to apply its stored cases more effectively, focusing on learning to refine two aspects of the case application process: the strategies used to adapt cases to fit new situations, and the criteria used to determine which prior cases are most similar to a new problem situation.





Andrew Kinley
Thu Apr 4 10:19:36 EST 1996