Source selection for analogical reasoning an empirical approach

  • Authors:

  • William A. Stubblefield;George F. Luger

  • Affiliations:

  • Sandia National Laboratories, Albuquerque, New Mexico;Department of Computer Science, University of New Mexico, Albuquerque, New Mexico

  • Venue:
  • AAAI'96 Proceedings of the thirteenth national conference on Artificial intelligence - Volume 1
  • Year:
  • 1996

Quantified Score

Hi-index 0.00

Visualization

Abstract

The effectiveness of an analogical reasoner depends upon its ability to select a relevant analogical source. In many problem domains, however, too little is known about target problems to support effective source selection. This paper describes the design and evaluation of SCAVENGER, an analogical reasoner that applies two techniques to this problem: (1) An assumption-based approach to matching that allows properties of candidate sources to match unknown target properties in the absence of evidence to the contrary. (2) The use of empirical learning to improve memory organization based on problem solving experience.