Date: Wed, 20 Nov 1996 19:16:38 GMT Server: Apache/1.0.3 Content-type: text/html Content-length: 2038 Last-modified: Sat, 06 Jul 1996 18:43:44 GMT
A central research challenge for case-based reasoning (CBR) is developing methods for adapting retrieved solutions to solve new problems. Current adaptation methods rely on hand-coded task-specific information, placing considerable burden on the system developer. This project studies how useful case adaptation knowledge can automatically be learned and reused. It models how CBR systems, starting with only domain- and task-independent adaptation information, can augment that general knowledge by learning domain-specific information from the adaptations they perform. Learning is done by storing traces of the solution process used in successful adaptations, to enable those traces to be reused for future adaptations. This method allows CBR systems to learn not only new solutions, but also how to make better use of existing solutions in their memories.
Associated Faculty: David B. Leake
Associated Graduate Students: Andrew Kinley, David Wilson.
Support: This research is supported by grant IRI-9409348 from the Knowledge Models and Cognitive Systems Program of the IRIS division of the National Science Foundation .
For more information about CBR research at Indiana, click here