Reports

Learning Control of Production Systems

Stolfo, Salvatore

One of the central problems in Artificial Intelligence is that of designing appropriate and efficient mechanisms for representing and learning real-world knowledge. We show that it is feasible to infer the control information from an analysis of traces of the successful executions of the production system program provided by a human trainer, thereby creating a system capable of improving its performance by experience.

Files

More About This Work

Academic Units
Computer Science
Publisher
Department of Computer Science, Columbia University
Series
Columbia University Computer Science Technical Reports, CUCS-006-79
Published Here
August 30, 2011