Abstract

A common dream for many AI researchers, present authors included, is the construction of a general purpose learning and reasoning system that given basic axiomatic knowledge of a domain is capable of becoming an expert problem solver.
Our machine learning approach, implemented in PRODIGY [21, starts with a general problem-solving engine based on a possibly incomplete domain theory. Learning in PRODIGY is driven by the single metalevel goal of improving the performance of the problem solver in its search for solutions to problems. All of PRODIGY's learning methods are deliberative in interpreting each and every choices made at PRODIGY's decision points. Learning results in the refinement of the initial domain knowledge and in the acquisition of knowledge to control the search process. This chapter is divided into two parts. The first part describes the basic architecture, including the problem solver …

Date

August 30, 1995

Authors

Jaime Carbonell, Oren Etzioni, Yolanda Gil, Robert Joseph, Craig Knoblock, Steven Minton, Manuela Veloso