Knowledge-based learning integrating acquisition and learning

Abstract

Empirical learning algorithms are hampered by their inability to use domain knowledge to guide the induction of new rules. This paper describes knowledge-based learning, an approach to learning that selects the examples and relevant attributes for an empirical algorithm. Knowledge-based learning can be used for developing rules for engineering expert systems. Engineers often have some rules for problem solving, but also many experiences (examples) that facilitate solving problems. Knowledge-based learning systems are able to use both forms of knowledge.