Adding Domain Knowledge to SBL Through Feature Construction

Abstract

This paper presents two methods for adding domain knowledge to similarity-based learning through feature construction, a form of representation change in which new features are constructed from relationships detected among existing features. In the first method, domain-knowledge constraints are used to eliminate less desirable new features before they are constructed. In the second method, domain-dependent transformations generalize new features in ways meaningful to the current problem. These two uses of domain knowledge are illustrated in CITRE where they are shown to improve hypothesis accuracy and conciseness on a tic-tac-toe classification problem. Introduction One advantage of explanation-based learning (EBL) is its ability to learn from few examples by exploiting domain-specific constraints represented in a domain theory. Similarity-based learning (SBL), on the other hand, requires relatively large numbers of training instances, but is more readily applicable because a domain ...