VALIDATION OF MACHINE LEARNING TECHNIQUES - DECISION TREES AND FINITETRAINING SET

There has been some recent interest in using machine learning techniqu es as part of pattern recognition systems. However, little attention i s typically given to the validity of the features and types of rules g enerated by these systems and how well they perform across a variety o f features and patterns. We focus on such issues of validity and compa rative performance using two different types of decision free techniqu es. In addition, we introduce the notion of including legal perturbati ons of objects in the training set and show that the performance of th e resulting classifiers was better than that those trained without suc h ''legal'' constructs in the data selection. (C) 1998 SPIE and IS&T. [S1017-9909(98)01101-5].