Categorizing Numeric lnformation for Generalization*

Learning programs that generalize from real-world examples will have to deal with many different kinds of data. Continuous numeric data can cause problems far algorithms that scorch for examples with identical property values. These problems can be surmounted by categorizing the numeric doto. However, this process has problems of its awn. In this paper, we look at the need for categorizing numeric data and several methods for doing so. We concentrate on the use of generalizofion-bosed memory, a memory organization where octuol examples ore stored along with generalizations, which leads to a generolizotion-bosed categorization algorithm. We also consider how to use a number heurlstlc, looking for gaps. These methods hove been implemented in the UNIMEM computer system. Examples ore presented of these algorithms categorizing doto about the states of the United States.