INVESTIGATIONS OF AN EXEMPLAR-BASED CONNECTIONIST MODEL OF CATEGORY LEARNING

This chapter discusses that learning to categorize objects stands among the most fundamental cognitive processes. Categorizing brings order and organization to our mental lives and is a building block of more complex cognitive processes such as reasoning, problem solving, and thinking. Central issues in the study of categorization include how categories are represented in memory, and what decision processes are involved when people make categorization judgments. It provides an overview of ALEX and discusses its relation to the context model. It explains applications of the model in a variety of category learning situations. The context model and ALEX are also compared. It applies ALEX to several previously published data sets that have been fitted accurately by the context model. The main goal is to provide some quantitative tests of ALEX and demonstrate that it performs well in situations in which the context model has performed well. It also discusses the advantages that are yielded by elaborating the context model as an exemplar-based network. The chapter demonstrates a variety of phenomena that are characterized well by ALEX but not by the context model. There is also some discussion about the limitations of the exemplar-based network, considering how the model might be further extended, and evaluating the model's applications. © 1992, Academic Press Inc.