Learning-induced categorical perception in a neural network model

Abstract: In human cognition, the expansion of perceived between-category distances and compression of within-category distances is known as categorical perception (CP). There are several hypotheses about the causes of CP (e.g., language, learning, evolution) but no functional model. Whether CP is essential to categorisation or simply a byproduct of it is not yet clear, but evidence is accumulating that CP can be induced by category learning. We provide a model for learning-induced CP as expansion and compression of distances in hidden-unit space in neural nets. Basic conditions from which the current model predicts CP are described, and clues as to how these conditions might generalize to more complex kinds of categorization begin to emerge. 1 Categorical Perception Categorical Perception (CP) is defined by the expansion of the perceived differences among members of different categories and/or the compression of the perceived differences among members of the same category (Harnad 1987). A clear consensus on the conditions generating CP has yet to be reached. According to the "Whorf Hypothesis" (Kay & Kempton 1984; Hussein 2012), the between-category separation and/or within-category compression that defines CP is cause by "language".