A connexionist model for category construction
Erik M. Altmann, Axel Cleeremans, Christian D. Schunn, Wayne D. Gray · Psychology Press eBooks · 2001
The cognitive model we started from comes from a psychological decision theory model, the Moving Basis Heuristics (MBH) (Barthélemy & Mullet, 1986). MBH is aimed to implement the bounded rationality of a Decision Maker (DM) as introduced by Simon (1969). It combines three basic principles. According to the parsimony principle, the DM extracts some subsets from a data set whose size is small enough to be compatible with human short-range abilities and with human computational abilities. The reliability principle states that the DM extracts from the data a subset large enough and composes values in such a way as to appear meaningful and to lead to reliable decisions. According to the decidability principle, the DM allows himself to change criterion if the current criterion does not lead to a decision (see (Barthélemy & Mullet, 1986) for details). We add a fourth principle to make our network achieve stable decisions: the resonance principle states that a decision made on an object is performed by a resonance between this object and an expectancy of what this decision should be. Here, resonance refers to of Grossberg ‘s Adaptive Resonance Theory (Carpenter & Grossberg, 1987).