Attentional Selection in the Unsupervised Generalized Context Model - eScholarship
Emmanuel M. Pothos, Todd M. Bailey · Proceedings of the Annual Meeting of the Cognitive Science Society · 2011
Attentional Selection in the Unsupervised Generalized Context Model Emmanuel Pothos Swansea University Todd Bailey Cardiff University Abstract: The distinction between supervised and unsupervised categorization has had a profound influence in related research. We consider a mechanism, attentional selection, possibly shared by models of supervised and un- supervised categorization. In supervised models such as the Generalized Context Model (GCM; Nosofsky, 1988), attentional selection emphasizes dimensions which are relevant for a taught classification. In a corresponding unsu- pervised version (the UGCM; Pothos & Bailey, 2009), attentional selection emphasizes stimulus dimensions which achieve the best separation of stimuli assigned to different categories. This approach unites the attentional mech- anisms for supervised and unsupervised categorization, subject to the constraint that the assignment of stimuli to categories is exogenous in supervised categorization, but stimulus-driven in unsupervised categorization. In both cases, attentional selection is rational as it facilitates optimal category separation. We suggest that particular stim- ulus dimensions are attended if they make the target classification as intuitive as possible, whether categorization is supervised or unsupervised.