Human categorization

My first interest was human categorization. I am particularly interested in formal models of human categorization (and category learning). I was most active on this topic while writing my doctoral dissertation (1994-1998). Then, for almost 8 years, I had not much time to pursue this line of research. Fortunately, since 2006, two of my PhD students (Maarten De Schryver and Katleen Vandist) are picking up my work from where I left it.

[This page is under construction]

• Mixture models of categorization

  If we assume that a category can be represented by a probability density function in feature space, we need a way to model the density. A convenient way to describe this density function is by using a mixture model. This was the theme of my doctoral dissertation. Two relevant papers on this topic are:
  
  
  • Rosseel, Y. (2002). Mixture models of categorization. Journal of Mathematical Psychology, 46, 178-210.
    
    
  • Rosseel, Y. (1996). Connectionist models of categorization: A statistical interpretation. Psychologica Belgica, 36, 93--112.
    
    
  Note that by varying the number (J) of components in the mixture, the mixture framework is able to span the whole continuum between prototype models (J=1) and exemplar models (J=number of exemplars).

• Reduced exemplar models

  The idea of reduced exemplar (REX) models is that a category is represented by a relatively small number of exemplars.

• Semi-supervised category learning

  How can we learn about a category without being supervised all the time?