🚀 go-pugleaf

>gal2@ellis.uchicago.edu (Jacob Galley) asks:

>>Can anyone direct me toward any viable theories of meaning that manage
>>to link symbolic / categorical representation to any form of
>>nonsymbolic / noncategorical representation? What kinds of nonsymbolic
>>representation are there?

You might want to have a look at Adaptive Resonance Theory (ART) to
see how stable categories are formed from "non-symbolic" representations or
activity patterns across neural populations.  For starters, try:

  Carpenter, G., and Grossberg, S., Pattern Recognition by Self-Organizing
      Neural Networks, MIT Press, Cambridge 1991

Assigning meaning to the categories entails embedding the pattern classifier,
e.g. an ART network, into a larger mechanism that is tailored to the specific
modality or problem being addressed.  For example, in a neural network
capable of evidencing real--time conditioning behavior, one combines
categorized stimuli with internal state information to yield (conditional!)
conditioned responses. See the first few chapters of:

  Grossberg, S., The Adaptive Brain I: Cognition, Learning, Reinforcement,
      and Rhythm, North--Holland, NY^2, 1988

As for the kinds of non--symbolic representations, Grossberg's point
that information is encoded as patterns of activity across neural populations
seems to be gaining a large following. The meaning of these patterns is 
very context dependent. In some places it is the sensory flux washing 
in from a given sensory modality, but in other places it is may be an
unfolding motor command. And, sometimes, it may be a category.

Cheers!

Chris Pribe
Department of Cognitive and Neural Systems
Boston University

cpribe@cns.bu.edu