Connectionists: New thesis -- Modeling the McCollough Effect

[Julien Ciroux]

Dear Colleagues,

It is my pleasure to announce the availability of my MSc dissertation,
completed at the Institute for Adaptive and Neural Computation at 
the University of Edinburgh, Scotland under the supervision of 
Dr. James A. Bednar.  

The dissertation describes how the McCollough color/orientation
aftereffect can be explained in remarkable detail, based on the same
Hebbian learning mechanisms that lead to the development of
biologically realistic topographic maps for orientation and color
preferences in V1.  The model thus provides a novel link between
long-term development and color vision in the adult.

For more details, please see the abstract below, or download the PDF
file using the link provided.

Julien Ciroux

_______________________________________________________________________________

             Simulating the McCollough Effect in a
       Self-Organizing Model of the Primary Visual Cortex

                      Julien Ciroux

The McCollough effect, an orientation-contingent color aftereffect, is
a useful tool for studying the functionality of the visual system.
There have been several attempts to model and understand the neural
mechanisms underlying it, but none of them give a satisfactory
functional explanation of the effect.  Moreover, numerous experiments
seem to indicate that this aftereffect involves the primary visual
cortex (V1), and there have been no reported efforts to simulate it
with a computational model of this cortical area. 

The present document constitutes the first computational study of the
McCollough effect with a self-organizing map model of V1. LISSOM
(Laterally Interconnected Synergetically Self-Organizing Map) is a
biologically plausible, self-organizing map model that has
successfully accounted for important anatomical and functional
features of V1, as well as for the tilt aftereffect, another visual
effect. Fundamental assumptions of LISSOM are that lateral connections
between cortical neurons play an important role in cortical
processing, and that the cortex self-organizes through Hebbian
learning.  LISSOM has been recently used to model the development of
color and orientation maps at the V1 level and is therefore
appropriate for modeling the McCollough effect.

This thesis presents a simulation of the effect with LISSOM, and the
first rigorous comparison of the simulated effect with available
psychophysical data on the McCollough effect.  The model suggests that
Hebbian learning of lateral inhibitory interactions between orientation
and color selective cells in V1 is the neural mechanism underlying the
effect, thus lending strong support to the theory that the processing
mechanisms involved in the ME are mostly located in the primary visual
cortex. 

Moreover, the structure of the orientation and color map in V1, as
well as the interactions between them, are also examined further, and
it will be shown that the model accounts for both structural
self-organization in color and orientation map and functional
characteristics of adult vision.  Thus, this thesis constitutes the
first psychophysically and biologically grounded simulation of the ME,
and the first to simulate complete topographic maps like those in the
visual cortex. 

Importantly, the model was not originally developed with the ME in
mind, which strongly suggests that the effect is a byproduct of the
general cortical processing mechanisms in the model.  The model also
shows for the first time how the short-term adaptation in the ME is
related to the long-term developmental processes, providing an
important link between development and adult visual function.


URL:
  http://www.inf.ed.ac.uk/publications/thesis/online/IM050308.pdf

Software:
  http://www.topographica.org

Keywords: 
  McCollough effect, orientation maps, color maps, development,
  self-organization, natural images, modeling