A double header (with eyes)

[Thomas Daniel <danielt@u.washington.edu>]

This week heralds the last presentations of the autumn quarter.


Jamie Theobold and Andrew Straw will co-present



The Eyes Have It (title supplied anonymously).

114 Kincaid Hall
1230 Wed and Friday

Animal sensory systems have evolved to extract the most useful
information
out of their environment.  Surprisingly, these systems are often poor at

measuring physical values, but excellent at detecting changes in these
values.  An example from vision is motion detection; in many animals, a
great deal of visual processing is concerned with detection of motion
against a background, and motion of the background itself.

This is theoretically difficult because every scheme for detecting
movements has shortcomings.  However, there is excellent experimental
evidence that many animals employ a correlator model of motion
detection,
a system that compares two points in the visual field through unequal
temporal filters.  These models have had good predictive success in
biology as well, but they lack two nearly ubiquitous features of sensory

systems: physiological adaptation and feedback integration.  Further,
they
have largely been explored on the single-unit level, while animals
certainly use thousands of units simultaneously.

We have implemented a spherical lattice of correlators with simultaneous

differential equations.  Each unit can adapt, respond to feedback, and
contribute to a wide-field signal.  Our goal is to examine the effects
of
adaptation, feedback, and unit interaction, with simulated optic flow of

natural scenes.