Dan Foygel

Project Proposal: Phased Firing in Cellular Automata Neurons

The problem:

In doing this project, I hope to answer two questions -
1) Can neurons be induced to fire in phase through purely local interactions?
2) Can an imposed segmentation force neurons to fire in phase with their 
neighbors within the segmentation and out of phase with all others, all through
purely local interactions?

First of all, by "neuron" I simply mean an abstract unit with an internal state
which is able to produce output (fire) and to perceive its neighbors' firings 
(and possibly the neighbors' internal states as well).  By two neurons "firing
in phase" I simply mean that the precise temporal instances of their firings 
are positively correlated.  Conversely, "firing out of phase" means that the
firings are negatively correlated.

The implementation:

From the description of these abstract "neurons" it's clear that they naturally
lend themselves to a cellular automata implementation.  I plan to implement the
project as an 2-dimensional array of cellular automata within CSL on the SGI
workstations.  At the moment I'm not sure if CSL is powerful enough to implement
everything I need; if I find out that it is not, I will need to extend it.  In
either case, I expect the bulk of the project to involve developing algorithms
to solve the two problems stated above and implementing them as automata rules 
within CSL.