Dynamics of Mammalian Neurons
HARRY R. ERWIN
herwin at mason1.gmu.edu
Sun Jul 31 14:22:42 EST 1994
William Calvin (wcalvin at u.washington.edu) wrote:
: herwin at mason1.gmu.edu (HARRY R. ERWIN) writes:
: >I would expect the sigmoid relationship at the population level to be
: >robust for almost any reasonable model of the neuron. The assymetry is
: One caution, and this from someone who used to emphasize f-I curves (the
: individual cell's version of that sigmoid): in cerebral cortex, there is
: a significant possibility that a relatively few inputs are producing
: AND-gate behavior. I say this from Softky & Koch's statistical analysis
: (Neural COmputation 1992, J. Neurophysiol. 1993) of single neuron
: firing. The sigmoid may be more appropriate to motorneuron pools where
: individual inputs are insignificant, where only hundreds at a time manage
: to get a repetitive output going.
: William H. Calvin WCalvin at U.Washington.edu
Thanks, Bill. This is approximately what the biologists have been trying
to tell me. I spotted another couple of papers coming to similar
conclusions this morning going through NIPS 4.
It's beginning to look like Christof Koch may well be the center of the
group that will be the first to make the big move from single-neuron
models to realistic neuronal system models. If I had a fast (100 MHz, 16
MB) PPC or equivalent workstation, I'd build a dynamic model of a
10-minicolumn patch of the OB (about 30,000 neurons) using simplified
(10-compartment) models of periglomerular, pyramidal and granule cell
neurons. I could model about a second's worth of inputs in about 7.5 hours
with that configuration. Then we would see whether we still get Katchalsky
net dynamics and what they mean. I think the way the NCA (my 'activation
network') interacts with the GUW ('great unwashed') mass of the neuronal
population will also come clear at that point. I have a gut-feel already
based on Bartlett Mel's and Christof Koch's papers, but there are too many
dynamic variables in the system for comfort.
It's a good time to be a system modeller!
Internet: herwin at gmu.edu
Just a dumb engineer working on Katchalsky nets....
More information about the Neur-sci