leaky integrator

Bruce Raoul Parnas brp at muttley
Fri May 21 14:30:56 EST 1993


In article <C7E22q.3uH at murdoch.acc.Virginia.EDU> cp5m at cajal.med.Virginia.EDU (Colin Prepscius) writes:
>Articles by Softky & Koch, 93, and Levine, 91, (and many more) describe  
>modeling of neurons by a "leaky integrator", or "integrate and fire model".   
>These are essentially capacitive models of neurons, but they all seem to  
>include "reset"; after the neuron fires, its input (voltage) is set back to 0  
>(or resting potential).
>
>My question is:  to what extent is this "reset" biologically plausible?  It  
>seems to me that such a reset is very different from a refractory period.   

The "reset" used in integrate and fire models, both leaky and not, is
intended to mimic the effects of repolarization that occurs in neurons
following a spike, mediated by the delayed potassium current.  It isn't
meant to model the refractory period specifically, although it does
result in a relative refractory period, simply to return the spike
initiator to a resting state so that it is ready to fire again.

As an example of what you can do with a version of the leaky integrate
and fire with a variable threshold, you may want to see (shameless
plug here):

BR Parnas and ER Lewis, "A Computationally Efficient Spike Initiator
Model that produces a Wide Variety of Neural Responses" in Neural
Systems: Analysis and Modeling, FH Eeckman (ed), Kluwer Academic
Publishers, pp. 67-76, 1993.


-- 
bruce
(brp at bandit.berkeley.edu)



More information about the Neur-sci mailing list