Liam Paninski, 10/17/01
<br>
<br>
I'll continue the discussion of the gain changes and neural
adaptation, relevant to the Fairhall et al. paper Jonathan presented
<a href="pillow10_10.html">last week</a>.  Brian Lau and I have done
some similar experiments in an in vitro slice preparation (in Alex
Reyes' lab); I'll (very briefly) review some of our results, then
spend most of the hour talking about some thoughts on modeling the
effects we see.  
<br>
<br>
Previous work has relied heavily on the LN (linear - nonlinear)
formalism we've discussed here in the past, in which the neuron is
modeled as implementing some memoryless nonlinear operation on a
filtered version of the input signal.  It is unclear, however, what a
simple dynamical model of a neuron might do with respect to the usual
LN estimation procedures - does a garden-variety integrate-and-fire
cell, for example, "adapt to input statistics" as Fairhall's (and many
others') cells appear to?  It turns out that a partial differential
equation approach (developed by Knight, Sirovich, Tranchina, Abbott,
etc. over the past few years) provides some surpisingly powerful tools
for us here.  I'll try to explain some of these methods and their
application to our data.