Jon Pillow, 9/2/03
<br>
<br>
Examining the neural code with a parametric model of neural response
<br>
<br>

I will discuss some recent progress of our work on the neural coding
problem.  This work examines the probabilistic relationship between a
sensory stimulus and the neural response using a model consisting of a
linear filter followed by noisy, leaky integrate-and-fire spiking.
Previous work has shown that this model can manifest a rich variety of
response behaviors, and that it more faithfully captures the spike train
statistics of real neurons than the standard 'LNP' model.  In recent work,
we have shown that the parameters of this model can be reliably and
effeciently estimated from extracellular spike train data using maximum
likelihood.  I will introduce the mathematical framework involved in
computing and maximizing the likelihood for this model, and will discuss
some future directions, which include applying the model to account for
variability of neural responses to a repeated stimulus.

<br>
<br>
Joint work with Liam Paninski