[CSBi-events] Talk on December 1st at 1:30

[CSBi events]

Talk:  Friday, December 1 Time     1:30-3:30
Place    46-3015

Dr. Liam Paninski

Statistical methods for understanding neural codes

The neural coding problem --- deciding which stimuli will cause a  
given neuron to spike, and with what probability --- is a fundamental  
question in systems neuroscience.  The high dimensionality of both  
stimuli and spike trains has spurred the development of a number of  
sophisticated statistical techniques for learning the neural code  
from finite experimental data.  In particular, modeling approaches  
based on maximum likelihood have proven to be flexible and powerful.

We present three such applications here.  One common thread is that  
the models we have chosen for these data each have concave  
loglikelihood surfaces, permitting tractable fitting (by maximizing  
the loglikelihood) even in high dimensional parameter spaces, since  
no local maxima can exist for the optimizer to get ``stuck'' in.

First we describe neural encoding models in which a linear stimulus  
filtering stage is followed by a noisy integrate-and-fire spike  
generation mechanism incorporating after-spike currents and spike- 
dependent conductance modulations.  This model provides a  
biophysically more realistic alternative to models based on Poisson  
(memoryless) spike generation, and can effectively reproduce a  
variety of spiking behaviors.  We use this model to analyze  
extracellular data from populations of retinal ganglion cells,  
simultaneously recorded during stimulation with dynamic light  
stimuli. Here the model provides insight into the biophysical factors  
underlying the reliability of these neurons' spiking responses, and  
provides a framework for analyzing the cross-correlations observed  
between these cells.  (Joint work with E.J. Chichilnisky, J. Pillow,  
J. Shlens, E. Simoncelli, and V. Uzzell, at NYU and Salk.)

Next we describe how to use this model to ``decode'' the underlying  
subthreshold somatic voltage dynamics, given only the superthreshold  
spike train.  We also point out some connections to spike-triggered  
averaging techniques.

We close by discussing recent extensions to highly biophysically- 
detailed, conductance-based models, which have the potential to allow  
us to estimate the density of active channels in a cell's membrane  
and also to decode the synaptic input to the cell as a function of  
time.  (With M. Ahrens and Q. Huys, Gatsby CNU.)
Host

Emery N. Brown, M.D., Ph.D.
Professor of Computational Neuroscience and Health Sciences
   and Technology
Department of Brain and Cognitive Sciences
MIT-Harvard Division of Health Science and Technology
Massachusetts Institute of Technology
77 Massachusetts Avenue, 46-6079
Cambridge, MA  02139
tel: 617 324 1880
email: enbrown1 at mit.edu

Associate Professor of Anaesthesia
Harvard Medical School
Department of Anesthesia and Critical Care
Massachusetts General Hospital
55 Fruit Street Clinics 3
Boston, MA 02114
tel: 617 726 8786
fax: 617 726 8410
email: brown at neurostat.mgh.harvard.edu