Current Research
The neural coding problem is perhaps the fundamental question in
systems neuroscience: given some input stimulus (or movement, or
thought, etc.), what is the conditional probability of a neural
response? The roadblock is that we want to know about these response
probabilities given any possible input, and there are typically more
such inputs than we can ever hope to sample. Thus the neural coding
problem is fundamentally a statistics problem: given a finite
number of samples of physiological data, how do we learn the neural
codebook?
Below are more detailed descriptions of some (overlapping) themes in
our recent neural coding research. These neural problems, in turn,
have led to a number of problems which are interesting from a purely
statistical point of view, with connections to machine learning,
latent variable methods, high-dimensional regression, etc.
Likelihood-based neural data
analysis: predicting and decoding spike trains
Estimation of information-theoretic
quantities given limited data
Estimation and analysis of stochastic
biophysically-based neural models
Population coding in primary motor
cortex
Neural coding in the retina
Optimal information-theoretic design of
experiments
Most of this work was done in collaboration.
A chronological list of our publications is available here.
Finally, see this page for
some lecture notes and background on the methods we're working on.
Liam Paninski's home