THREE POSTDOCTORAL POSITIONS are available in Sophie
Deneve’s team at the Group Neural Theory, Paris, France (see www.gnt.ens.fr).
The GNT is highly interactive and dynamic, is situated in central Paris,
and is embedded within the strong Parisian theoretical neuroscience community. The ideal candidate should have a PhD with a
quantitative background (ideally in fields such as machine learning
and/or computational neuroscience).
We will investigate information coding and learning in
spiking neural networks, combining theoretical approaches, simulations and
analysis of neurophysiological datasets. Possible projects are described in
more details below.
Starting dates are flexible. The positions are for two
years, with net salaries from 2500 to 2800 euro/month depending on prior
experience. We will also provide generous travel funds. Possibilities exists to
get subsidized housing (especially for families).
Candidates should send a letter of motivation (2 pages
max), the contact information of 2 to 3 referees and their CVs to sophie.deneve@ens.fr
BEFORE OCTOBER 10, 2012. Interviews of short-listed candidates will be
conducted in the fall either in Paris, at SFN in New Orleans or by
video-conferences.
Description of projects:
Dealing with uncertainties is necessary for the survival
of any living organism. Indeed, recent years have seen the growing application
of probabilistic inference models to perception and action. Excitable neural
structures face similar uncertainties: they receive noisy and ambiguous inputs
and must accumulate evidence over time, combine unreliable cues and decide
among alternative interpretations of the sensory input. Probabilistic model can
thus be used to further our understanding not only of behavior, but also of the
function and dynamics of biological neural networks.
Our working hypotheses are two-fold. First, we suppose
that neural networks are tuned to estimate sensory or motor variables as
reliably as possible. And second, firing dynamics insure self-consistency,
i.e. these estimates can be extracted by postsynaptic integration of output
spike trains. These two principles entirely constrain the structure, dynamics
and plasticity of the corresponding spiking neural network. In particular, this
purely functional approach captures many aspects of cortical dynamics and
sensory responses (Boerlin and Deneve Plos Comp Bio 2011, Lochman, Ernst and
Deneve J Neurosci 2012, Lochman and Deneve, Curr Opin Neurobiol. 2011).
The projects will consist in
1. Developing and generalizing this framework to explore
its implications for neural coding, dynamics and sensory representations
2. Designing new methods of data analysis able to extract
a network’s function from multi-electrode neural recordings.
3. Applying this approach to neural datasets (multielectrode
recordings – optical imaging data) from sensory and motor areas.
