Connectionists: Post-doctoral Position in Computational Neuroscience at INRIA

[Pierre Kornprobst]

*Post-doctoral Position in Computational Neuroscience at INRIA*

*Title: Bio-inspired image and video compression schemes*

*Supervised by* Pierre Kornprobst, INRIA Researcher

The INRIA, Neuromathcomp Project Team is looking for highly motivated 
candidates to work in the field of computational neuroscience with a 
focus on information theory. This work is to provide strong basis to 
conceive novel bio-inspired video compression schemes based on realistic 
spiking retinal simulators. It is a very challenging and motivating 
subject for two main reasons: The first is that video compression 
techniques are now essential for most standart equipments such as HDTV 
and DVD, and that recent technical progress allow us to imagine more 
elaborated coding schemes. The second is that neuroscience and recent 
discoveries about the nervous system could be a source of inspiration to 
propose new ideas, especially if one is able to better understand the 
statistics of spike trains.

The first goal will be to better define for a spiking retinal simulator 
to be "realistic". To do so, we will start from the Virtual Retina 
simulator, which is a large-scale spiking simulator developed in the 
team (see [1]). Virtual Retina already proved its capacity to reproduce 
accurately several retinal cells behaviors such as contrast gain 
control. We want to extend this validation to additional retinal cells 
behaviors by comparing the statistics of simulated and real spike trains 
thanks to statistical tools developed in the team (see [2]). Based on 
these results and, we will consider to make the parameters of the 
simulator evolve according to plasticity rules (see [3]).

The second goal will be to investigate decoding strategies based on 
simulated realistic spike trains obtained from the Virtual Retina. This 
work will start from some recent efforts from the team to understand how 
to "invert" retinal operations when still images are presented (see 
[4,5]) and it will be based on other related contribution not 
necessarily focus on the visual system such as [6].

Successful candidate is expected to interact with several researchers 
and PhD students coming from different disciplines and already working 
on the different aspects mentioned above. Mainly, the candidate will 
interact with Bruno Cessac (theoretical physics and mathematics), Marc 
Antonini (electrical engineering), and Pierre Kornprobst (mathematics 
and computational neuroscience). Also, other interactions corresponding 
to current proposals will be encouraged, for example with Adrian 
Palacios and Institut de la Vision.

*Related references*

[1] Virtual Retina: A biological retina model and simulator, with 
contrast gain control A. Wohrer and P. Kornprobst, Journal of 
Computational Neuroscience, Volume 26:2, pp. 219-249 (2009)
[2] ENAS: Event neural assembly Simulation
[3] How Gibbs distributions may naturally arise from synaptic adaptation 
mechanisms, B. Cessac, H. Rostro, J.C. Vasquez, T. Viéville, Journal of 
Statistical Physics, 136, (3), 565-602 (2009).
[4] Retinal filtering and image reconstruction, A. Wohrer, P. Kornprobst 
and M. Antonini, INRIA Research report no 6960, 2009
[5] A novel bio-inspired static image compression scheme for noisy data 
transmission over low-bandwidth channels, K. Masmoudi, M. Antonini, P. 
Kornprobst and L. Perrinet, ICASSP 2010, to appear.
[6] Spikes: Exploring the Neural Code. F Rieke, D Warland, R de Ruyter 
van Steveninck & W Bialek (MIT Press, Cambridge, 1997)

*Deadline to apply: February, 20*

*To know more*, please check job offers in the Neuromathcomp project 
team website [http://www-sop.inria.fr/neuromathcomp]

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://mailman.srv.cs.cmu.edu/mailman/private/connectionists/attachments/20100120/d3ec283b/attachment-0001.html