Physics
Scientific paper
May 2008
adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=2008njph...10e5019l&link_type=abstract
New Journal of Physics, Volume 10, Issue 5, pp. 055019 (2008).
Physics
Scientific paper
Spike trains of cortical neurons resulting from repeatedpresentations of a stimulus are variable and exhibit Poisson-like statistics. Many models of neural coding therefore assumed that sensory information is contained in instantaneous firing rates, not spike times. Here, we ask how much information about time-varying stimuli can be transmitted by spiking neurons with such input and output variability. In particular, does this variability imply spike generation to be intrinsically stochastic? We consider a model neuron that estimates optimally the current state of a time-varying binary variable (e.g. presence of a stimulus) by integrating incoming spikes. The unit signals its current estimate to other units with spikes whenever the estimate increased by a fixed amount. As shown previously, this computation results in integrate and fire dynamics with Poisson-like output spike trains. This output variability is entirely due to the stochastic input rather than noisy spike generation. As a result such a deterministic neuron can transmit most of the information about the time varying stimulus. This contrasts with a standard model of sensory neurons, the linear nonlinear Poisson (LNP) model which assumes that most variability in output spike trains is due to stochastic spike generation. Although it yields the same firing statistics, we found that such noisy firing results in the loss of most information. Finally, we use this framework to compare potential effects of top-down attention versus bottom-up saliency on information transfer with spiking neurons.
Denève Sophie
Lochmann Timm
No associations
LandOfFree
Information transmission with spiking Bayesian neurons does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.
If you have personal experience with Information transmission with spiking Bayesian neurons, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Information transmission with spiking Bayesian neurons will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-1075808