Dresden 2014 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

BP: Fachverband Biologische Physik

BP 43: Neurosciences

BP 43.5: Talk

Friday, April 4, 2014, 10:45–11:00, ZEU 250

Statistics of neural spiking under non-Poissonian stimulation — •Tilo Schwalger¹ and Benjamin Lindner^2,3 — ¹EPFL, Lausanne, Switzerland — ²Humboldt-Universität zu Berlin, Berlin, Germany — ³Bernstein-Center for Computational Neuroscience, Berlin, Germany

Nerve cells in the brain generate sequences of spikes with a complex statistics. To understand this statistics, the synaptic input received from other neurons is often modeled by temporally uncorrelated input (Poissonian shot noise), which possesses a flat (white) power spectrum. However, realistic input is temporally correlated because presynaptic neurons exhibit refractoriness, bursting or adaptation, carry a time-dependent signal in their spikes and are subject to short-term synaptic plasticity. The effect of such ``colored noise'' input is poorly understood theoretically because the the associated first-passage-time problem with colored noise is generally a hard problem. Based on a weak-noise expansion of a multi-dimensional Fokker-Planck equation, we derive simple analytical formulas for essential spike train statistics for a tonically firing neuron driven by arbitrarily correlated synaptic input. We show that synaptic input with power-law correlations also leads to power laws in the interspike interval correlations and the Fano factor. Furthermore, input spikes that are more regular than Poisson, as well as neurons with short-term synaptic depression, cause negative interval correlations similar to neurons with adaptation. Our results provide a framework for the interpretation of spiking statistics measured in vivo.