Shigeru Shinomoto (Department of Physics, Kyoto University, Kyoto, Japan)

Abstract:
Temporally irregular sequences of spikes emitted by in vivo cortical neurons had often been approximated by Poisson random events, but it was recently revealed that each neuron expresses a rather specific non-Poisson feature; the non-Poisson irregularity does not vary significantly in time for each neuron but varies among neurons in a manner strongly correlated to cortical functions; neuronal firing patterns are regular in motor areas, random in the visual areas, and bursty in the prefrontal area [1].

Because spike times are determined according to their inputs in a mostly deterministic manner, it may be possible to inquire into the dynamical origin of such firing patterns. Though problems of estimating inputs from output signals are generally ill-posed, a large number of randomly arriving input spikes make it possible to extract some information from an output spike train. Mathematical methods of estimating inputs have been developed on the assumption that activities of presynaptic excitatory and inhibitory neurons are constant over time [2].

Recently, we proposed a method of tracking temporal variations in neuronal input parameters from an evoked spike train [3]. I would like to introduce a method of characterizing instantaneous firing characteristics such as the firing rate and non-Poisson irregularity [4], and suggest utilizing such information for inferring time-varying activities of presynaptic excitatory and inhibitory neurons.

[1] Shinomoto S et al., (2009) PLoS Comput. Biol. 5, e1000433.
[2] Lansky P and Ditlevsen S (2008) Biol. Cybern. 99, 253-262.
[3] Kim H and Shinomoto S (2012) under review.
[4] Shimokawa T and Shinomoto S (2009) Neural Comput 21, 1931-1951.