A Point Process Approach to Cortical Networks

Abstract

The “cognitive” properties of some artificial neuronal networks have introduced attractive models for cortical function. We discuss an extended framework for the description of biological nerve nets such that a direct comparison with the signals from electrophysiological recordings on the level of individual nerve cells becomes feasible. The mathematical analysis of these models leads to explicit conditions on their biophysical parameters giving rise to unexpected conclusions. We demonstrate that the “dynamic repertory” of a system of interacting spiking neurons is dramatically enhanced, if signals are admitted to have a time structure. Some possibilities of a spatio-temporal code in presence of plastic synapses and an appropriate learning rule are discussed.