Long-Tailed Statistics of Corticocortical EPSPs: Origin and Computational Role of Noise in Cortical Circuits

Abstract

Neurons in the brain exhibit highly irregular asynchronous firing even without sensory stimulation. Here, we study the recently proposed hypothesis that a highly non-homogeneous distribution, typically lognormal distribution, of cortico-cortical EPSP (excitatory postsynaptic potential) accounts for the low-rate spontaneous irregular activity observed in vivo. When amplitude distribution of EPSPs among excitatory neuron pairs obeys the lognormal distribution, networks of leaky integrate-and-fire model neurons robustly show ongoing firing state with low firing rate. Moreover, consistent with cortical neurobiology, the obtained activity had high irregularity, low synchronicity, and dynamically balanced excitation-inhibition population activity. We derive effective evolution equations for excitatory and inhibitory population activities from a recurrent network of the leaky integrate-and-fire neurons coupled with highly non-homogeneous connections. Based on the evolution equation, we perfume stability analysis of nontrivial solutions of the equation and reveal underling mechanisms and computational functions of the noise in cortical circuits.