Higher-order correlations in common input shapes the output spiking activity of a neural population

Abstract

Recent neurophysiological experiments suggest that populations of neurons use a computational scheme in which spike timing is regulated by common non-Gaussian inputs across neurons. The presence of beyond-pairwise correlations in the neuronal inputs and the spiking outputs following a non-Gaussian statistics elicits the need of developing a new theoretical framework taking into account the complexity of synchronous activity patterns. To this end, we quantify the amount of higher-order correlations in the common neuronal inputs and outputs of a population of neurons. We provide a novel formalism, of easy numerical implementation, that can capture the subtle changes of the inputs heterogeneities. Within our approach, correlations across neurons arise from -Gaussian inputs into threshold neurons and higher-order correlations in the spiking outputs activity are quantified by the parameter . We present an exhaustive analysis of how input statistics are transformed in this threshold process into output statistics, and we show under which conditions higher-order correlations can lead to either bigger or smaller number of synchronized spikes in the neural population outputs.