Mechanisms That Modulate the Transfer of Spiking Correlations

Abstract

Correlations between neuronal spike trains affect network dynamics and population coding. Overlapping afferent populations and correlations between presynaptic spike trains introduce correlations between the inputs to downstream cells. To understand network activity and population coding, it is therefore important to understand how these input correlations are transferred to output correlations.Recent studies have addressed this question in the limit of many inputs with infinitesimal postsynaptic response amplitudes, where the total input can be approximated by gaussian noise. In contrast, we address the problem of correlation transfer by representing input spike trains as point processes, with each input spike eliciting a finite postsynaptic response. This approach allows us to naturally model synaptic noise and recurrent coupling and to treat excitatory and inhibitory inputs separately.We derive several new results that provide intuitive insights into the fundamental mechanisms that modulate the transfer of spiking correlations.

Poisson-Like Spiking in Circuits with Probabilistic Synapses

“…It has also been suggested that to produce in vivo high spiking variability, presynaptic spikes need to be synchronous [49], but it was unknown how large input variability caused by synchrony can be generated in recurrent networks. Previous models have also explored the role of synaptic noise in neuronal computations [50], [51], [52], [53], in up-down state transitions [54] and in the spiking variability of single cells or pairs of cells [50], [55], [56], [57], but the role of probabilistic synapses on Poisson-like variability in large recurrent networks or over a broad continuum of firing rates was not studied. As we have shown here, probabilistic synapses generate multiplicative noise that is amplified by recurrent connections without fine-tuning of the network parameters.…”

The mechanics of state-dependent neural correlations

“…In particular, the example of a pair of uncoupled neurons driven by partially correlated inputs has been extensively studied. Formally, the nonlinear transfer between continuous input and spike response ensures that ρ < ρ x 92,37,39,93,94 , yet the influence of the nonlinearity can be controlled by several factors. In many neuron models L increases with the firing rate of a neuron resulting in a relationship between firing rates and ρ 37,38,95 .…”

Finite volume and asymptotic methods for stochastic neuron models with correlated inputs

“…Even weak correlations between individual cells can significantly impact the ensemble activity of a population (Shadlen and Newsome 1998;Rosenbaum et al 2010). Recent studies explore the joint statistics of integrate-and-fire receiving correlated stochastic inputs (Burak et al 2009;de la Rocha et al 2007;Rosenbaum and Josić 2011;Schneider et al 2006;Shea-Brown et al 2008;Tchumatchenko et al 2010;Moreno-Bote and Parga 2006;Ostojić et al 2009;Vilela and Lindner 2009;Moreno-Bote 2010). Here, we develop numerical and asymptotic analytical techniques to study the joint response of two cell populations (or a cell pair) receiving correlated inputs.…”