Distribution of orientation selectivity in recurrent networks of spiking neurons with different random topologies.

Sadra Sadeh,Stefan Rotter

doi:10.1371/journal.pone.0114237

Abstract

Neurons in the primary visual cortex are more or less selective for the orientation of a light bar used for stimulation. A broad distribution of individual grades of orientation selectivity has in fact been reported in all species. A possible reason for emergence of broad distributions is the recurrent network within which the stimulus is being processed. Here we compute the distribution of orientation selectivity in randomly connected model networks that are equipped with different spatial patterns of connectivity. We show that, for a wide variety of connectivity patterns, a linear theory based on firing rates accurately approximates the outcome of direct numerical simulations of networks of spiking neurons. Distance dependent connectivity in networks with a more biologically realistic structure does not compromise our linear analysis, as long as the linearized dynamics, and hence the uniform asynchronous irregular activity state, remain stable. We conclude that linear mechanisms of stimulus processing are indeed responsible for the emergence of orientation selectivity and its distribution in recurrent networks with functionally heterogeneous synaptic connectivity.

Highlights

When arriving at the cortex from the sensory periphery, sensory signals are further processed by local recurrent networks
In the primary visual cortex of mammals like carnivores and primates, for instance, it has been proposed that the recurrent network might be mainly responsible for the amplification of orientation selectivity [2, 3]
The network with these parameters operates in the fluctuation-driven regime, which shows asynchronousirregular (AI) dynamics (Fig. 1A), with low firing rates (Fig. 1B) and high variance of inter-spike intervals (ISI) (Fig. 1C)

Summary

Introduction

When arriving at the cortex from the sensory periphery, sensory signals are further processed by local recurrent networks. In the primary visual cortex of mammals like carnivores and primates, for instance, it has been proposed that the recurrent network might be mainly responsible for the amplification of orientation selectivity [2, 3]. A small bias provided by the feedforward afferents would be enough, and selectivity is amplified by a non-linear mechanism implemented by the recurrent network. This mechanism is a result of the feature-specific connectivity assumed in the model, where neurons with similar input selectivities are connected to each other with a higher probability. As nearby neurons are connected with a higher likelihood than distant neurons, feature-specific connectivity is a straight-forward result in this scenario

Methods

Results

Conclusion