LAUSR

Abstract Multiple sensory signal, such as various visual stimuli, can be simultaneously transmitted and processed through different cortical areas. Using a biologically inspired model, we investigate the mechanism underlying such multiplexing in cortex. The network model is comprised of five feedforward-connected neural population of excitatory (E) and inhibitory (I) spiking neurons, each representing a cortical area. Numerical results indicate that multiple input signals are independently transmitted through feedforward neural networks via stochastic resonance (SR), and the transmission of information between adjacent cortical areas is altered by E-I relation of local cortical circuits. Due to different intrinsic properties of neurons, excitatory population can induce stochastic resonance to respond to low-frequency signal, and inhibitory neurons tend to respond to high-frequency signal through resonance. The E-I coupled neural ensemble shows selectivity for different input signal, which is gated by the gain between excitatory and inhibitory neurons. To be specific, neural network is inclined to gate-on signal with low-frequency when the excitation exceeds inhibition, whereas the high-frequency signal is selected. Moreover, neural signal can be transmitted from area to area only when the input frequency coincides with the inherent frequency of receivers in posterior areas, thus the independent conduction pathway is established through resonance. The transmission efficiency largely depends on the gain between excitatory and inhibitory input. Mean-field theory is further applied to validate the multiplexing in cortical networks and demonstrate the effective transmission of multiple information via SR in cortical networks.