Fast Propagation of Firing Rates through Layered Networks of Noisy Neurons

We model the propagation of neural activity through a feedforward network consisting of layers of integrate-and-fire neurons. In the presence of a noisy background current and spontaneous background firing, firing rate modulations are transmitted linearly through many layers, with a delay proportional to the synaptic time constant and with little distortion. Single neuron properties and firing statistics are in agreement with physiological data. The proposed mode of propagation allows for fast computation with population coding based on firing rates, as is demonstrated with a local motion detector.