We study the phenomenon of stochastic resonance when the noise is generated
by a network of integrate-and-fire neurons. A network with two populations
of neurons (one excitatory and one inhibitory) and a dispersion of intrins
ic firing rates can display a great amount of variability and its output ca
n be used to improve the signal-to-noise ratio of another neuron that is re
ceiving a subthreshold signal. We compare the performance of this system to
that of a system using white noise with different distributions. We find n
etwork parameters for which the performance is similar to that of a system
receiving white noise. In other cases, when the network output displays an
oscillatory component, the signal-to-noise ratio has several peaks, We also
analyze the relation between the synchrony of the network and the signal-t
o-noise ratio. [S1063-651X(99)11502-2].