The effect of noise on a neural network with spiking neurons
Mario E. Inchiosa · AIP conference proceedings · 1993
We study a class of neural network associative memories which include noise and transmission delays, code information in the timing of spikes, use long‐range Hebbian couplings plus local, inhibitory couplings, and feature low, biologically realistic neuronal activity. Recall of a pattern consists of a synchronized, periodic firing of neurons. We find a Lyapunov functional for the noiseless network dynamics, and using statistical mechanics and numerical simulation, we find that noisy dynamics improves the network’s ability to discriminate stored from unknown patterns.