System size resonance in attractor neural networks

We consider the dynamics of an attractor neural network of a finite size N, trained with two patterns, which is subject to the action of an external stimulus (or field). This field drives the system to one of the patterns or to another for alternating intervals of duration T. It is observed that, for not too strong fields, the response of the network to the evolving field is optimal for some finite size, decreasing for smaller or larger systems. This is the so-called system size resonance, already reported for the Ising model. The explanation of this results is related to the phenomenon of stochastic resonance.