On Pattern Separating Function in a Two-Layered Random Nerve Net with Feedforward Inhibitory Connections

In this paper, we derive a theory for investigating the pattern separating function in the nerve net. Then, we consider some effects of the stochastic parameters constructing the nerve net, firing rates of the second layer and input patterns on the pattern separating function through the theory. As a result, we reveal that a nerve net excellent in pattern separating function is obtained when the stochastic parameters are selected properly. We also show that the pattern separating function is enhanced by controlling the firing rate of the second layer to a small value. Furthermore, the input patterns are separated largely when the firing rates of them are large.