Pattern separating functioning of two-layered random nerve nets with feedforward inhibitory connections

A two-layered random nerve net with feedforward inhibitory connections has the function of separating input patterns. The function depends largely on the nerve connection form between the layers. The nerve connection form is specified by a distribution of the number of inputs of an element on the second layer. A theory is derived that is able to treat the pattern separating functioning in nerve nets with various nerve connection forms. Also, considered are properties of the pattern separating functioning in three nerve nets, in which the nerve connection forms are specified by the δ-, Poisson, and compound Poisson distribution, respectively. From this consideration, the effect of the nerve connection form on pattern separating functioning is shown as well as the influence of firing rate and of standard deviation of the threshold of an element on the second layer