Contribution to the probabilistic theory of neural nets: III. Specific inhibition

Abstract

The input-output formula is derived for a neuron upon which converge the axones of two other neurons (one excitatory, the other inhibitory) which are themselves subjected to a “Poisson shower” of excitatory stimuli. If the period of latent inhibition, σ, does not exceed one half the refractory period, δ, the input-output curve has no maximum. If, however, σ>δ/2, a maximum exists in the input-output curve. As the outside frequencyx increases without bound, the output frequencyx ₃ approaches an asymptotic value which ranges from 1/δ to 0, depending on the ratio σ/δ. The maximum output (if it exists) is also derived as a function of σ and δ.