Stochastic automaton models for the temporal pattern discrimination of nerve impulse sequences.

Abstract

The application of stochastic automata to the input-output relations of single neurons is considered. For this, some stochastic properties of temporal pattern discrimination in single synaptic cells are used to suggest stochastic automaton models. The models have only three possible states, the active, the absolute refractory and the relative refractory states, which are sufficient for temporal pattern sensitivity. From such an application, it was found that the temporal pattern discriminating structures in the models are similar to those used for experimental data and computer simulation (real-time neuron models). Extensions related to temporal pattern learning are discussed.