An introduction to speed independent circuit theory

A neuron model is given whose properties adhere as closely as possible to the known properties of living neurons. Essential use is made of the concept that information to the nerve is in the form of pulse-rate information. Since this kind of information can assume more than two values, the model becomes a multivalued-logic device with a threshold. The output value of the neuron model is equal to the summation of its inputs, provided the summation is greater than the threshold. With this model, it is possible to construct networks which display a logically stable output although the elements comprising the net are not themselves logically stable. Two example networks are given that also demonstrate this property. The first example network is composed of unreliable model neurons whose thresholds are independently changing between two values. Criteria are given to aid in the selection of other networks of this type. The second example network is composed of model neurons which undergo a common shift of threshold, over nearly the entire threshold range. An algorithm for finding other networks of this type is given. The examples are given in three- and four-valued logic, but the matrix methods utilized are extensible to any n-value logic.