Frequency based computation in neural networks

A model of neural network computation is presented in which the frequency of firing of a neuron is the basis upon which numerical values are measured. The author first discusses the model of the individual firing units, and then presents the constructions of the modules which purport to compute the arithmetic operations. A large network was built using these modules. It can be used to find near-optimal solutions to the task assignment problem. The proposed model is shown to produce excellent results.<>