Single-Neuron Models

Abstract

An important aspect of the modeling of neural networks is the choice of the individual neuron model. When the neural network is very large the details of the modeling of the single neurons become less important. Even the fact that the outputs of the individual neurons are highly irregular can be neglected and some models consider only the average firing rate as the determining input or output value. For very large numbers of neurons, the firing activity thus becomes a rather deterministic process and the neural net shows a behavior comparable to the macroscopic parameters of a gas consisting of a large number of individual molecules. In this case temperature and pressure are the externally measurable parameters that reflect the intermolecular interactions. Although on the microscopic level everything behaves stochastically, at the macroscopic level the process can be described deterministically. Most neural network models have been extensively reviewed by Levine (1983). We will deal with such models in the next chapter.