neural network

In biology, a network of neurons, or nerve cells, in a human being or other animal; in computer technology, a network of processors that attempts to mimic a biological neural network.

A basic network has three layers of processors: an input layer, an output layer, and a ‘hidden’ layer in between. Each processor is connected to others in the network by a system of ‘synapses’; each processor in the top layer connects to every one in the hidden layer, and each of these connects to every processor in the output layer. This means that each nerve cell in the middle and bottom layers receives input from several different sources; only when the amount of input exceeds a critical level does the cell fire an output signal. The chief characteristic of neural networks is their ability to sum up large amounts of imprecise data and decide whether they match a pattern or not. They can ‘learn’ to perform such tasks without needing to be explicitly programmed to do so. Networks of this type may be used in developing robot vision, matching fingerprints, and analysing fluctuations in stock-market prices.