Neural Networks

Abstract

Neural networks, as a discipline, studies the information processing capabilities of networks — “neural networks” — of simple processors which are in some way like the living neurons of the brain. It uses a distributed representation of the information stored in the network, so leading to robustness against damage and corresponding fault tolerance. Training is a very important component of neural networks. By this process there is modification of the strengths with which one neuron affects another - the so-called connection weights - during exposure to an appropriate environment as opposed to programming in the required responses of the network to inputs from the environment. This feature is very important, since a neural network can thereby be trained to give a desired response to a set of inputs even though there are no explicit rules for this response to be achieved.