Dictionary:Neural net, neural network
Artificial neural net. A pattern recognition network that is used to divide data into classes (see classification). The network consists of three layers of nodes (input, hidden, and output layers) that are interconnected by artificial neurons that develop weights that increase or decrease the signals that come over them. Each node fires an impulse when the sum of its inputs exceeds a threshold. Supervised classification involves using a training set (learning set) consisting of examples of inputs and corresponding desired outputs to ‘‘train’’ the network. Training involves adjusting the weights of the artificial neurons so that the desired outputs are achieved. Unsupervised classification seeks to find the natural classes into which the data may be subdivided. Useful for classification of nonlinear, statistical, large, poorly understood data sets.