Neural network refers to a system of input, hidden, and output units that is capable of learning if the mathematical weights among the units are systematically modified either according to Hebb's rule or by back-propagation. Please see also Hebb's rule, Connectionism. And Back-propagation systems