A concept for advanced computer calculations developed by Alan Turing to mimic some of the operations of the neurons in a brain. Memory elements (neurons) are conceptually interconnected by multiple paths connected with on-off switches to emulate the synapses of the brain. The original intent was to build a data-processing machine.
Modern applications reduce the concept to structured digital software processing models. Repeated processing through a neural network allows the network to learn from the data it processes. The learned process obtained from a set of training data with solutions can then be applied to other data sets for which no solution exists. An oilfield example includes training a network with wireline log and core data and then using the network to interpret further log data in terms of the core data. Neural networks are also being used in seismic processing, geological mapping and petrophysical analysis.