Artificial neural networks for pattern recognition
Simon A. Corne · Concepts in Magnetic Resonance · 1996
Patterns of several kinds exist in NMR spectra. Examples include cross-peak shapes in multidimensional spectra and functional-group “fingerprints” in one-dimensional spectra. This article reviews the application of neural networks to solve pattern recognition problems in NMR. Neural network models are inspired by the highly parallel and interconnected organization of biological brains and their fault-tolerant processing capabilities. The most widely applied model, the multiplayer perceptron, which incorporates backpropagation learning, is described together with its predecessor, the perceptron; algorithms are presented for their training. Artificial neural network applications in one- and two-dimensional NMR are reviewed. © 1996 John Wiley & Sons, Inc.