Neural network models for pattern recognition and associative memory
Written by Gail A. Carpenter
This review outlines some fundamental neural network modules for associative memory, pattern recognition, and category learning. Included are discussions of the McCulloch-Pitts neuron, perceptrons, adaline and madaline, back propagation, the learning matrix, linear associative memory, embedding fields, instars and outstars, the avalanche, shunting competitive networks, competitive learning, computational mapping by instarl outstar families, adaptive resonance theory, the cognitron and neocognitron, and simulated annealing. Adaptive filter formalism provides a unified notation. Activation laws include additive and shunting equations. Learning laws include back-coupled error correction, Hebbian learning, and gated instar and outstar equations. Also included are discussions of real-time and off-line modeling, stable and unstable coding, supervised and unsupervised learning, and self-organization.