Optimal training of thresholded linear correlation classifiers
Thomas Hildebrandt · IEEE Transactions on Neural Networks · 1991
A closed-form solution for improved pattern recognition that reduces the training time to a single epoch (one presentation of each of the training patterns) is presented. It is shown that the corresponding hardware requirements are no greater than those for regular recognition under certain conditions. A simple example which shows that the generalization obtained with the closed-form method exceeds that obtained by a model that admits only diagonal transformations is discussed.