Design and implementation of optimized nearest neighbor classifiers for handwritten digit recognition

A method is described for handwritten digit recognition based on an optimized nearest-neighbor classification rule. In this method, a set of prototypes is obtained from training samples and is used to build a nearest-neighbor classifier. The classifier is then mapped to a multilayer perceptron. After training, the neural network is mapped back to a nearest-neighbor classifier with new and optimized prototypes. The classification procedure can be efficiently implemented without any multiplications