Handwritten digit recognition using an optimized nearest neighbor classifier

In this paper we present a method for handwritten digit recognition using a nearest neighbor classifier. In this method a set of prototypes are obtained from the training samples and used to build a nearest neighbor classifier. The classifier is then mapped to a multi-layer perceptron. After training the neural network is mapped back to a nearest neighbor classifier with new and optimized prototypes. Using this method we were able to obtain recognition accuracy from 93.7% with 100 prototypes to 96.2% with 500 prototypes for samples not used for training. © 1994.