A fast algorithm for the nearest-neighbor classifier

A fast algorithm that finds the nearest neighbor (NN) of an unknown sample from a design set of labeled samples is proposed. This algorithm requires a quite moderate preprocessing effort and a rather excessive storage, but it accomplishes substantial computational savings during classification. The performance of the algorithm is described and compared to the performance of the conventional one. Results on simulated data are provided to illustrate the computational savings that may be achieved using this fast algorithm.