A fast classifier for image data

摘要

A technique for nearest neighbor classification is described which is approximately three times as fast as the usual method. Although the new technique was developed specifically for image data, tests on simulated random data show a significant improvement can be achieved using periodic reordering of class attractors.