ADAPTIVE VOTING RULES FOR K-NEAREST NEIGHBORS CLASSIFIERS

A simple form of cooperation between the k-nearest neighbors (NN) appr oach to classification and the neural-like property of adaptation is e xplored. A tunable, high level k-nearest neighbors decision rule is de fined that comprehends most previous generalizations of the common maj ority rule. A learning procedure is developed that applies to this rul e and exploits those statistical features that can be induced from the training set. The overall approach is tested on a problem of handwrit ten character recognition. Experiments show that adaptivity in the dec ision rule may improve the recognition and rejection capability of sta ndard k-NN classifiers.