Classification Using NN Method based on Mean of Norm in Closest Prototype from Input Pattern and Its κ Neighbor Prototypes

Abstract

For accurate classification in arbitrary distributed patterns when the number of patterns is very low, we have created a classification method using the NN method based on the mean of norm in the closest prototype from an input pattern and its κ neighbor prototypes. Because this method takes into consideration the weight by the difference of variance in prototypes around the discrimination boundary, it can be used to precisely classify input patterns when the number of patterns is very low.