Creating of Quasioptimal General Classifiers Based on Distance Function

In this paper the work of general classifiers based on distance function that work in small test sample conditions has been investigated. The way of identical estimations obtaining on small samples has been presented. Influence of the main factors that conditioned such classifiers work has been established. Basing on face recognition system example the nearest neighbors number has been determine. Such an approach allows one to make more reliable decisions.