An adaptive algorithm for training of a Nearest Neighbor (NN) classifier is developed in this paper. This learning rule has got some similarity to the well-known LVQ method, but using the nearest centroid neighborhood concept to estimate optimal locations of the code-book vectors. The aim of this approach is to improve the performance of the standard LVQ algorithms when using a very small code-book. The behavior of the learning technique proposed here is experimentally compared to those of the plain k-NN decision rule and the LVQ algorithms.