Extended nearest feature line classifier

Abstract

An extended nearest feature line (ENFL) classifier is proposed to handle the discrimination problems. The ENFL borrows the concept of feature line spaces from the nearest feature line (NFL) method, to make use of the information implied by the interaction between each pair of points in the same class. Instead of the NFL distance, a better distance metric is designed in the ENFL. The ENFL is very effective in the cases with a small training set. The experimental evaluation shows that in the given feature space, the ENFL consistently achieves better performance than NFL and conventional nearest neighbor methods. © Springer-Verlag Berlin Heidelberg 2004.