Comparative analysis of support vector machine and nearest boundary vector classifier

The paper will present the original NBV (Nearest Boundary Vector) classifier whose structure has been inspired by the structure of CP (Counter Propagation) neural network, which uses the methods applied in the minimum-distance classification while in its operation drawn on the idea of functioning of SVM (Support Vector Machines) classifiers. The classification algorithm which is used by it relies on the original concept of a set of Boundary Vectors. It is characterized by the possibility of creation of various shapes of decision-making regions and it enables effective multi-class recognition. Recognition efficiency of NBV classifier will be confronted with efficiency of SVM classifiers.