Quadratic boundaries in N-N classifiers with dissimilarity-based representations

The performance of the nearest neighbor (N-N) classifiers with the dissimilarity-based representations is studied. The dissimilarity-based representations cause quadratic decision boundaries instead of usual piecewise linear boundaries. The correct classification ratio of the N-N classifiers is improved with the use of the dissimilarity-based representations when variations within class differ from each other and the dimensionality of patterns is high.