Fig. 1

Illustration of a two-dimensional feature space. Inset a shows the distance between two classes in a two-dimensional features space. The ellipses representing the classes are constructed according to the covariance of the two-dimensional data. The figure emphasizes the overlap of classes, which is a big challenge in pattern recognition. Inset b shows the six nearest neighbors of the marked target data point. NNS is based on the fraction of the neighbors from the same class as the target point