An optimal basis for feature extraction with support vector machine classification using the radius-margin bound Conferences