On the Selection of the Globally Optimal Prototype Subset for Nearest-Neighbor Classification | Publicación