Metric learning by directly minimizing the k-NN training error.

Konstantin Chernoff, Marco Loog, Mads Nielsen

Browse the full ICPR paper archive.