Metric learning by directly minimizing the k-NN training error

More Info
expand_more