Learning to Rank Recommendations with the k-Order Statistic Loss

   Abstract