Empirical risk minimization, where the underlying loss function depends on a pair of data
points, covers a wide range of application areas in statistics including pairwise ranking and …