multi-class classification problem is among the most popular and well-studied
statistical frameworks. Modern multi-class datasets can be extremely ambiguous
and single-output predictions fail to deliver satisfactory performance. By
allowing predictors to predict a set of label candidat