Regret Bounds for Multilabel Classification in Sparse Label Regimes

Interestingly, the theoretical understanding of MLC learning problems is still limited. This might be due to the daunting presence of multiple and diverse loss functions that can be defined over subsets of labels.