Instance-dependent noise, label reconstruction, sample selection, identity mapping
Exemplar-free class-incremental learning, self-supervised multitask learning, training-time augmentation, test-time augmentation