Topic: representation-learning
-
Contrastive self-supervised learning: pairs, augmentation and collapse
Learn representations by comparing related and contrasting views, choose transformations that preserve useful information, and evaluate representations beyond the training loss.
Machine-readable: JSON