Guillotine Regularization: Why removing layers is needed to improve generalization in Self-Supervised Learning

Open in new window