Skip to main content

Self-Supervised Learning Caps the Unbounded Gap

·38 words·1 min · Download pdf

Supervised deep learning has unbounded generalization gap in theory and shown empirically on few datasets. Below work shows that self-supervised deep learning is much more better than supervised learning and comes with theoretically bounded generalization gap! https://x.com/boazbaraktcs/status/1317987242980331520

Discussion