Improving Self-Supervised Learning by Characterizing Idealized Representations

Open in new window