On Linear Separation Capacity of Self-Supervised Representation Learning

Open in new window