Preventing Dimensional Collapse in Self-Supervised Learning via Orthogonality Regularization

Open in new window