VISReg: Variance-Invariance-Sketching Regularization for JEPA training 事件

VISReg: Variance-Invariance-Sketching Regularization for JEPA training arXiv:2606.02572v1 Announce Type: new Abstract: Self-supervised learning methods prevent embedding collapse via modeling heuristics or explicit regularization of the embedding space. Among the latter, VICReg decomposes regularization into variance and covariance objectives, offering flexibility and interpretability. However, covariance captures only second-order statistics -- encouraging decorrelation but failing to enforce