VICReg: Variance-Invariance-Covariance Regularization for Self-Supervised Learning

TL;DR

VICReg introduces a regularization-based approach to prevent collapse in self-supervised image representation learning, improving stability and performance by explicitly controlling embedding variance and decorrelation.

Contribution

The paper proposes VICReg, a novel self-supervised learning method that explicitly avoids collapse through variance and covariance regularization, enhancing training stability and downstream task performance.

Findings

Achieves state-of-the-art results on several downstream tasks.

Incorporating the variance term improves training stability.

Method effectively prevents embedding collapse.

Abstract

Recent self-supervised methods for image representation learning are based on maximizing the agreement between embedding vectors from different views of the same image. A trivial solution is obtained when the encoder outputs constant vectors. This collapse problem is often avoided through implicit biases in the learning architecture, that often lack a clear justification or interpretation. In this paper, we introduce VICReg (Variance-Invariance-Covariance Regularization), a method that explicitly avoids the collapse problem with a simple regularization term on the variance of the embeddings along each dimension individually. VICReg combines the variance term with a decorrelation mechanism based on redundancy reduction and covariance regularization, and achieves results on par with the state of the art on several downstream tasks. In addition, we show that incorporating our new variance term into other methods helps stabilize the training and leads to performance improvements.