T-REGS: Minimum Spanning Tree Regularization for Self-Supervised Learning

TL;DR

T-REGS introduces a regularization method for self-supervised learning that uses the length of the Minimum Spanning Tree to prevent collapse and promote uniformity in learned representations, backed by theoretical and experimental validation.

Contribution

The paper proposes T-REGS, a novel regularization framework based on MST length, which theoretically and empirically improves representation quality in SSL.

Findings

T-REGS reduces dimensional collapse in learned features.

T-REGS promotes uniform distribution of representations.

Experimental results show improved performance on SSL benchmarks.

Abstract

Self-supervised learning (SSL) has emerged as a powerful paradigm for learning representations without labeled data, often by enforcing invariance to input transformations such as rotations or blurring. Recent studies have highlighted two pivotal properties for effective representations: (i) avoiding dimensional collapse-where the learned features occupy only a low-dimensional subspace, and (ii) enhancing uniformity of the induced distribution. In this work, we introduce T-REGS, a simple regularization framework for SSL based on the length of the Minimum Spanning Tree (MST) over the learned representation. We provide theoretical analysis demonstrating that T-REGS simultaneously mitigates dimensional collapse and promotes distribution uniformity on arbitrary compact Riemannian manifolds. Several experiments on synthetic data and on classical SSL benchmarks validate the effectiveness of our approach at enhancing representation quality.