English

Modulate Your Spectrum in Self-Supervised Learning

Machine Learning 2024-01-23 v2 Computer Vision and Pattern Recognition Signal Processing

Abstract

Whitening loss offers a theoretical guarantee against feature collapse in self-supervised learning (SSL) with joint embedding architectures. Typically, it involves a hard whitening approach, transforming the embedding and applying loss to the whitened output. In this work, we introduce Spectral Transformation (ST), a framework to modulate the spectrum of embedding and to seek for functions beyond whitening that can avoid dimensional collapse. We show that whitening is a special instance of ST by definition, and our empirical investigations unveil other ST instances capable of preventing collapse. Additionally, we propose a novel ST instance named IterNorm with trace loss (INTL). Theoretical analysis confirms INTL's efficacy in preventing collapse and modulating the spectrum of embedding toward equal-eigenvalues during optimization. Our experiments on ImageNet classification and COCO object detection demonstrate INTL's potential in learning superior representations. The code is available at https://github.com/winci-ai/INTL.

Keywords

Cite

@article{arxiv.2305.16789,
  title  = {Modulate Your Spectrum in Self-Supervised Learning},
  author = {Xi Weng and Yunhao Ni and Tengwei Song and Jie Luo and Rao Muhammad Anwer and Salman Khan and Fahad Shahbaz Khan and Lei Huang},
  journal= {arXiv preprint arXiv:2305.16789},
  year   = {2024}
}

Comments

Accepted at ICLR 2024. The code is available at https://github.com/winci-ai/intl

R2 v1 2026-06-28T10:47:21.807Z