An Embedding-Dynamic Approach to Self-Supervised Learning

Suhong Moon, Domas Buracas, Seunghyun Park, Jinkyu Kim, John Canny

Research output: Chapter in Book/Report/Conference proceedingConference contribution


A number of recent self-supervised learning methods have shown impressive performance on image classification and other tasks. A somewhat bewildering variety of techniques have been used, not always with a clear understanding of the reasons for their benefits, especially when used in combination. Here we treat the embeddings of images as point particles and consider model optimization as a dynamic process on this system of particles. Our dynamic model combines an attractive force for similar images, a locally dispersive force to avoid local collapse, and a global dispersive force to achieve a globally-homogeneous distribution of particles. The dynamic perspective highlights the advantage of using a delayed-parameter image embedding (a la BYOL) together with multiple views of the same image. It also uses a purely-dynamic local dispersive force (Brownian motion) that shows improved performance over other methods and does not require knowledge of other particle coordinates. The method is called MSBReg which stands for (i) a Multiview centroid loss, which applies an attractive force to pull different image view embeddings toward their centroid, (ii) a Singular value loss, which pushes the particle system toward spatially homogeneous density, (iii) a Brownian diffusive loss. We evaluate downstream classification performance of MSBReg on ImageNet as well as transfer learning tasks including fine-grained classification, multi-class object classification, object detection, and instance segmentation. In addition, we also show that applying our regularization term to other methods further improves their performance and stabilize the training by preventing a mode collapse.

Original languageEnglish
Title of host publicationProceedings - 2023 IEEE Winter Conference on Applications of Computer Vision, WACV 2023
PublisherInstitute of Electrical and Electronics Engineers Inc.
Number of pages9
ISBN (Electronic)9781665493468
Publication statusPublished - 2023
Event23rd IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2023 - Waikoloa, United States
Duration: 2023 Jan 32023 Jan 7

Publication series

NameProceedings - 2023 IEEE Winter Conference on Applications of Computer Vision, WACV 2023


Conference23rd IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2023
Country/TerritoryUnited States

Bibliographical note

Publisher Copyright:
© 2023 IEEE.


  • Algorithms: Machine learning architectures
  • and algorithms (including transfer)
  • formulations

ASJC Scopus subject areas

  • Artificial Intelligence
  • Computer Science Applications
  • Computer Vision and Pattern Recognition


Dive into the research topics of 'An Embedding-Dynamic Approach to Self-Supervised Learning'. Together they form a unique fingerprint.

Cite this