What shapes the loss landscape of self supervised learning?

Liu Ziyin, Ekdeep Singh Lubana, Masahito Ueda, Hidenori Tanaka

Keywords: loss landscape, self-supervised learning, collapse

Abstract: Prevention of complete and dimensional collapse of representations has recently become a design principle for self-supervised learning (SSL). However, questions remain in our theoretical understanding: When do those collapses occur? What are the mechanisms and causes? We answer these questions by deriving and thoroughly analyzing an analytically tractable theory of SSL loss landscapes. In this theory, we identify the causes of the dimensional collapse and study the effect of normalization and bias. Finally, we leverage the interpretability afforded by the analytical theory to understand how dimensional collapse can be beneficial and what affects the robustness of SSL against data imbalance.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

TL;DR: We analytically solve the loss landscape of self-supervised learning and identify the causes of complete and dimensional collapse

Please Choose The Closest Area That Your Submission Falls Into: Deep Learning and representational learning