AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

The Role of Entropy and Reconstruction in Multi-View Self-Supervised Learning

Borja Rodríguez Gálvez, Arno Blaas, Pau Rodriguez, Adam Golinski, Xavier Suau, Jason Ramapuram, Dan Busbridge, Luca Zappella

ICML 2023 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Contrastive versus non-contrastive theory · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace objective with ER bound (SimCLR + ER)SimCLR · ImageNet linear evalTop-1 accuracy: +0.72Improved
add ER bound to BYOL / replace DINO loss with ERBYOL and DINO · ImageNet linear evalTop-1 accuracy: -1.5 (BYOL), -1.65 (DINO)Hurt
train with ER bound instead of original lossBYOL, DINO, SimCLR · ImageNet, batch size 512 degradationAccuracy drop from bs 4096 to 512: BYOL 23.65 -> 2.35; DINO 6.76 -> 2.35; SimCLR 2.17 -> 1.01Improved
train with ER bound instead of original lossBYOL, DINO · ImageNet, EMA coefficient 0.8Accuracy drop: DINO 8.25 -> 0.92; BYOL 2.63 -> 0.41Improved
remove softmax centering from DINO (with and without ER) (ablation)DINO · ImageNetentropy and accuracyHurt
add ER bound to BYOL (entropy maximization check)BYOL · ImageNet entropy trajectoryentropy during trainingNo clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy