AI Paper Atlas

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

Self-Supervised Learning via Maximum Entropy Coding

Xin Liu, Zhongdao Wang, Ya-Li Li, Shengjin Wang

NeurIPS 2022 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Data-efficient and robust self-supervision · 14 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
Use maximum entropy coding with view consistency instead of established SSL objectivesResNet-50 pre-training · ImageNet val, linear probetop-1 accuracy: 70.6 vs 68.1Improved
Use maximum entropy coding pre-training instead of the strongest listed SSL baselineResNet-50 pre-training · ImageNet val, 1% labelstop-1 accuracy: 55.9 vs 55.0Improved
Use maximum entropy coding pre-training instead of MoCo v2ResNet-50 C4 · COCO detectionAP: +0.5 APImproved
Reduce pre-training from 100 to 50 epochs (ablation)MEC · ImageNet valtop-1 accuracy: 70.6 -> 67.5Hurt
Reduce batch size from 1024 to 128 (ablation)MEC · ImageNet valtop-1 accuracy: 70.6 -> 69.5Hurt
Set distortion εd² to 0.02 instead of 0.06 (ablation)MEC · ImageNet valtop-1 accuracy: 70.6 -> 70.5No clear effect
Set distortion εd² to 1.0 instead of 0.06 (ablation)MEC · ImageNet valtop-1 accuracy: 70.6 -> 69.5Hurt
Remove the momentum encoder (ablation)MEC · ImageNet valtop-1 accuracy: 70.6 -> 69.0Hurt
Use first-order instead of fourth-order Taylor expansion (ablation)MEC · ImageNet valtop-1 accuracy: 70.6 -> 70.0Hurt
Increase Taylor expansion order from 1 to 4 (ablation)MEC · COCO detectionAP: +1.5 APImproved
Use symmetric rather than asymmetric projector networks (ablation)MEC · ImageNet valtop-1 accuracy: 70.6 -> 70.1No clear effect
Increase projector dimension to 8192 (ablation)MEC · ImageNet valtop-1 accuracy: 70.6 -> 71.2Improved

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