AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Prototypical Contrastive Learning of Unsupervised Representations

Junnan Li, Pan Zhou, Caiming Xiong, Steven Hoi

ICLR 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Contrastive representation theory · 469 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 instance-wise contrastive learning with PCLResNet-50 representations · ImageNet, 1% labelstop-5 accuracy: 56.9 -> 75.3Improved
Replace instance-wise contrastive learning with PCLResNet-50 representations · ImageNet linear classificationtop-1 accuracy: 60.6 -> 61.5Improved
Replace MoCo representations with PCL representationsImageNet kNN classification · ImageNetaccuracy: 47.1 -> 54.5Improved
Replace MoCo representations with PCL representationsImageNet feature clustering · ImageNet, k-means with k=25000AMI: 0.285 -> 0.410Improved
Replace MoCo pretraining with PCL pretrainingResNet-50-FPN Faster R-CNN · VOC07 test, detector trained on VOC07mAP: 66.4 -> 71.7Improved
Add the prototypical contrastive term (ablation)instance-only PCL objective · ImageNet, 1% labelstop-5 accuracy: 56.9 -> 75.3Improved
Add the prototypical contrastive term (ablation)instance-only PCL objective · ImageNet linear classificationtop-1 accuracy: 60.6 -> 61.5Improved
Add warm-up using the instance-wise loss (ablation)proto-only objective · ImageNet, 1% labelstop-5 accuracy: 60.7 -> 72.3Improved
Add warm-up using the instance-wise loss (ablation)proto-only objective · ImageNet linear classificationtop-1 accuracy: 60.4 -> 60.9Improved
Add warm-up using the instance-wise loss (ablation)instance + proto objective · ImageNet, 1% labelstop-5 accuracy: 74.6 -> 75.3Improved
Add warm-up using the instance-wise loss (ablation)instance + proto objective · ImageNet linear classificationtop-1 accuracy: 61.3 -> 61.5Improved

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