AI Paper Atlas

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

Long-Tailed Partial Label Learning via Dynamic Rebalancing

Feng Hong, Jiangchao Yao, Zhihan Zhou, Ya Zhang, Yanfeng Wang

ICLR 2023 · Label-efficient and robust learning › Recognition and long-tail · approach: Decoupled and calibrated long-tail training · 8 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
Add RECORDS dynamic rebalancingCORR · CIFAR-10-LT (ρ=50, q=0.3)Top-1 accuracy: 76.12 -> 82.57Improved
Add RECORDS dynamic rebalancingCORR · CIFAR-10-LT (ρ=100, q=0.7)Top-1 accuracy: 38.11 -> 57.46Improved
Add RECORDS dynamic rebalancingCORR · CIFAR-100-LT (ρ=50, q=0.03)Top-1 accuracy: 42.29 -> 48.06Improved
Add RECORDS dynamic rebalancingCORR · CIFAR-100-LT (ρ=100, q=0.07)Top-1 accuracy: 31.05 -> 38.65Improved
Add RECORDS dynamic rebalancingCORR · PASCAL VOCTop-1 accuracy: 24.43 -> 56.46Improved
Apply oracle logit adjustment during trainingCORR · CIFAR-10-LT (ρ=50, q=0.3)Top-1 accuracy: 76.12 -> 36.27Hurt
Apply oracle logit adjustment during trainingCORR · PASCAL VOCTop-1 accuracy: 24.43 -> 52.51Improved
Replace RECORDS with temperature-scaled oracle logit adjustment (ablation)CORR · CIFAR-100-LT (ρ=100, q=0.05)Top-1 accuracy: 34.09 -> 43.46Hurt
Estimate the class distribution from the latest epoch's predictions instead of momentum-updated batch features (ablation)CORR + RECORDS · CIFAR-100-LT (ρ=100, q=0.05)Top-1 accuracy: 37.14 vs. 40.59Hurt
Set momentum coefficient m to 1.0, eliminating prototype updates (ablation)CORR + RECORDS · LT-PLL benchmarksTop-1 accuracyHurt
Vary momentum coefficient m from the reported best value (ablation)CORR + RECORDS · CIFAR-100-LT (ρ=100, q=0.05)Top-1 accuracyHurt

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