AI Paper Atlas

Research map / Label-efficient and robust learning / Noisy labels

Neighborhood Collective Estimation for Noisy Label Identification and Correction

Jichang Li, Guanbin Li, Feng Liu, Yizhou Yu

ECCV 2022 · Label-efficient and robust learning › Noisy labels · approach: Noise-aware loss and pairwise methods · 57 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 NCNV neighborhood noise verification instead of GMM (ablation)NCE · CIFAR-10/100 meanaccuracy: 79.2 -> 85.1Improved
add NCNV instead of GMM (ablation)NCE · CIFAR-10 sym 0.8accuracy: 79.0 -> 94.1Improved
add NCNV instead of GMM (ablation)NCE · CIFAR-100 sym 0.8accuracy: 56.8 -> 65.2Improved
add NCLC neighborhood label correction instead of confidence thresholding 0.95 (ablation)NCE · CIFAR-10/100 meanaccuracy: 81.0 -> 85.1Improved
add NCLC instead of CT(0.95) (ablation)NCE · CIFAR-10 sym 0.8accuracy: 86.1 -> 94.1Improved
add mixup regularization on clean samples instead of plain cross-entropy (ablation)NCE · CIFAR-10/100 meanaccuracy: 77.8 -> 85.1Improved
add mixup regularization instead of cross-entropy (ablation)NCE · CIFAR-10 sym 0.8accuracy: 78.5 -> 94.1Improved
add consistency loss L^lab on relabeled samples (ablation)NCE · CIFAR-10/100 meanaccuracy: 80.8 -> 85.1Improved
add L^lab (ablation)NCE · CIFAR-10 sym 0.8accuracy: 86.7 -> 94.1Improved
add input perturbation (AutoAugment) in consistency loss (ablation)NCE · CIFAR-10/100 meanaccuracy: 80.4 -> 85.1Improved
add both NCNV and L^lab versus GMM without L^lab (ablation)NCE · CIFAR-10/100 meanaccuracy: 75.7 -> 85.1Improved
NCE versus best prior method at highest noiseNCE vs NGC · CIFAR-10 sym 0.9accuracy: 80.5 -> 88.4Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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