AI Paper Atlas

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

Class-Dependent Label-Noise Learning with Cycle-Consistency Regularization

De Cheng, Yixiong Ning, Nannan Wang, Xinbo Gao, Heng Yang, Yuxuan Du, Bo Han, Tongliang Liu

NeurIPS 2022 · Label-efficient and robust learning › Noisy labels · approach: Instance-dependent noise modeling · 2 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 backward transition matrix T' with L2 loss to forward loss L1 (ablation)Ours · CIFAR-100 Sym-60%test accuracy (%): 40.89 -> 41.94Improved
add cycle-consistency regularization on top of T + T' (ablation)Ours · CIFAR-100 Sym-60%test accuracy (%): 41.94 -> 49.30Improved
add cycle-consistency regularization on top of T + T' (ablation)Ours · CIFAR-10 Pair-45%test accuracy (%): 78.62 -> 91.08Improved
add cycle-consistency regularization on top of T + T' (ablation)Ours · CIFAR-10 Sym-60%test accuracy (%): 73.39 -> 81.01Improved
add cycle-consistency regularization on top of T + T' (ablation)Ours · CIFAR-100 Asym-60%test accuracy (%): 42.72 -> 50.29Improved
use only the backward loss L2(w,T') instead of only forward loss (ablation)Ours · CIFAR-10 Asym-20%test accuracy (%): 89.46 -> 89.97No clear effect
plug the proposed module into DivideMixDivideMix · Clothing1Maccuracy (%): 74.67 -> 75.12Improved
plug the proposed module into DivideMixDivideMix · Food-101Naccuracy (%): 84.39 -> 86.11Improved
replace VolMinNet by Ours when combined with DivideMixDivideMix · Clothing1Maccuracy (%): 74.83 -> 75.12Improved
replace CE training by OursResNet-50 · Clothing1Maccuracy (%): 68.94 -> 70.73Improved
replace CE training by OursResNet-50 · Food-101Naccuracy (%): 81.44 -> 83.71Improved
set lambda to 0.3 (ablation)Ours · CIFAR-10/100 hyper-parameter studytest accuracy (%)Improved

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