AI Paper Atlas

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

Foster Adaptivity and Balance in Learning with Noisy Labels

Mengmeng Sheng, Zeren Sun, Tao Chen, Shuchao Pang, yucheng wang, Yazhou Yao

ECCV 2024 · Label-efficient and robust learning › Noisy labels · approach: Noise-aware loss and pairwise methods · 16 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 self-adaptive class-balanced sample selection (SCS) to standard training (ablation)Standard training · CIFAR100N Sym-50%test accuracy (%): 34.10 -> 58.21Improved
add local (class-specific) thresholds to SCS (ablation)SCS · CIFAR100N Sym-50%test accuracy (%): 53.36 -> 58.21Improved
add global threshold to SCS (ablation)SCS · CIFAR100N Sym-50%test accuracy (%): 55.64 -> 58.21Improved
add EMA to threshold estimation (ablation)SCS · CIFAR100N Sym-50%test accuracy (%): 54.72 -> 58.21Improved
add self-adaptive class-balanced re-weighting (SCR) (ablation)Standard+SCS · CIFAR100N Sym-50%test accuracy (%): 58.21 -> 60.43Improved
add confidence-based re-weighting inside SCR (ablation)Standard+SCS+SCR · CIFAR100N Sym-50%test accuracy (%): 59.75 -> 60.43Improved
add consistency regularization on clean samples (ablation)Standard+SCS+SCR · CIFAR100N Sym-50%test accuracy (%): 60.43 -> 62.65Improved
add class-balanced design (ablation)SED · CIFAR100N Asym-40%test accuracy (%): 52.78 -> 58.80Improved
add class-balanced design (ablation)SED · CIFAR100N Sym-80%test accuracy (%): 38.08 -> 39.32Improved
use SED instead of DISC (best prior)noisy label learning · Web-Aircraft/Bird/Car averagetest accuracy (%): 84.89 -> 85.83Improved
use SED instead of Co-LDLnoisy label learning · CIFAR100N Asym-40%test accuracy (%): 52.28 -> 58.29Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 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