Research map / Label-efficient and robust learning / Noisy labels
Learning From Noisy Labels by Regularized Estimation of Annotator Confusion
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| Add trace norm regularization to the joint noisy-label loss (ablation) | Our method · CIFAR-10, dense labels | accuracy: 80.29 \pm 0.65 -> 81.23 \pm 0.21 | Helped |
| Add trace norm regularization to the joint noisy-label loss (ablation) | Our method · CIFAR-10, dense labels | CM error: 1.37 \pm 0.12 -> 0.72 \pm 0.01 | Helped |
| Add trace norm regularization to the joint noisy-label loss (ablation) | Our method · CIFAR-10, 1 label per image | accuracy: 76.31\pm0.49 -> 77.65\pm0.31 | Helped |
| Add trace norm regularization to the joint noisy-label loss (ablation) | Our method · CIFAR-10, 1 label per image | CM error: 1.46 \pm 0.27 -> 1.22\pm0.01 | Helped |
| Use trace regularization (ablation) | Our method · MNIST, simulated annotators | classification accuracy | Helped |
| Use trace regularization (ablation) | Our method · MNIST, simulated annotators | CM estimation error | Helped |
| Use trace regularization (ablation) | Our method · MNIST with low label noise | classification accuracy | No clear effect |
| Use an individual confusion matrix for each annotator rather than one shared confusion matrix | Noise-robust classifier · CIFAR-10, simulated annotators | accuracy | Helped |
| Use the proposed joint trace-regularized estimator instead of MBEM | CIFAR-10 · CIFAR-10, dense labels | accuracy: 73.33 \pm 0.46 -> 81.23 \pm 0.21 | Helped |
| Use the proposed joint trace-regularized estimator instead of MBEM | CIFAR-10 · CIFAR-10, 1 label per image | accuracy: 55.97 \pm 1.23 -> 77.65\pm0.31 | Helped |
| Set trace regularizer scale to values in {0.1, 0.01, 0.001} (ablation) | Our method · MNIST, mean annotator skill p=0.35 | validation accuracy | No clear effect |
Closest papers by meaning
- Is one annotation enough? - A data-centric image classification benchmark for noisy and ambiguous label estimation (NeurIPS 2022 · similarity 0.88)
- Collaborative Refining for Learning from Inaccurate Labels (NeurIPS 2024 · similarity 0.87)
- Noisy Label Learning with Instance-Dependent Outliers: Identifiability via Crowd Wisdom (NeurIPS 2024 · similarity 0.87)
- Learning from Noisy Labels via Conditional Distributionally Robust Optimization (NeurIPS 2024 · similarity 0.87)
- Pre-Trained Vision-Language Models as Noisy Partial Annotators (AAAI 2025 · similarity 0.86)
- Estimating Noisy Class Posterior with Part-level Labels for Noisy Label Learning (CVPR 2024 · similarity 0.86)
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.