Research map / Label-efficient and robust learning / Noisy labels
Extended : Learning with Mixed Closed-set and Open-set Noisy Labels
Xiaobo Xia, Bo Han, Nannan Wang, Jiankang Deng, Jiatong Li, Yinian Mao, Tongliang Liu
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 changed | Where | Measured | Effect |
|---|---|---|---|
| use 3 cluster-dependent transition matrices instead of 1 (ablation) | Extended T · MS1MV0 face verification (average) | accuracy (%): 89.62 -> 90.93 | Improved |
| use 2 cluster-dependent transition matrices instead of 1 (ablation) | Extended T · MS1MV0 face verification (average) | accuracy (%): 89.62 -> 90.42 | Improved |
| use 3 clusters instead of 1 (ablation) | Extended T · VggFace-2 face verification (average) | accuracy (%): 91.30 -> 91.51 | No clear effect |
| use 3 clusters instead of 1 (ablation) | Extended T · CIFAR-10+SVHN (tau=0.8, rho=0.25) | accuracy (%): 83.68 -> 84.42 | Improved |
| use 3 clusters instead of 1 (ablation) | Extended T · CIFAR-10+CIFAR-100 (tau=0.8, rho=0.75) | accuracy (%): 84.62 -> 84.59 | No clear effect |
| extend transition matrix to open-set noise vs T-Revision (closed-set only) | Extended T · CIFAR-10+CIFAR-100 (tau=0.8, rho=0.75) | accuracy (%): 82.27 -> 84.62 | Improved |
| extend transition matrix to open-set noise vs T-Revision (closed-set only) | Extended T · CIFAR-10+ImageNet32 (tau=0.8, rho=0.75) | accuracy (%): 81.37 -> 84.57 | Improved |
| extend transition matrix to open-set noise vs T-Revision (closed-set only) | Extended T · CIFAR-10+SVHN (tau=0.8, rho=0.75) | accuracy (%): 82.98 -> 84.83 | Improved |
| use GCE instead of Extended T at low noise | GCE · CIFAR-10+CIFAR-100 (tau=0.4, rho=0.25) | accuracy (%): 89.61 vs 89.52 | Hurt |
| extend transition matrix to open-set noise vs T-Revision | Extended T · MS1MV0 face verification (average) | accuracy (%): 89.55 -> 89.62 | No clear effect |
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.
- Co-teaching — Co-teaching: Robust training of deep neural networks with extremely noisy labels (NeurIPS 2018) · better on 1 of 1 shared result
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
- Dual T: Reducing Estimation Error for Transition Matrix in Label-noise Learning (NeurIPS 2020 · similarity 0.94)
- Estimating Noise Transition Matrix with Label Correlations for Noisy Multi-Label Learning (NeurIPS 2022 · similarity 0.94)
- Estimation of the Label-Noise Transition Matrix with Performance Guarantees via Selective Classification (NeurIPS 2026 · similarity 0.92)
- Learning the Latent Causal Structure for Modeling Label Noise (NeurIPS 2024 · similarity 0.91)
- Estimating Noisy Class Posterior with Part-level Labels for Noisy Label Learning (CVPR 2024 · similarity 0.91)
- Learning Causal Transition Matrix for Instance-dependent Label Noise (AAAI 2025 · similarity 0.91)
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.
