Research map / Label-efficient and robust learning / Noisy labels
Identifiability of Label Noise Transition Matrix
Read the paperCheck 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 |
|---|---|---|---|
| Replace weakly supervised features with SimCLR self-supervised features | HOC transition-matrix estimator · CIFAR-10 asymm. 0.3, asymm. 0.4, inst. 0.4, inst. 0.5, inst. 0.6 | transition-matrix estimation error: 14.51 $\pm$ 0.4 -> 4.42 $\pm$ 0.01; 15.2 $\pm$ 0.02 -> 4.41 $\pm$ 0.01; 8.39 $\pm$ 0.05 -> 2.91 $\pm$ 0.02; 6.91 $\pm$ 0 | Helped |
| Use IPIRM rather than SimCLR self-supervised features | HOC transition-matrix estimator · CIFAR-10 asymm. 0.3, asymm. 0.4, inst. 0.4, inst. 0.5, inst. 0.6 | transition-matrix estimation error: 4.42 $\pm$ 0.01 -> \textbf{3.73 $\pm$ 0.02}; 4.41 $\pm$ 0.01 -> \textbf{3.74 $\pm$ 0.01}; 2.91 $\pm$ 0.02 -> \textbf{2.4 | Helped |
| Use IPIRM rather than SimCLR features to estimate the transition matrix | Forward loss correction on CIFAR-10 · CIFAR-10 inst. 0.3, inst. 0.4, inst. 0.5, inst. 0.6 | test accuracy: 66.61 -> 73.24; 65.82 -> 72.54; 64.51 -> 71.33; 62.81 -> 69.42 | Helped |
| Initialize CE training with SimCLR features instead of random initialization | CE training on CIFAR-100 · CIFAR-100 inst. 0.3, inst. 0.4, inst. 0.5, inst. 0.6 | test accuracy: 43.47 -> 58.95; 35.17 -> 49.7; 27.07 -> 36.87; 18.25 -> 25.07 | Helped |
| Initialize CE training with IPIRM features instead of random initialization | CE training on CIFAR-100 · CIFAR-100 inst. 0.3, inst. 0.4, inst. 0.5, inst. 0.6 | test accuracy: 43.47 -> 64.92; 35.17 -> 56.18; 27.07 -> 43.75; 18.25 -> 30.36 | Helped |
| Use IPIRM rather than SimCLR initialization | CE training on CIFAR-100 · CIFAR-100 inst. 0.3, inst. 0.4, inst. 0.5, inst. 0.6 | test accuracy: 58.95 -> 64.92; 49.7 -> 56.18; 36.87 -> 43.75; 25.07 -> 30.36 | Helped |
Closest papers by meaning
- Estimating Noise Transition Matrix with Label Correlations for Noisy Multi-Label Learning (NeurIPS 2022 · similarity 0.92)
- Learning Causal Transition Matrix for Instance-dependent Label Noise (AAAI 2025 · similarity 0.91)
- Learning the Latent Causal Structure for Modeling Label Noise (NeurIPS 2024 · similarity 0.91)
- Estimation of the Label-Noise Transition Matrix with Performance Guarantees via Selective Classification (NeurIPS 2026 · similarity 0.90)
- Noisy Label Learning with Instance-Dependent Outliers: Identifiability via Crowd Wisdom (NeurIPS 2024 · similarity 0.88)
- 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.