Research map / Label-efficient and robust learning / Noisy labels
Part-dependent Label Noise: Towards Instance-dependent Label Noise
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 |
|---|---|---|---|
| model instance-dependent noise with part-dependent transition matrices and revise the matrix | PTD-R-V · CIFAR-10 IDN-50% | classification accuracy: 53.98$\pm$2.34 vs 44.72$\pm$7.72 | Helped |
| model instance-dependent noise with part-dependent transition matrices and revise the matrix | PTD-R-V · F-MNIST IDN-50% | classification accuracy: 75.96$\pm$1.38 vs 68.99$\pm$1.04 | Helped |
| model instance-dependent noise with part-dependent transition matrices and revise the matrix | PTD-R-V · SVHN IDN-50% | classification accuracy: 58.09$\pm$2.57 vs 49.02$\pm$4.33 | Helped |
| model instance-dependent noise with part-dependent transition matrices and revise the matrix | PTD-R-V · NEWS IDN-50% | classification accuracy: 62.77$\pm$0.98 vs 59.29$\pm$2.07 | Helped |
| model instance-dependent noise with part-dependent transition matrices and revise the matrix | PTD-R-V · MNIST IDN-50% | classification accuracy: 88.67$\pm$1.25 vs 87.52$\pm$1.03 | Helped |
| model instance-dependent noise with part-dependent transition matrices and revise the matrix | PTD-R-V · Clothing1M | classification accuracy: 71.67 vs 70.97 | Helped |
| add transition-matrix revision (ablation) | PTD-R on CIFAR-10 · CIFAR-10 IDN-50% | classification accuracy: 49.07$\pm$2.56 -> 53.98$\pm$2.34 | Helped |
| use Reweight rather than Forward loss correction (ablation) | PTD-F-V on CIFAR-10 · CIFAR-10 IDN-50% | classification accuracy: 48.76$\pm$2.01 -> 53.98$\pm$2.34 | Helped |
| vary the number of parts from r=10 to r=20 (ablation) | PTD-R-V · CIFAR-10 IDN-50% | classification accuracy: 53.78$\pm$2.77 to 53.98$\pm$2.34 | No clear effect |
| use part-dependent rather than class-dependent transition matrices (ablation) | PTD · CIFAR-10 IDN-50% | transition-matrix approximation error: 0.945$\pm$0.051 -> 0.840$\pm$0.030 | Helped |
Closest papers by meaning
- Confidence Scores Make Instance-dependent Label-noise Learning Possible (ICML 2021 · similarity 0.92)
- Instance-Dependent Label-Noise Learning With Manifold-Regularized Transition Matrix Estimation (CVPR 2022 · similarity 0.92)
- Estimating Noisy Class Posterior with Part-level Labels for Noisy Label Learning (CVPR 2024 · similarity 0.90)
- Learning Causal Transition Matrix for Instance-dependent Label Noise (AAAI 2025 · similarity 0.90)
- Learning the Latent Causal Structure for Modeling Label Noise (NeurIPS 2024 · similarity 0.89)
- Label Noise: Ignorance Is Bliss (NeurIPS 2024 · similarity 0.87)
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.