Research map / Label-efficient and robust learning / Noisy labels
Joint Optimization Framework for Learning With Noisy Labels
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 |
|---|---|---|---|
| jointly optimize network parameters and reassign training labels | PreAct ResNet-32 · SN-CIFAR, symmetric noise | test accuracy: 36.6 -> 86.0 | Helped |
| jointly optimize network parameters and reassign training labels | PreAct ResNet-32 · AN-CIFAR, asymmetric noise 40% | test accuracy: 86.4 -> 91.7 | Helped |
| jointly optimize network parameters and reassign training labels | ResNet-50 · Clothing1M test | accuracy: 69.84 -> 72.23 | Helped |
| use soft-label updates instead of hard-label updates (ablation) | label optimization method · SN-CIFAR, noise rate 0.7 | test accuracy: 85.7 -> 86.0 | Helped |
| use soft-labels rather than hard-labels (ablation) | label optimization method · SN-CIFAR, noise rate 0.7 | recovery accuracy: 85.9% -> 86.0% | No clear effect |
| train with a high learning rate rather than a low learning rate (ablation) | CNN trained on noisy SN-CIFAR · SN-CIFAR | test accuracy over training: null | Helped |
| use pseudo-labels recovered by the method instead of ground-truth labels (ablation) | PreAct ResNet-32 · PL-CIFAR | test accuracy: null | Hurt |
| start label updating at epoch 0 instead of epoch 70 (ablation) | joint optimization · AN-CIFAR, noise rate 0.4 | validation accuracy: 58.4 vs 91.3 | Hurt |
| vary the learning rate from 0.03 to 0.2 (ablation) | joint optimization · AN-CIFAR, noise rate 0.4 | validation accuracy: 92.0 vs 88.5 | Hurt |
| increase beta from 0.8 to 2.0 (ablation) | joint optimization · SN-CIFAR, noise rate 0.7 | validation accuracy: 85.5 vs 30.3 | Hurt |
Closest papers by meaning
- Learning with Structural Labels for Learning with Noisy Labels (CVPR 2024 · similarity 0.92)
- Toward Robustness against Label Noise in Training Deep Discriminative Neural Networks (NeurIPS 2017 · similarity 0.91)
- Coresets for Robust Training of Deep Neural Networks against Noisy Labels (NeurIPS 2020 · similarity 0.91)
- Augmentation Strategies for Learning With Noisy Labels (CVPR 2021 · similarity 0.90)
- Distilling Effective Supervision From Severe Label Noise (CVPR 2020 · similarity 0.90)
- Handling Label Noise via Instance-Level Difficulty Modeling and Dynamic Optimization (NeurIPS 2025 · similarity 0.88)
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.