Research map / Label-efficient and robust learning / Noisy labels
IRNet: Iterative Refinement Network for Noisy Partial Label Learning
Zheng Lian, Mingyu Xu, Lan Chen, Licai Sun, Bin Liu, Lei Feng, Jianhua Tao
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 |
|---|---|---|---|
| add IRNet to PiCO | PiCO · CIFAR-10 (q=0.3, eta=0.3) | accuracy: 82.25 -> 91.35 | Improved |
| add IRNet to CC | CC · CIFAR-10 (q=0.1, eta=0.3) | accuracy: 84.46 -> 89.92 | Improved |
| add IRNet to PiCO | PiCO · CIFAR-10 (q=0.5, eta=0.3) | accuracy: 68.75 -> 86.19 | Improved |
| add IRNet to PiCO | PiCO · Kuzushiji-MNIST (last-epoch gap) | best-last accuracy gap: 9.42%~12.46% -> 0.12%~0.14% | Improved |
| add IRNet to PiCO | PiCO · CIFAR-100 (q=0.01, eta=0.3) | best accuracy: 58.97 -> 68.77 | Improved |
| add weak augmentation smoothness constraint (ablation) | IRNet · CIFAR-10 | accuracy: 0.50%~1.90% improvement; 90.78 -> 92.38 at q=0.1 | Improved |
| add weak augmentation smoothness constraint (ablation) | IRNet · CIFAR-100 (q=0.01) | accuracy: 65.40 -> 68.77 | Improved |
| replace weak with strong augmentation (ablation) | IRNet · CIFAR-10 (q=0.1) | accuracy: 92.38 -> 91.43 | Hurt |
| use two strong augmentations instead of none (ablation) | IRNet · CIFAR-100 (q=0.01) | accuracy: 65.40 -> 61.70 | Hurt |
| increase weak augmentations from 1 to 2 (ablation) | IRNet · CIFAR-10 (q=0.1) | accuracy: 92.38 -> 92.68 | Mixed |
| add swapping (move lowest-confidence label out of candidates) (ablation) | IRNet · CIFAR-10 (q=0.5) | accuracy: 86.19 -> 84.15 | Hurt |
| add swapping (move lowest-confidence label out of candidates) (ablation) | IRNet · CIFAR-100 (q=0.05) | accuracy: 68.09 -> 67.19 | Hurt |
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.
- PiCO+ — PiCO: Contrastive Label Disambiguation for Partial Label Learning (ICLR 2022) · better on 9 of 9 shared results
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
- Mutual Partial Label Learning with Competitive Label Noise (ICLR 2023 · similarity 0.89)
- Realistic Evaluation of Deep Partial-Label Learning Algorithms (ICLR 2025 · similarity 0.88)
- Adaptive Integration of Partial Label Learning and Negative Learning for Enhanced Noisy Label Learning (AAAI 2024 · similarity 0.88)
- Partial Label Learning with a Partner (AAAI 2024 · similarity 0.86)
- Conformal Prediction for Partial Label Learning (AAAI 2025 · similarity 0.86)
- Distilling Reliable Knowledge for Instance-Dependent Partial Label Learning (AAAI 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.
