Research map / Label-efficient and robust learning / Recognition and long-tail
Making Partial-Label Datasets Easier: A Simple Yet Highly Effective Data Augmentation for Deep Partial-Label Learning
Dong-Dong Wu, Zhaoyi Li, Xiang Li, Zhiqiang Shen
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 UMiP augmentation | PRODEN · SVHN instance-dependent q1=1.0 | test accuracy: 83.74 -> 95.64 | Improved |
| add UMiP augmentation | PRODEN · SVHN instance-dependent q1=0.6 | test accuracy: 91.79 -> 95.82 | Improved |
| add UMiP augmentation | LWS · CUB-200 uniform | test accuracy: +4.57%, +6.53%, +10.2% | Improved |
| add UMiP augmentation | CRDPLL · SVHN instance-dependent q1=0.6 | test accuracy: 79.45 -> 79.70 | No clear effect |
| add UMiP augmentation | PaPi · CIFAR-10 instance-dependent q1=1.0 | test accuracy: 52.25 -> 60.02 | Improved |
| add UMiP augmentation | PRODEN · Lost real-world | test accuracy: 71.15 -> 73.81 | Improved |
| use UMiP instead of PLDA augmentation | LWS · Lost real-world | test accuracy: 74.12 -> 76.28 | Improved |
| add UMiP augmentation | instance-dependent CIFAR-10 high ambiguity · local ambiguity degree | LAD: 0.993 -> 0.974 | Improved |
| add UMiP augmentation | original model · SVHN instance-dependent | expected calibration error: 11.01 -> 5.54 | Improved |
| add label renormalization step to UMiP (ablation) | PRODEN, LWS, PiCO, ABLE, DIRK, PaPi · uniform PLL q3=0.9 | test accuracy: up to 8.85% | Improved |
| add warm-up epochs before enabling UMiP (ablation) | PRODEN · SVHN q1=0.8 | final accuracy: 95.82 -> 95.97 | No clear effect |
| change Beta parameter alpha away from 1.0 (0.5 or 2.0) (ablation) | UMiP · SVHN q1=0.6 | test accuracy | 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 2 of 2 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
- Distilling Reliable Knowledge for Instance-Dependent Partial Label Learning (AAAI 2024 · similarity 0.91)
- Adaptive Integration of Partial Label Learning and Negative Learning for Enhanced Noisy Label Learning (AAAI 2024 · similarity 0.90)
- Realistic Evaluation of Deep Partial-Label Learning Algorithms (ICLR 2025 · similarity 0.89)
- Navigating Real-World Partial Label Learning: Unveiling Fine-Grained Images with Attributes (AAAI 2024 · similarity 0.89)
- Partial Label Causal Representation Learning for Instance-Dependent Supervision and Domain Generalization (AAAI 2025 · similarity 0.89)
- Controller-Guided Partial Label Consistency Regularization with Unlabeled Data (AAAI 2024 · similarity 0.89)
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.
