Research map / Label-efficient and robust learning / Semi-supervised learning
Unsupervised Selective Labeling for More Effective Semi-Supervised Learning
Xudong Wang, Long Lian, Stella X. Yu
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 |
|---|---|---|---|
| select labeled instances with USL instead of random | FixMatch · CIFAR-10 (40 labels) | accuracy: 82.9 -> 90.4 | Improved |
| select labeled instances with USL-T instead of random | FixMatch · CIFAR-10 (40 labels) | accuracy: 82.9 -> 93.5 | Improved |
| select labeled instances with USL instead of random | MixMatch · CIFAR-10 (40 labels) | accuracy: 43.4 -> 61.6 | Improved |
| select with USL instead of random | SimCLRv2-CLD · CIFAR-10 (40 labels) | accuracy: 60.8 -> 76.6 | Improved |
| select with USL instead of random as budget grows | SimCLRv2-CLD · CIFAR-10 (250 labels) | accuracy: 79.4 -> 82.1 | Improved |
| select with USL instead of random as budget grows | FixMatch · CIFAR-10 (250 labels) | accuracy: 93.3 -> 94.0 | Improved |
| replace random selection by USL | FixMatch · CIFAR-100 (400 labels) | accuracy: 48.7 -> 55.1 | Improved |
| replace random selection by USL | SimCLRv2 · ImageNet-100 (400 labels) | accuracy: 62.2 -> 67.5 | Improved |
| replace random selection by USL-MoCo | FixMatch · ImageNet-1k (0.2% labels) | accuracy: 34.3 -> 48.6 | Improved |
| use CLIP features instead of MoCov2 features for USL selection (ablation) | FixMatch · ImageNet-1k (0.2% labels) | accuracy: 48.6 -> 47.5 | Hurt |
| use CLIP features instead of MoCov2 features for USL selection (ablation) | FixMatch · ImageNet-1k (1% labels) | accuracy: 61.6 -> 62.2 | Improved |
| use AL/SSAL selection (MMA+) instead of random in the 20 random + 20 selected setting | FixMatch · CIFAR-10 (40 labels) | accuracy: 82.9 -> 71.3 | Hurt |
Closest papers by meaning
- All Labels Are Not Created Equal: Enhancing Semi-Supervision via Label Grouping and Co-Training (CVPR 2021 · similarity 0.92)
- InterLUDE: Interactions between Labeled and Unlabeled Data to Enhance Semi-Supervised Learning (ICML 2024 · similarity 0.90)
- CGMatch: A Different Perspective of Semi-supervised Learning (CVPR 2025 · similarity 0.90)
- Learning with Partial-Label and Unlabeled Data: A Uniform Treatment for Supervision Redundancy and Insufficiency (ICML 2024 · similarity 0.90)
- OwMatch: Conditional Self-Labeling with Consistency for Open-World Semi-Supervised Learning (NeurIPS 2024 · similarity 0.89)
- Scaling Up Semi-supervised Learning with Unconstrained Unlabelled Data (AAAI 2024 · 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.
