Research map / Label-efficient and robust learning / Semi-supervised learning
Towards Discovering the Effectiveness of Moderately Confident Samples for Semi-Supervised Learning
Check 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 |
|---|---|---|---|
| Add gradient synchronization filtering for moderately confident samples | FixMatch · CIFAR-10, 40 labels | error rate (%): 13.81±3.37 -> 7.73±2.50 | Helped |
| Add prototype proximity filtering for moderately confident samples | FixMatch · CIFAR-100, 400 labels | error rate (%): 48.85±1.75 -> 42.08±1.10 | Helped |
| Add gradient synchronization filtering | FixMatch · SVHN, 1000 labels | error rate (%): 2.28±0.11 -> 2.38±0.10 | Hurt |
| Add prototype proximity filtering | FixMatch · SVHN, 1000 labels | error rate (%): 2.28±0.11 -> 2.39±0.12 | Hurt |
| Use GSF or PPF rather than FixMatch | FixMatch · CIFAR-100, 400 labels | error rate (%): 6.89% and 6.77% | Helped |
| Use GSF or PPF rather than FixMatch | FixMatch · CIFAR-100, 2500 labels | error rate (%): 1.97% and 1.87% | Helped |
| Use GSF or PPF rather than FixMatch | FixMatch · CIFAR-100, 10000 labels | error rate (%): 1.15% and 1.26% | Helped |
| Use moderate-confidence bounds of 0.75 and 0.95 rather than including all unlabeled samples (ablation) | GSF and PPF · CIFAR-10, 40- and 250-label settings | error rate | Helped |
| Discard highly confident samples by setting the upper bound to 1 (ablation) | GSF and PPF · CIFAR-10 ablation settings | error rate | Hurt |
| Increase the GSF synchronization threshold to 1, degenerating to FixMatch (ablation) | GSF · CIFAR-10, 40-label setting | error rate | Hurt |
| Vary the moderate-confidence lower or upper bound (ablation) | GSF and PPF · CIFAR-10, 250-label setting | error rate | No clear effect |
| Vary the GSF synchronization threshold (ablation) | GSF · CIFAR-10, 250-label setting | error rate | No clear effect |
Closest papers by meaning
- Revisiting Semi-Supervised Learning in the Era of Foundation Models (NeurIPS 2025 · similarity 0.89)
- CGMatch: A Different Perspective of Semi-supervised Learning (CVPR 2025 · similarity 0.88)
- Enhancing Semi-Supervised Learning via Representative and Diverse Sample Selection (NeurIPS 2024 · similarity 0.88)
- Rethinking Confidence Scores and Thresholds in Pseudolabeling-based SSL (ICML 2025 · similarity 0.88)
- Towards Realistic Model Selection for Semi-supervised Learning (ICML 2024 · similarity 0.87)
- InterLUDE: Interactions between Labeled and Unlabeled Data to Enhance Semi-Supervised Learning (ICML 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.