Research map / Label-efficient and robust learning / Semi-supervised learning
Enhancing Sample Utilization through Sample Adaptive Augmentation in Semi-Supervised Learning
Guan Gui, Zhen Zhao, Lei Qi, Luping Zhou, Lei Wang, Yinghuan Shi
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 SAA to FixMatch | FixMatch · CIFAR-10 (40 labels) | accuracy (%): 92.50 -> 94.76 | Improved |
| add SAA to FlexMatch | FlexMatch · CIFAR-10 (40 labels) | accuracy (%): 95.01 -> 95.31 | Improved |
| add SAA to FixMatch | FixMatch · STL-10 (1000 labels) | accuracy (%): 85.27 -> 87.92 | Improved |
| add SAA to FlexMatch | FlexMatch · SVHN (40 labels) | accuracy (%): 92.42 -> 93.15 | Improved |
| apply diverse augmentation to all samples instead of only naive samples (ablation) | FixMatch + SAA · STL-10 (1000 labels) | accuracy (%): 85.27 -> 83.19 | Hurt |
| apply diverse augmentation to all samples (ablation) | FixMatch · CIFAR-10 (40 labels) | accuracy (%): 92.50 -> 92.98 | No clear effect |
| apply diverse augmentation to random 50% samples (ablation) | FixMatch · CIFAR-10 (40 labels) | accuracy (%): 92.50 -> 94.05 | Mixed |
| select naive samples with OTSU instead of fixed proportion (25%) (ablation) | SAA on FixMatch · CIFAR-10 (40 labels) | accuracy (%): 93.10 -> 94.50 | Improved |
| select naive samples with OTSU instead of fixed threshold (ablation) | SAA on FixMatch · STL-10 (1000 labels) | accuracy (%): 85.98 -> 87.92 | Improved |
| add model warm-up of 100k iterations before applying SAA (ablation) | SAA on FixMatch · CIFAR-10 (40 labels) | accuracy (%) | Improved |
| augment whole image then cut, instead of cutting then augmenting patches (ablation) | SAA · CIFAR-10 (40 labels) and STL-10 | accuracy (%) | Improved |
| add SAA to FixMatch: speed of improvement | FixMatch · CIFAR-10 (40 labels), 200k iterations | accuracy (%): 86.72 -> 91.15 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 4 on most shared results.
- FixMatch — FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence (NeurIPS 2020) · better on 6 of 6 shared results
- FlexMatch — FlexMatch: Boosting Semi-Supervised Learning with Curriculum Pseudo Labeling (NeurIPS 2021) · better on 4 of 4 shared results
- NP-Match — NP-Match: When Neural Processes meet Semi-Supervised Learning (ICML 2022) · better on 1 of 1 shared result
- ReMixMatch — ReMixMatch: Semi-Supervised Learning with Distribution Matching and Augmentation Anchoring (ICLR 2020) · better on 0 of 1 shared result
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
- Enhancing Semi-Supervised Learning via Representative and Diverse Sample Selection (NeurIPS 2024 · similarity 0.88)
- Data Augmentation with Diffusion for Open-Set Semi-Supervised Learning (NeurIPS 2024 · similarity 0.85)
- CGMatch: A Different Perspective of Semi-supervised Learning (CVPR 2025 · similarity 0.85)
- Towards Cost-Effective Learning: A Synergy of Semi-Supervised and Active Learning (CVPR 2025 · similarity 0.85)
- RegMixMatch: Optimizing Mixup Utilization in Semi-Supervised Learning (AAAI 2025 · similarity 0.85)
- Enhancing Sample Selection Against Label Noise by Cutting Mislabeled Easy Examples (NeurIPS 2025 · similarity 0.85)
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.
