Research map / Label-efficient and robust learning / Semi-supervised learning
Class-Imbalanced Semi-Supervised Learning with Adaptive Thresholding
Read the paperCheck 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 |
|---|---|---|---|
| adaptive class-dependent thresholds (Adsh) replacing FixMatch fixed 0.95 threshold | FixMatch on imbalanced CIFAR-10 (γl=γu, N1=1500) · CIFAR-10 N1=1500 γ=50/100/150 | accuracy: +4.28, +5.02, +3.02 points | Helped |
| adaptive class-dependent thresholds (Adsh) replacing FixMatch fixed threshold | FixMatch on imbalanced CIFAR-10 (N1=500) · CIFAR-10 N1=500 γ=50/100/150 | accuracy: +1.93, +2.52, +1.94 points | Helped |
| adaptive class-dependent thresholds (Adsh) | SVHN with γ=100 · SVHN γ=100 | accuracy: +2.79 over FixMatch | Helped |
| adaptive class-dependent thresholds (Adsh) | STL-10 with unknown γu · STL-10 γl=10 and γl=20 | accuracy: +6.00 (γl=10), +7.49 (γl=20) over FixMatch | Helped |
| adaptive thresholding (Adsh) vs class-imbalanced SSL baselines DARP and CReST | CIFAR-10 γl=γu · CIFAR-10 N1=1500 γ=50/100/150 and N1=500 | accuracy: Adsh best in all settings; e.g., 83.38 vs 82.03 (CReST) at γ=50 | Helped |
| adaptive thresholding with mismatched labeled/unlabeled imbalance | CIFAR-10 γl≠γu (24 settings) · CIFAR-10 γl≠γu | accuracy: not reported numerically | Helped |
| adaptive thresholding sensitivity to τ1 (0.95 to 0.99) (ablation) | Adsh on CIFAR-10 N1=1500 γ=100 · CIFAR-10 hyperparameter τ1 | accuracy: best at 0.96; other values did not hurt much | No clear effect |
Closest papers by meaning
- CReST: A Class-Rebalancing Self-Training Framework for Imbalanced Semi-Supervised Learning (CVPR 2021 · similarity 0.92)
- ABC: Auxiliary Balanced Classifier for Class-imbalanced Semi-supervised Learning (NeurIPS 2021 · similarity 0.92)
- Imbalanced Semi-supervised Learning with Bias Adaptive Classifier (ICLR 2023 · similarity 0.91)
- CDMAD: Class-Distribution-Mismatch-Aware Debiasing for Class-Imbalanced Semi-Supervised Learning (CVPR 2024 · similarity 0.90)
- Twice Class Bias Correction for Imbalanced Semi-supervised Learning (AAAI 2024 · similarity 0.89)
- OwMatch: Conditional Self-Labeling with Consistency for Open-World Semi-Supervised Learning (NeurIPS 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.