Research map / Label-efficient and robust learning / Semi-supervised learning
DC-SSL: Addressing Mismatched Class Distribution in 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 |
|---|---|---|---|
| Estimate the unlabeled class distribution from EMA predictions and apply distribution consistency | FixMatch on CIFAR10 with 40 labels and γu=50 · CIFAR10, |Dx|=40, γu=50 | test accuracy (%): 57.54 -> 63.95 | Helped |
| Estimate the unlabeled class distribution from EMA predictions and apply distribution consistency | FixMatch on CIFAR10 with 40 labels and γu=100 · CIFAR10, |Dx|=40, γu=100 | test accuracy (%): 54.82 -> 57.16 | Helped |
| Estimate the unlabeled class distribution from EMA predictions and apply distribution consistency | FixMatch on CIFAR10 with 40 labels and γu=200 · CIFAR10, |Dx|=40, γu=200 | test accuracy (%): 50.67 -> 53.27 | Helped |
| Use EMA-estimated reference distribution rather than labeled-distribution alignment | FixMatch on CIFAR10 with 40 labels and γu=50 · CIFAR10, |Dx|=40, γu=50 | test accuracy (%): 54.08 -> 62.44 | Helped |
| Estimate the unlabeled class distribution from EMA predictions and apply distribution consistency | FixMatch on CIFAR100 with 2500 labels and γu=100 · CIFAR100, |Dx|=2500, γu=100 | test accuracy (%): 52.46 -> 59.02 | Helped |
| Estimate the unlabeled class distribution from EMA predictions and apply distribution consistency | FixMatch on Mini-ImageNet with 1000 labels and γu=100 · Mini-ImageNet, |Dx|=1000, γu=100 | test accuracy (%): 21.65 -> 29.12 | Helped |
| Use EMA-estimated reference distribution for training-free pseudo-label correction | FixMatch on imbalanced-labeled CIFAR10 with Γ0=100 · CIFAR10, |Dx|=250, Γ0=100 | test accuracy (%): 69.76 -> 72.21 | Helped |
| Estimate the unlabeled class distribution from EMA predictions and apply distribution consistency | FixMatch on imbalanced-labeled Mini-ImageNet with Γ0=40 · Mini-ImageNet, |Dx|=1000, Γ0=40 | test accuracy (%): 36.20 -> 40.13 | Helped |
| Add distribution consistency to FixMatch | STL-10 with 1000 labeled samples · STL-10, |Dx|=1000 | test accuracy (%): 65.38 -> 82.47 | Helped |
| Use DC-SSL (TF) rather than DC-SSL (TB) | CIFAR10 with 40 labels and matched distributions · CIFAR10, |Dx|=40, matched distributions | test accuracy (%): 95.31 vs 93.89 | Mixed |
| Vary EMA momentum coefficient α (ablation) | DC-SSL (TF) on matched CIFAR10 with 40 labels · CIFAR10, |Dx|=40, matched distributions | test accuracy (%): 93.14; 94.82; 95.38; 95.07 | Mixed |
| Increase distribution-loss weight λd from 1.0 to 7.0 (ablation) | DC-SSL (TB) on matched CIFAR10 with 40 labels · CIFAR10, matched distributions | test accuracy (%): 93.92 -> 94.81 | Helped |
Closest papers by meaning
- Towards Realistic Long-Tailed Semi-Supervised Learning: Consistency Is All You Need (CVPR 2023 · similarity 0.90)
- RankMatch: A Novel Approach to Semi-Supervised Label Distribution Learning Leveraging Rank Correlation between Labels (NeurIPS 2025 · similarity 0.90)
- On Non-Random Missing Labels in Semi-Supervised Learning (ICLR 2022 · similarity 0.90)
- Boosting Semi-Supervised Learning by Exploiting All Unlabeled Data (CVPR 2023 · similarity 0.90)
- CDMAD: Class-Distribution-Mismatch-Aware Debiasing for Class-Imbalanced Semi-Supervised Learning (CVPR 2024 · similarity 0.89)
- Contrastive Credibility Propagation for Reliable Semi-supervised Learning (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.