Research map / Label-efficient and robust learning / Semi-supervised learning
CHMATCH: Contrastive Hierarchical Matching and Robust Adaptive Threshold Boosted Semi-Supervised Learning
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 |
|---|---|---|---|
| Use memory-bank dynamic-proportion adaptive thresholds instead of fixed thresholds (ablation) | CHMatch · CIFAR-100, 400 labels | error rate: 49.56 -> 45.23 | Helped |
| Add a coarse-grained classification head (ablation) | CHMatch · CIFAR-100, 400 labels | error rate: 47.14 -> 45.23 | Helped |
| Add graph matching between fine- and coarse-grained pseudo-label graphs (ablation) | CHMatch · CIFAR-100, 400 labels | error rate: 48.83 -> 45.23 | Helped |
| Add hierarchical-label guided contrastive matching (ablation) | CHMatch · CIFAR-100, 400 labels | error rate: 50.14 -> 45.23 | Helped |
| Use linear rather than exponential or log-type increase for K (ablation) | CHMatch · CIFAR-100, 40 labels | error rate: 46.63; 47.51; linear increasing manner 45.23 | Helped |
| Add graph matching to the coarse-label predictions (ablation) | CHMatch · CIFAR-100, 400 labels | inconsistent rate of coarse labels: 30.94% -> 27.91% | Helped |
| Add CHMatch's hierarchical matching and adaptive-threshold approach | CIFAR-10 · CIFAR-10, 1000 labels | error rate: 4.19± 0.01 -> 4.48± 0.10 | Hurt |
| Use CHMatch rather than FlexMatch | CIFAR-100 · CIFAR-100, 400 labels, WRN-28-2 | error rate: 49.23± 2.58 -> 45.23± 0.28 | Helped |
| Use CHMatch rather than CoMatch | STL-10 · STL-10, 1000 labels | error rate: 20.20± 0.38 -> 10.36± 0.31 | Helped |
| Use CHMatch rather than CoMatch | ImageNet subset · ImageNet subset, 1% labels | Top-1 error rate: 42.88 -> 34.18 | Helped |
| Increase the dynamic-threshold duration to 150 epochs (ablation) | CHMatch · CIFAR-100 | error rate | Helped |
Closest papers by meaning
- HyperMatch: Noise-Tolerant Semi-Supervised Learning via Relaxed Contrastive Constraint (CVPR 2023 · similarity 0.90)
- RegMixMatch: Optimizing Mixup Utilization in Semi-Supervised Learning (AAAI 2025 · similarity 0.87)
- OwMatch: Conditional Self-Labeling with Consistency for Open-World Semi-Supervised Learning (NeurIPS 2024 · similarity 0.86)
- Scaling Up Semi-supervised Learning with Unconstrained Unlabelled Data (AAAI 2024 · similarity 0.85)
- Continuous Contrastive Learning for Long-Tailed Semi-Supervised Recognition (NeurIPS 2024 · similarity 0.85)
- CGMatch: A Different Perspective of Semi-supervised Learning (CVPR 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.