Research map / Label-efficient and robust learning / Semi-supervised learning
Geometric Regularization for Long-Tailed Semi-Supervised Learning via Gaussian Feature Bridges
Hongyang He, Xinyuan Song, Yan Zhong, Daizong Liu, Yanbin Li, Yangfan He, Wenqiao Zhang
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 bridge consistency loss (ablation) | GBC · CIFAR10-LT (consistent, gamma_l=100) | Top-1 accuracy: 86.1 -> 92.3 | Improved |
| add Gaussian noise to the bridge (ablation) | GBC · CIFAR10-LT (consistent, gamma_l=100) | Top-1 accuracy: 90.2 -> 92.3 | Improved |
| use residual fusion instead of hard replacement (ablation) | GBC · CIFAR10-LT (consistent, gamma_l=100) | Top-1 accuracy: 88.4 -> 92.3 | Improved |
| add class-aware reweighting (ablation) | GBC · CIFAR10-LT (consistent, gamma_l=100) | Top-1 accuracy: 85.3 -> 92.3 | Improved |
| add BridgeMix (ablation) | GBC · CIFAR10-LT (consistent, gamma_l=100) | Top-1 accuracy: 87.4 -> 92.3 | Improved |
| add BridgeMix (ablation) | GBC · CIFAR10-LT (reversed, gamma_l=100) | Top-1 accuracy: 81.7 -> 86.8 | Improved |
| inject bridge at layer4 instead of layer2 (ablation) | GBC · CIFAR10-LT (consistent, gamma_l=100) | Top-1 accuracy: 87.8 -> 92.3 | Improved |
| inject bridge at layer4 instead of fc (ablation) | GBC · CIFAR10-LT (consistent, gamma_l=100) | Top-1 accuracy: 91.2 -> 92.3 | Improved |
| prototype atlas size 2% instead of default (ablation) | GBC · CIFAR10-LT (consistent, gamma_l=100) | Top-1 accuracy: 92.3 -> 85.7 | Hurt |
| prototype atlas size 8% instead of default (ablation) | GBC · CIFAR10-LT (consistent, gamma_l=100) | Top-1 accuracy: 92.3 -> 91.1 | Hurt |
| noise level nu 0.10 versus 0.00 (ablation) | GBC · CIFAR10-LT (consistent, gamma_l=100) | Top-1 accuracy: 91.8 -> 92.3 | Improved |
| beta(2,2) sampling instead of uniform (alpha 1.0) (ablation) | GBC · CIFAR10-LT (consistent, gamma_l=100) | Top-1 accuracy: 91.5 -> 92.3 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FixMatch — FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence (NeurIPS 2020) · better on 1 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
- ConMatch: Semi-Supervised Learning with Confidence-Guided Consistency Regularization (ECCV 2022 · similarity 0.88)
- Adaptive Consistency Regularization for Semi-Supervised Transfer Learning (CVPR 2021 · similarity 0.87)
- Continuous Contrastive Learning for Long-Tailed Semi-Supervised Recognition (NeurIPS 2024 · similarity 0.87)
- MaxMatch: Semi-Supervised Learning With Worst-Case Consistency (TPAMI 2022 · similarity 0.87)
- Keep It on a Leash: Controllable Pseudo-label Generation Towards Realistic Long-Tailed Semi-Supervised Learning (NeurIPS 2025 · similarity 0.86)
- RegMixMatch: Optimizing Mixup Utilization in Semi-Supervised Learning (AAAI 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.
