Research map / Label-efficient and robust learning / Semi-supervised learning
Dual-Decoupling Learning and Metric-Adaptive Thresholding for Semi-Supervised Multi-Label Learning
Jia-Hao Xiao, Ming-Kun Xie, Heng-Bo Fan, Gang Niu, Masashi Sugiyama, Sheng-Jun Huang
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 |
|---|---|---|---|
| replace CAP class-proportion thresholding with metric-adaptive thresholding (MAT) (ablation) | CAP baseline · COCO (p=0.05) | mAP: 62.43 -> 65.03 | Improved |
| add correlative/discriminative feature decoupling (dual-head classifier with patches) (ablation) | CAP + MAT · COCO (p=0.05) | mAP: 65.03 -> 66.07 | Improved |
| add generation/utilization of pseudo-labels decoupling (separate classifier heads) (ablation) | CAP + MAT + CDD · COCO (p=0.05) | mAP: 66.07 -> 69.30 | Improved |
| add GUD decoupling (ablation) | CAP + MAT + CDD · VOC (p=0.05) | mAP: 77.11 -> 79.26 | Improved |
| add MAT to CAP (ablation) | CAP · NUS (p=0.20) | mAP: 51.06 -> 51.20 | Improved |
| use F-beta as MAT metric instead of precision or recall | MAT · COCO | mAP | Improved |
| vary beta in F-beta metric | MAT · COCO | mAP | No clear effect |
| increase number of image patches from 2x2 to 3x3 or 4x4 | D2L · VOC, COCO, NUS (p=0.05) | mAP | Hurt |
| vary temperature alpha in spatial weighting | D2L · COCO, NUS | mAP | No clear effect |
| use too large or too small temperature alpha | D2L · VOC | mAP | Hurt |
| use D2L+MAT instead of CAP at very low label ratio | SSMLL · NUS (p=0.01) | mAP: 24.75 -> 38.09 | Improved |
| use D2L+MAT instead of CAP on F1 metrics | SSMLL · VOC (CF1/OF1) | CF1: 69.86 -> 68.50 (p=0.05) | Mixed |
Closest papers by meaning
- Correlation-Induced Label Prior for Semi-Supervised Multi-Label Learning (ICML 2024 · similarity 0.91)
- Bi-Level Optimization for Semi-Supervised Learning with Pseudo-Labeling (AAAI 2025 · similarity 0.89)
- Semi-supervised Multi-label Learning with Balanced Binary Angular Margin Loss (NeurIPS 2024 · similarity 0.89)
- Keep It on a Leash: Controllable Pseudo-label Generation Towards Realistic Long-Tailed Semi-Supervised Learning (NeurIPS 2025 · similarity 0.89)
- Decouple then Classify: A Dynamic Multi-view Labeling Strategy with Shared and Specific Information (ICML 2024 · similarity 0.89)
- Semi-Supervised Regression with Heteroscedastic Pseudo-Labels (NeurIPS 2025 · 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.
