Research map / Label-efficient and robust learning / Semi-supervised learning
Density-Aware Graph for Deep Semi-Supervised Visual Recognition
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 |
|---|---|---|---|
| add density-aware neighborhood aggregation (ablation) | DAG feature learning · CIFAR10, 1000 labeled | error rate: 10.96 -> 9.18 | Helped |
| add density-aware neighborhood aggregation (ablation) | DAG feature learning · CIFAR100, 4000 labeled | error rate: 43.34 -> 40.33 | Helped |
| use density information in neighborhood aggregation instead of similarity alone (ablation) | neighborhood aggregation · CIFAR10 and CIFAR100 | error rate: 9.46 -> 9.18; 40.76 -> 40.33 | Helped |
| use two-hop instead of one-hop neighbors (ablation) | DNA · CIFAR10, 1000 labeled | error rate: 9.18 -> 9.20 | No clear effect |
| use two-hop instead of one-hop neighbors (ablation) | DNA · CIFAR100, 4000 labeled | error rate: 40.33 -> 39.60 | Helped |
| use three-hop neighbors (ablation) | DNA · CIFAR10 and CIFAR100 | error rate: 9.18 -> 9.26; 39.60 -> 40.28 | Hurt |
| construct density-ascending paths from labeled samples only (ablation) | DPLP · CIFAR10 and CIFAR100, 4000 labeled | error rate: 10.96 -> 9.57; 43.34 -> 40.94 | Helped |
| also construct paths from remaining unlabeled samples (ablation) | DPLP · CIFAR10 and CIFAR100, 4000 labeled | error rate: 9.57 -> 9.14; 40.94 -> 40.08 | Helped |
| combine DNA and DPLP (ablation) | DAG · CIFAR10 and CIFAR100, 4000 labeled | error rate: 9.14 -> 8.35; 40.08 -> 38.04 | Helped |
| use DAG instead of the strongest listed baseline | semi-supervised visual recognition · CIFAR10, 2000 labeled | error rate: 7.03 -> 7.16 | Hurt |
| use DAG instead of the strongest listed baseline | semi-supervised visual recognition · CIFAR10, 4000 labeled | error rate: 7.29 -> 6.13 | Helped |
| use DAG instead of the strongest listed baseline | semi-supervised visual recognition · CIFAR100, 4000 labeled | error rate: 43.73 -> 37.38 | Helped |
Closest papers by meaning
- Probability-Density-aware Semi-supervised Learning (AAAI 2025 · similarity 0.89)
- HGDL: Heterogeneous Graph Label Distribution Learning (NeurIPS 2024 · similarity 0.88)
- GLDL: Graph Label Distribution Learning (AAAI 2024 · similarity 0.86)
- Neighbor Does Matter: Density-Aware Contrastive Learning for Medical Semi-supervised Segmentation (AAAI 2025 · similarity 0.86)
- Roll with the Punches: Expansion and Shrinkage of Soft Label Selection for Semi-supervised Fine-Grained Learning (AAAI 2024 · similarity 0.85)
- InterLUDE: Interactions between Labeled and Unlabeled Data to Enhance Semi-Supervised Learning (ICML 2024 · similarity 0.84)
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.