Research map / Graph learning / GNN architectures
Learning Graphs for Knowledge Transfer With Limited Labels
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 |
|---|---|---|---|
| Learn and update the input graph with triplet loss | GCN semi-supervised learning · Citeseer | accuracy: 72.0% -> 74.3% | Helped |
| Learn and update the input graph with triplet loss | GCN semi-supervised learning · Pubmed | accuracy: 77.8% -> 79.8% | Helped |
| Learn and update the input graph with triplet loss | GCN semi-supervised learning · Cora | accuracy: 80.0% -> 83.6% | Helped |
| Learn and update the input graph with triplet loss | GCN zero-shot/few-shot action recognition · UCF101 and HMDB51, all three input-graph configurations | mean accuracy: 49.14 -> 53.27; 57.04 -> 60.57; 64.24 -> 65.49; 38.01 -> 41.05; 45.07 -> 48.07; 47.69 -> 49.17 | Helped |
| Add triplet loss without updating adjacency matrix (ablation) | GCN on UCF101 with V-KG · UCF101, V-KG | mean accuracy: 57.04 -> 58.57 | Helped |
| Update adjacency matrix without triplet loss (ablation) | GCN on UCF101 with V-KG · UCF101, V-KG | mean accuracy: 57.04 -> 59.39 | Helped |
| Use both triplet loss and adjacency updates (ablation) | GCN on UCF101 with V-KG · UCF101, V-KG | mean accuracy: 57.04 -> 60.57 | Helped |
| Set input-graph averaging weight λ to 0.8 rather than 1.0 (ablation) | GCN on Pubmed · Pubmed validation | accuracy: 76.2% -> 80.6% | Helped |
| Vary the number of epochs between adjacency updates (ablation) | GCN on UCF101 with A-KG · UCF101, A-KG | mean accuracy: 52.89; 50.17; 54.41; 50.72; 48.71 | Mixed |
| Vary the negative-set class index range for triplet loss (ablation) | GCN on UCF101 with A-KG · UCF101, A-KG | mean accuracy: 49.22; 48.74; 51.51; 54.41; 49.27 | Mixed |
| Learn and update the input graph with triplet loss | GCN zero-shot action recognition · UCF101, 23-78 split, A-KG | mean accuracy | Mixed |
Closest papers by meaning
- Graph Few-Shot Learning via Knowledge Transfer (AAAI 2020 · similarity 0.90)
- Distilling Knowledge From Graph Convolutional Networks (CVPR 2020 · similarity 0.90)
- Semi-Supervised Learning With Graph Learning-Convolutional Networks (CVPR 2019 · similarity 0.90)
- Deeper Insights Into Graph Convolutional Networks for Semi-Supervised Learning (AAAI 2018 · similarity 0.90)
- Semi-Supervised Classification with Graph Convolutional Networks (ICLR 2017 · similarity 0.89)
- Unlocking the Potential of Black-box Pre-trained GNNs for Graph Few-shot Learning (AAAI 2025 · similarity 0.87)
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.