Research map / Graph learning / GNN architectures
GraphSAINT: Graph Sampling Based Inductive Learning Method
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 graph-level Node sampling instead of the layer-sampling baseline | GCN minibatch training · PPI, Flickr, Reddit, Yelp, Amazon | F1-micro: 0.637 -> 0.960 (PPI); 0.501 -> 0.507 (Flickr); 0.953 -> 0.962 (Reddit); 0.634 -> 0.641 (Yelp); 0.758 -> 0.782 (Amazon) | Helped |
| Use random edge sampling rather than random node sampling | GraphSAINT · PPI, Flickr, Reddit, Yelp, Amazon | F1-micro: 0.960 -> 0.981 (PPI); 0.507 -> 0.510 (Flickr); 0.962 -> 0.966 (Reddit); 0.641 -> 0.653 (Yelp); 0.782 -> 0.807 (Amazon) | Helped |
| Use random-walk sampling rather than random node sampling | GraphSAINT · PPI, Flickr, Reddit, Yelp, Amazon | F1-micro: 0.960 -> 0.981 (PPI); 0.507 -> 0.511 (Flickr); 0.962 -> 0.966 (Reddit); 0.641 -> 0.653 (Yelp); 0.782 -> 0.815 (Amazon) | Helped |
| Increase random-walk root count from 250 to 750 (ablation) | GraphSAINT-RW · PPI | F1-micro | Helped |
| Increase random-walk root count beyond 750 (ablation) | GraphSAINT-RW · All evaluated datasets | F1-micro | No clear effect |
| Increase JK-GraphSAINT depth from 2 layers to 4 layers (ablation) | JK-GraphSAINT with edge sampler · Reddit | F1-micro: 0.966 -> 0.970 | Helped |
| Increase depth of GraphSAGE-minibatch GAT and JK-net from 2 to 4 layers (ablation) | GAT-SAGE and JK-SAGE · Reddit | F1-micro | Hurt |
| Use graph sampling instead of GraphSAGE layer sampling for 4-layer architectures | GAT and JK-net · Reddit | training speed: two orders of magnitude | Helped |
| Use GraphSAINT rather than ClusterGCN for the 5-layer PPI model | GCN, 5×2048 · PPI (large version) | F1-micro: 0.994 -> 0.995 | Helped |
| Use GraphSAINT rather than ClusterGCN for the 4-layer Reddit model | GCN, 4×128 · Reddit | F1-micro: 0.966 -> 0.970 | Helped |
Closest papers by meaning
- L2-GCN: Layer-Wise and Learned Efficient Training of Graph Convolutional Networks (CVPR 2020 · similarity 0.92)
- Deeper Insights Into Graph Convolutional Networks for Semi-Supervised Learning (AAAI 2018 · similarity 0.91)
- Adaptive Sampling Towards Fast Graph Representation Learning (NeurIPS 2018 · similarity 0.91)
- Simplifying Graph Convolutional Networks (ICML 2019 · similarity 0.91)
- On the Stability of Graph Convolutional Neural Networks: A Probabilistic Perspective (NeurIPS 2025 · similarity 0.89)
- Generalizing CNNs to graphs with learnable neighborhood quantization (NeurIPS 2024 · similarity 0.89)
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.