Research map / Graph learning / GNN architectures
Bandit Samplers for Training Graph Neural Networks
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 a bandit sampler instead of the strongest reported GCN sampling baseline | GCN · Cora | Micro F1: 0.828 (± 0.001) -> 0.855 (± 0.005) | Helped |
| Use a bandit sampler instead of the strongest reported GCN sampling baseline | GCN · Pubmed | Micro F1: 0.899 (± 0.002) -> 0.903 (± 0.001) | Helped |
| Use a bandit sampler instead of the strongest reported GCN sampling baseline | GCN · PPI | Micro F1: 0.853(± 0.001) -> 0.905(± 0.003) | Helped |
| Use a bandit sampler instead of the strongest reported GCN sampling baseline | GCN · Reddit | Micro F1: 0.965 (± 0.001) -> 0.957 (± 0.000) | Hurt |
| Use a bandit sampler instead of the strongest reported GCN sampling baseline | GCN · Flickr | Micro F1: 0.506 (± 0.012) -> 0.513 (± 0.001) | Helped |
| Use a bandit sampler instead of GraphSAINT-GAT | GAT · Cora | Micro F1: 0.773 (± 0.036) -> 0.857 (± 0.003) | Helped |
| Use a bandit sampler instead of GraphSAINT-GAT | GAT · Pubmed | Micro F1: 0.886 (± 0.016) -> 0.894 (± 0.001) | Helped |
| Use a bandit sampler instead of GraphSAINT-GAT | GAT · PPI | Micro F1: 0.789 (± 0.001) -> 0.867 (± 0.003) | Helped |
| Use a bandit sampler instead of GraphSAINT-GAT | GAT · Reddit | Micro F1: 0.933 (± 0.012) -> 0.962 (± 0.001) | Helped |
| Use a bandit sampler instead of GraphSAINT-GAT | GAT · Flickr | Micro F1: 0.470 (± 0.002) -> 0.513 (± 0.001) | Helped |
| Use multiple-play sampling (GNN-BS.M) rather than repeated single-arm sampling (GNN-BS) (ablation) | GAT · PPI | Micro F1: 0.841(± 0.001) -> 0.867(± 0.003) | Helped |
| Use multiple-play sampling (GNN-BS.M) rather than repeated single-arm sampling (GNN-BS) (ablation) | GAT · Cora | Micro F1: 0.857 (± 0.003) -> 0.857 (± 0.003) | No clear effect |
Closest papers by meaning
- A Biased Graph Neural Network Sampler with Near-Optimal Regret (NeurIPS 2021 · similarity 0.93)
- Stochastic Training of Graph Convolutional Networks with Variance Reduction (ICML 2018 · similarity 0.90)
- Adaptive Sampling Towards Fast Graph Representation Learning (NeurIPS 2018 · similarity 0.90)
- MuseGNN: Forming Scalable, Convergent GNN Layers that Minimize a Sampling-Based Energy (ICLR 2025 · similarity 0.87)
- On the Stability of Graph Convolutional Neural Networks: A Probabilistic Perspective (NeurIPS 2025 · similarity 0.86)
- A Manifold Perspective on the Statistical Generalization of Graph Neural Networks (ICML 2025 · similarity 0.86)
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.