Research map / Graph learning / GNN architectures
Rethinking Graph Neural Architecture Search From Message-Passing
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 |
|---|---|---|---|
| Use GNAS instead of the strongest reported manual or search-based baseline | Graph neural architecture search · PATTERN | test accuracy: 85.59 ± 0.01 -> 86.80 ± 0.10 | Helped |
| Use GNAS instead of the strongest reported manual or search-based baseline | Graph neural architecture search · CLUSTER | test accuracy: 60.40 ± 0.42 -> 62.21 ± 0.20 | Helped |
| Use GNAS instead of the strongest reported manual or search-based baseline | Graph neural architecture search · MNIST | test accuracy: 97.34 ± 0.14 -> 98.01 ± 0.10 | Helped |
| Use GNAS instead of the strongest reported manual or search-based baseline | Graph neural architecture search · CIFAR10 | test accuracy: 67.31 ± 0.31 -> 70.10 ± 0.44 | Helped |
| Use GNAS instead of the strongest reported manual or search-based baseline | Graph neural architecture search · ZINC | MAE: 0.292 -> 0.242 | Helped |
| Optimize architecture parameters using training loss rather than validation loss (ablation) | GNAS · Five datasets | identity-operation collapse | Helped |
| Increase initial search depth (ablation) | GNAS · ZINC | searched message-passing depth: 12-14 | Mixed |
| Use depth 12 rather than depth 4 (ablation) | GNAS · ZINC | MAE: 0.276 -> 0.242 | Helped |
| Use depth 16 rather than depth 12 (ablation) | GNAS · ZINC | MAE: 0.242 -> 0.260 | Hurt |
| Use mean aggregation rather than the selected aggregators (ablation) | GNAS · Five datasets | test performance | Hurt |
Closest papers by meaning
- One-shot Graph Neural Architecture Search with Dynamic Search Space (AAAI 2021 · similarity 0.92)
- Deep and Flexible Graph Neural Architecture Search (ICML 2022 · similarity 0.92)
- Graph Differentiable Architecture Search with Structure Learning (NeurIPS 2021 · similarity 0.92)
- Neural Architecture Retrieval (ICLR 2024 · similarity 0.89)
- AutoGFM: Automated Graph Foundation Model with Adaptive Architecture Customization (ICML 2025 · similarity 0.88)
- Can Classic GNNs Be Strong Baselines for Graph-level Tasks? Simple Architectures Meet Excellence (ICML 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.