Research map / Graph learning / GNN architectures
Graph Differentiable Architecture Search with Structure Learning
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 |
|---|---|---|---|
| add hidden-feature-smoothness graph structure learning (vs GASSO-V1 plain DARTS) (ablation) | GASSO · Citeseer | accuracy (%): 78.71 -> 79.61 | Helped |
| add hidden-feature-smoothness graph structure learning (vs GASSO-V1) (ablation) | GASSO · Cora | accuracy (%): 87.19 -> 87.63 | Helped |
| use hidden feature smoothness instead of original feature smoothness (ablation) | GASSO · Cora | accuracy (%): 87.17 -> 87.63 | Helped |
| use hidden feature smoothness instead of one-hot predicted label smoothness (ablation) | GASSO · Citeseer | accuracy (%): 78.80 -> 79.61 | Helped |
| update neural architecture weights instead of freezing them (ablation) | GASSO · Citeseer | accuracy (%): 77.08 -> 79.61 | Helped |
| search architecture instead of fixing GCN (ablation) | GASSO · Cora | accuracy (%): 86.78 -> 87.63 | Helped |
| joint training of structure and architecture instead of separate stages (V8) (ablation) | GASSO · Pubmed | accuracy (%): 90.40 -> 90.52 | Helped |
| joint training instead of architecture-then-structure (V9) (ablation) | GASSO · Citeseer | accuracy (%): 78.22 -> 79.61 | Helped |
| compute edge weights directly from node similarity (V4) instead of learned smoothness (ablation) | GASSO · Pubmed | accuracy (%): 89.12 -> 90.52 | Helped |
| replace DARTS with GASSO | DARTS · Cora/CiteSeer with 100% random edge perturbation | accuracy | Helped |
| replace DARTS with GASSO | DARTS · Physics, CoraFull, ogbn-arxiv | accuracy (%): 69.52 -> 70.52 (ogbn-arxiv) | Helped |
| DARTS denoising ability under edge noise | DARTS · PubMed vs Cora/CiteSeer perturbation | accuracy | Mixed |
Closest papers by meaning
- Deep and Flexible Graph Neural Architecture Search (ICML 2022 · similarity 0.92)
- Rethinking Graph Neural Architecture Search From Message-Passing (CVPR 2021 · similarity 0.92)
- Graph HyperNetworks for Neural Architecture Search (ICLR 2019 · similarity 0.90)
- AutoGFM: Automated Graph Foundation Model with Adaptive Architecture Customization (ICML 2025 · similarity 0.88)
- On the Interplay between Graph Structure and Learning Algorithms in Graph Neural Networks (ICML 2025 · similarity 0.88)
- Data-Augmented Curriculum Graph Neural Architecture Search under Distribution Shifts (AAAI 2024 · 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.