Research map / Graph learning / Graph self-supervised learning
Feature-Aware (Hyper)graph Generation via Next-Scale Prediction
Dorian Gailhard, Enzo Tartaglione, Lirida Naviner, Jhony H. Giraldo
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 hierarchical scale encoding instead of concatenating target size to node embeddings (ablation) | FAHNES · ModelNet Bookshelf | Node Num: 0.940±0.917 -> 0.135±0.276 | Improved |
| use hierarchical scale encoding (ablation) | FAHNES · ModelNet Piano | Node Num: 3.622±1.822 -> 0.846±1.009 | Improved |
| use hierarchical scale encoding (ablation) | FAHNES · ManifoldNet Airplane | ChamDist: 0.079±0.019 -> 0.048±0.003 | Improved |
| add multi-scale graph OT coupling (ablation) | FAHNES · ManifoldNet Airplane | ChamDist: 0.050±0.005 -> 0.048±0.003 | No clear effect |
| add multi-scale graph OT coupling (ablation) | FAHNES · ManifoldNet Bench | ChamDist: 0.085±0.056 -> 0.064±0.005 | Improved |
| add multi-scale graph OT coupling (ablation) | FAHNES · ModelNet Bookshelf | Spectral: 0.014±0.007 -> 0.024±0.015 | Hurt |
| remove both scale encoding and OT coupling (ablation) | FAHNES · ModelNet Piano | Node Num: 5.490±8.847 -> 0.846±1.009 | Improved |
| joint topology and feature generation instead of sequential topology-then-features | FAHNES · ManifoldNet Airplane | ChamDist: 0.082±0.074 -> 0.048±0.003 | Improved |
| hierarchical generation instead of flat models | graph point cloud datasets · Airplane Point Clouds | ChamDist: OOM for flat baselines -> 0.094±0.006 | Improved |
| use FAHNES instead of flat DeFoG on small SBM graphs | DeFoG · SBM graphs | Valid (%): 90.0±5.1 -> 50.0±5.0 | Hurt |
| use FAHNES instead of HSpectre on unfeatured graphs | HSpectre · Point cloud graphs | Ratio: 7.0 -> 3.2±0.5 | Improved |
| use FAHNES instead of HYGENE on featureless hypergraphs | HYGENE · SBM hypergraphs | Valid (%): 65.0 -> 87.8±3.1 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- DeFoG — DeFoG: Discrete Flow Matching for Graph Generation (ICML 2025) · better on 0 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Towards Addressing Frontiers in Graph Generation (AAAI 2025 · similarity 0.87)
- HypeBoy: Generative Self-Supervised Representation Learning on Hypergraphs (ICLR 2024 · similarity 0.87)
- HiGen: Hierarchical Graph Generative Networks (ICLR 2024 · similarity 0.87)
- Efficient and Scalable Graph Generation through Iterative Local Expansion (ICLR 2024 · similarity 0.87)
- On the Role of Edge Dependency in Graph Generative Models (ICML 2024 · similarity 0.87)
- Quality Measures for Dynamic Graph Generative Models (ICLR 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.
