Research map / Graph learning / Graph algorithms and clustering
PolyGraph Discrepancy: a classifier-based metric for graph generation
Markus Krimmel, Philip Hartout, Karsten Borgwardt, Dexiong Chen
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 PGD instead of MMD metrics for tracking validity across denoising steps | DiGress on Planar-L · Planar-L denoising steps | neg. Pearson corr with validity: 99.29 vs 82.52 (best MMD, GIN RBF) | Improved |
| use PGD instead of MMD for tracking training progress | DiGress on SBM-L · SBM-L training | Spearman corr with training iterations: 77.27 vs 68.18 (Clust. RBF), 23.64 (Orbit RBF) | Improved |
| use PGD instead of MMD on Lobster-L training | DiGress on Lobster-L · Lobster-L training | neg. Pearson corr with validity: 86.54 vs 87.05 (Clust. RBF) | Mixed |
| use PGD instead of MMD on perturbation experiments | Graph metrics · Synthetic perturbations | Spearman corr with perturbation magnitude | No clear effect |
| use multiple descriptors in PGD to detect edge swapping | PGD · Edge-swapping perturbation | Spearman corr with perturbation magnitude | Improved |
| use larger datasets (SBM-L, Planar-L, Lobster-L) with unbiased MMD | MMD evaluation · Synthetic graph benchmarks | MMD bias and variance: 20-40 test graphs: bias dominates | Improved |
| use feature concatenation instead of max-reduction over descriptors | PGD · Benchmark | JSD estimate tightness | Mixed |
| pretrain AutoGraph on PubChem | AutoGraph · GuacaMol | PGD (x100): 27.3 -> 11.5 | Improved |
| use PGD-TV instead of PGD-JS | PGD · Training progress | Spearman corr with training | Hurt |
Closest papers by meaning
- Quality Measures for Dynamic Graph Generative Models (ICLR 2025 · similarity 0.88)
- A Non-Asymptotic Convergent Analysis for Scored-Based Graph Generative Model via a System of Stochastic Differential Equations (ICML 2025 · similarity 0.86)
- On the Role of Edge Dependency in Graph Generative Models (ICML 2024 · similarity 0.86)
- No Metric to Rule Them All: Toward Principled Evaluations of Graph-Learning Datasets (ICML 2025 · similarity 0.85)
- Towards Addressing Frontiers in Graph Generation (AAAI 2025 · similarity 0.85)
- Generative Graph Pattern Machine (NeurIPS 2025 · similarity 0.84)
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.
