Research map / Graph learning / Graph self-supervised learning
Score-based Generative Modeling of Graphs via the System of Stochastic Differential Equations
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 |
|---|---|---|---|
| Model node-edge dependency jointly over time instead of generating nodes and edges sequentially (ablation) | GDSS · Ego-small | Avg. MMD: 0.023 -> 0.017 | Helped |
| Model node-edge dependency jointly over time instead of generating nodes and edges sequentially (ablation) | GDSS · Community-small | Avg. MMD: 0.073 -> 0.046 | Helped |
| Model node-edge dependency jointly over time instead of generating nodes and edges sequentially (ablation) | GDSS · Enzymes | Avg. MMD: 0.111 -> 0.032 | Helped |
| Model node-edge dependency jointly over time instead of generating nodes and edges sequentially (ablation) | GDSS · Grid | Avg. MMD: 0.135 -> 0.062 | Helped |
| Model node-edge dependency jointly over time instead of generating nodes and edges sequentially (ablation) | GDSS · QM9 | Validity w/o corr. (%): 94.47 -> 95.72 | Helped |
| Model node-edge dependency jointly over time instead of generating nodes and edges sequentially (ablation) | GDSS · ZINC250k | Validity w/o corr. (%): 92.39 -> 97.01 | Helped |
| Use discrete-step perturbation instead of continuous-time diffusion (ablation) | GDSS-discrete · Molecule generation | Generation quality | Hurt |
| Use GMH layers instead of simple GNN layers in EDP-GNN (ablation) | EDP-GNN · Graph generation | Generation quality | No clear effect |
| Use GMH instead of GNN for the node score model (ablation) | GDSS · Graph generation | Generation quality | No clear effect |
| Use S4 instead of predictor-only EM or Reverse samplers (ablation) | GDSS · Graph generation | Generation quality | Helped |
| Use S4 instead of PC samplers (ablation) | GDSS · Graph generation | Generation quality | Helped |
| Use EMA for larger graph datasets (ablation) | GDSS · Enzymes and Grid | Generation performance | Helped |
Closest papers by meaning
- Graph Generation with Diffusion Mixture (ICML 2024 · similarity 0.92)
- A Non-Asymptotic Convergent Analysis for Scored-Based Graph Generative Model via a System of Stochastic Differential Equations (ICML 2025 · similarity 0.92)
- Discrete-state Continuous-time Diffusion for Graph Generation (NeurIPS 2024 · similarity 0.91)
- Pard: Permutation-Invariant Autoregressive Diffusion for Graph Generation (NeurIPS 2024 · similarity 0.89)
- Generative Modelling of Structurally Constrained Graphs (NeurIPS 2024 · similarity 0.89)
- SBGD: Improving Graph Diffusion Generative Model via Stochastic Block Diffusion (ICML 2025 · similarity 0.89)
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.