Research map / Graph learning / Graph self-supervised learning
Discrete Bayesian Sample Inference for Graph Generation
Ole Petersen, Marcel Kollovieh, Marten Lienen, Stephan Günnemann
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 |
|---|---|---|---|
| increase noise level gamma from 0 (probability flow ODE) to tuned values (ablation) | GraphBSI · Moses | SNN, Filters, FCD | Improved |
| use Ornstein-Uhlenbeck instead of Euler-Maruyama discretization at high noise (ablation) | GraphBSI · Moses | sample quality | Improved |
| finer discretization at later timesteps (rho<1) (ablation) | GraphBSI · Moses | FCD | Improved |
| non-uniform timestepping rho (ablation) | GraphBSI · Moses | SNN, Filters | No clear effect |
| exponential instead of linear precision schedule (ablation) | GraphBSI · Moses | key metrics | No clear effect |
| set final precision just high enough for perfect reconstruction (ablation) | GraphBSI · Moses | key metrics | Improved |
| increase initial variance beta_0 from 0.05 to 1.0 (ablation) | OU sampler vs Flowback sampler · Moses | key metrics | Mixed |
| increase compute budget (NFEs 25 to 500) (ablation) | GraphBSI · Moses | novelty | Hurt |
| gamma to infinity sampler (Flowback-like) with higher budget (ablation) | GraphBSI · Moses | molecule metrics | Mixed |
| use EM sampler at 500 steps | GraphBSI · Moses | FCD: 1.07 -> 0.72 | Improved |
Closest papers by meaning
- Smooth Interpolation for Improved Discrete Graph Generative Models (ICML 2025 · similarity 0.92)
- DeFoG: Discrete Flow Matching for Graph Generation (ICML 2025 · similarity 0.89)
- Discrete-state Continuous-time Diffusion for Graph Generation (NeurIPS 2024 · similarity 0.89)
- Graph Generation with Diffusion Mixture (ICML 2024 · similarity 0.87)
- Pard: Permutation-Invariant Autoregressive Diffusion for Graph Generation (NeurIPS 2024 · similarity 0.86)
- Variational Flow Matching for Graph Generation (NeurIPS 2024 · similarity 0.86)
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.
