Research map / Graph learning / Graph self-supervised learning
Fitting Autoregressive Graph Generative Models through Maximum Likelihood Estimation
Xu Han, Xiaohui Chen, Francisco J. R. Ruiz, Li-Ping Liu
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 Rout-VI instead of ROS-VI inference network | GraphRNN · Community-small | test log-likelihood (importance sampling): -53.7 -> -25.2 | Improved |
| use Rout-VI instead of ROS-VI inference network | GraphRNN · Enzymes | test log-likelihood (importance sampling): -274.9 -> -92.6 | Improved |
| use a learned variational order distribution (Rout-VI) instead of uniform orders | GraphRNN · Community-small | test log-likelihood (importance sampling): -154.6 -> -25.2 | Improved |
| use Rout-VI instead of uniform node orders | GraphDF · Citeseer-small | test log-likelihood (importance sampling): -181.4 -> -93.5 | Improved |
| use Rout-VI instead of the DFS canonical order | GraphGEN · Community-small | test log-likelihood (importance sampling): -263.74 -> -22.5 | Improved |
| use Rout-VI instead of ROS-VI | GraphGEN · Enzymes | test log-likelihood (importance sampling): -189.7 -> -89.5 | Improved |
| replace GraphGEN with the DAGG attention generative model, both trained with Rout-VI | DAGG · Enzymes | test log-likelihood (importance sampling): -89.5 -> -75.1 | Improved |
| use Rout-VI instead of DFS order | GraphGEN · Community-small | MMD Deg: 0.070 -> 0.094 | Hurt |
| use Rout-VI instead of BFS order | GraphRNN · Cora | MMD Deg: 1.125 -> 0.049 | Improved |
| use Rout-VI instead of ROS-VI inference network | GraphRNN · Inference time | time to sample node orders: 20-times speed-up | Improved |
| use Rout-VI instead of ROS-VI | GraphRNN · Yeast | log-likelihood vs exact (importance sampling check) | No clear effect |
| increase graph size | DAGG and GraphRNN · Cora and Community-small subgraphs | MMD Orbit and Cluster | Hurt |
Closest papers by meaning
- Learning-Order Autoregressive Models with Application to Molecular Graph Generation (ICML 2025 · similarity 0.87)
- Pard: Permutation-Invariant Autoregressive Diffusion for Graph Generation (NeurIPS 2024 · similarity 0.86)
- A Non-Asymptotic Convergent Analysis for Scored-Based Graph Generative Model via a System of Stochastic Differential Equations (ICML 2025 · similarity 0.86)
- Quality Measures for Dynamic Graph Generative Models (ICLR 2025 · similarity 0.85)
- BARNN: A Bayesian Autoregressive and Recurrent Neural Network (ICML 2025 · similarity 0.85)
- SBGD: Improving Graph Diffusion Generative Model via Stochastic Block Diffusion (ICML 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.
