Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Scalable Deep Gaussian Markov Random Fields for General Graphs
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 |
|---|---|---|---|
| increase DGMRF depth from 1 to 5 layers (ablation) | DGMRF · Chameleon, Squirrel, and Crocodile | RMSE: 1.589 -> 1.465; 1.643 -> 1.374; 1.311 -> 1.169 | Helped |
| increase DGMRF depth from 1 to 5 layers (ablation) | DGMRF · Chameleon, Squirrel, and Crocodile | CRPS: 0.883 -> 0.804; 0.924 -> 0.765; 0.704 -> 0.614 | Helped |
| increase DGMRF depth from 1 to 3 layers (ablation) | DGMRF · California housing, no features | RMSE: 6.909 -> 6.853 | Helped |
| increase DGMRF depth from 1 to 3 layers (ablation) | DGMRF · Wind speed, spatial mask | RMSE: 1.246 -> 1.075 | Helped |
| increase DGMRF depth from 1 to 3 layers (ablation) | DGMRF · Wind speed, random mask | RMSE: 0.272 -> 0.272 | No clear effect |
| make γ trainable rather than fixed at 0 or 1 (ablation) | DGMRF · Synth. DGMRF (L=3) | MAEμ (×100): 0.370±0.008 -> 0.163±0.002 (γ=1); 0.404±0.000 -> 0.163±0.002 (γ=0) | Helped |
| make γ trainable rather than fixed at 0 or 1 (ablation) | DGMRF · Mix synthetic | RMSE (×100): 1.945±0.002 -> 1.584±0.001 (γ=1); 1.587±0.000 -> 1.584±0.001 (γ=0) | Mixed |
| use γ=1 instead of trainable γ (ablation) | DGMRF · Chameleon, L=3 and L=5 | RMSE | Hurt |
| add one DGMRF layer to the variational distribution instead of mean field (ablation) | DGMRF variational inference · Synth. DGMRF (L=3) | ELBO: 2.107±0.000 -> 2.129±0.000 | Helped |
| add one DGMRF layer to the variational distribution instead of mean field (ablation) | DGMRF variational inference · Chameleon | RMSE: 1.527±0.002 -> 1.511±0.006 | Helped |
| use two rather than one DGMRF layer in the variational distribution (ablation) | DGMRF variational inference · Chameleon | RMSE: 1.511±0.006 -> 1.515±0.014 | No clear effect |
| increase DGMRF depth on Mix synthetic data (ablation) | DGMRF · Mix synthetic | RMSE | Helped |
Closest papers by meaning
- Deep Gaussian Markov Random Fields for Graph-Structured Dynamical Systems (NeurIPS 2023 · similarity 0.92)
- Learning in Deep Factor Graphs with Gaussian Belief Propagation (ICML 2024 · similarity 0.87)
- Graph Neural Network Generalization With Gaussian Mixture Model Based Augmentation (ICML 2025 · similarity 0.87)
- Neural Gaussian Similarity Modeling for Differential Graph Structure Learning (AAAI 2024 · similarity 0.86)
- Graph Neural Networks with a Distribution of Parametrized Graphs (ICML 2024 · similarity 0.86)
- Uncertainty Estimation on Graphs with Structure Informed Stochastic Partial Differential Equations (NeurIPS 2025 · 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.