Research map / Probabilistic and scientific modeling / Causal discovery
Learning Gaussian DAGs from Network Data
Hangjian Li, Oscar Hernan Madrid Padilla, Qing Zhou
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 |
|---|---|---|---|
| model sample correlation jointly via BCD instead of assuming independent samples | DAG estimation with known ordering · Simulated equi-cor, n<p | SHD: 523.0 -> 514.2 | Improved |
| model sample correlation jointly via BCD instead of assuming independent samples | DAG estimation with known ordering · Simulated toeplitz, n<p | SHD: 599.9 -> 347.8 | Improved |
| model sample correlation jointly via BCD instead of assuming independent samples | DAG estimation with known ordering · Simulated toeplitz, n>p | SHD: 176.3 -> 80.5 | Improved |
| model sample correlation jointly via BCD instead of assuming independent samples | DAG estimation with known ordering · Simulated equi-cor, n>p | SHD: 147.6 -> 93.2 | Improved |
| use BCD with L1 on DAG instead of KGLasso | DAG estimation with known ordering · Simulated equi-cor, n<p | SHD: 1166.0 -> 514.2 | Improved |
| use BCD instead of KGLasso | DAG estimation with known ordering · Simulated AR(5), n>p | SHD: 102.4 -> 90.3 | Improved |
| use BCD instead of KGLasso | DAG estimation with known ordering · Simulated SBM, n>p | SHD: 251.0 -> 84.5 | Improved |
| increase lambda_p from 0.01/p to 0.1 in graphical Lasso step when n>p | BCD · Simulated n>p ROC/AUC | AUC | Improved |
| use one-iteration estimate Theta^(1) instead of converged Theta^(infinity) | BCD · Simulated ordered DAG, Theta estimation | err(Theta) | No clear effect |
| model sample correlation via BCD | DAG estimation with known ordering · Andes+facebook real network, n=500 | SHD: 235.9 -> 218.6 | Improved |
| model sample correlation via BCD | DAG estimation with known ordering · Hepar2+celegans_n306, n=100 | SHD: 357.9 -> 343.7 | Improved |
| model sample correlation via BCD with non-block-diagonal Theta | DAG estimation with known ordering · Andes+facebook general Theta, n=100 | SHD: 398.6 -> 387.9 | Improved |
Closest papers by meaning
- DAGs with NO TEARS: Continuous Optimization for Structure Learning (NeurIPS 2018 · similarity 0.93)
- Learning Gaussian DAG Models without Condition Number Bounds (ICML 2025 · similarity 0.90)
- Causal Discovery via Bayesian Optimization (ICLR 2025 · similarity 0.89)
- Boosting Causal Structure Learning: An Asymmetric Exponential Modulation Gaussian-Based Adaptive Sample Reweighting Framework (AAAI 2025 · similarity 0.89)
- Deep Gaussian Processes on Directed Acyclic Graphs (NeurIPS 2026 · similarity 0.89)
- ProDAG: Projected Variational Inference for Directed Acyclic Graphs (NeurIPS 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.
