Research map / Learning theory and algorithms / Sample complexity
Learning Identifiable Gaussian Bayesian Networks in Polynomial Time and Sample Complexity
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 |
|---|---|---|---|
| Use terminal-vertex elimination with CLIME precision estimation and OLS instead of PC | Gaussian Bayesian network structure learning · Synthetic GBNs (p = 50, 100, 150, 200) | precision: PC: 0.587 ± 0.015, 0.587 ± 0.008, 0.572± 0.006, 0.573± 0.005; Ours: 1.000 ± 0.000, 1.000 ± 0.000, 1.000± 0.000, 1.000± 0 | Helped |
| Use terminal-vertex elimination with CLIME precision estimation and OLS instead of MMHC | Gaussian Bayesian network structure learning · Synthetic GBNs (p = 50, 100, 150, 200) | precision: MMHC: 0.581± 0.038, 0.529± 0.019, 0.566± 0.014, 0.582± 0.012; Ours: 1.000 ± 0.000, 1.000 ± 0.000, 1.000± 0.000, 1.000± 0 | Helped |
| Use terminal-vertex elimination with CLIME precision estimation and OLS instead of GES | Gaussian Bayesian network structure learning · Synthetic GBNs (p = 50, 100, 150, 200) | recall: GES: 0.396 ± 0.031, 0.372 ± 0.020, 0.333± 0.017, 0.310± 0.011; Ours: 1.000 ± 0.000, 1.000 ± 0.000, 1.000± 0.000, 1.000± | Helped |
| Use terminal-vertex elimination with CLIME precision estimation and OLS instead of PC | Gaussian Bayesian network structure learning · Synthetic GBNs (p = 50, 100, 150, 200) | recall: PC: 0.996 ± 0.004, 0.999 ± 0.001, 0.996± 0.002, 0.997± 0.001; Ours: 1.000 ± 0.000, 1.000 ± 0.000, 1.000± 0.000, 1.000± 0 | Helped |
| Use terminal-vertex elimination with CLIME precision estimation and OLS instead of MMHC | Gaussian Bayesian network structure learning · Synthetic GBNs (p = 50, 100, 150, 200) | recall: MMHC: 0.583± 0.038, 0.533± 0.019, 0.577± 0.015, 0.593± 0.012; Ours: 1.000 ± 0.000, 1.000 ± 0.000, 1.000± 0.000, 1.000± 0 | Helped |
| Use terminal-vertex elimination with CLIME precision estimation and OLS instead of GES | Gaussian Bayesian network structure learning · Synthetic GBNs (p = 50, 100, 150, 200) | running time: p = 50: 0.089 ± 0.005 vs 0.206 ± 0.025; p = 100: 0.534 ± 0.004 vs 0.557 ± 0.045; p = 150: 1.988 ± 0.010 vs 1.031 ± 0.036 | Mixed |
| Use terminal-vertex elimination with CLIME precision estimation and OLS instead of MMHC | Gaussian Bayesian network structure learning · Synthetic GBNs (p = 50, 100, 150, 200) | running time: MMHC: 0.460± 0.049, 1.417± 0.141, 2.934± 0.241, 5.511± 0.355; Ours: 0.089 ± 0.005, 0.534 ± 0.004, 1.988 ± 0.010, 5.130 ± | Helped |
| Allow small differences in noise variances rather than requiring equal noise variance | The proposed GBN learning algorithm · Synthetic GBNs with perturbed noise variance | precision and recall: 0.125; 0.9375 and 1.0625; γ = 0 | Mixed |
| Order nodes by marginal variance instead of identifying terminal vertices | Marginal-variance algorithm · Synthetic GBNs (p = 200) | structure recovery failure rate: 50% | Hurt |
Closest papers by meaning
- Beyond normality: Learning sparse probabilistic graphical models in the non-Gaussian setting (NeurIPS 2017 · similarity 0.90)
- Minimal I-MAP MCMC for Scalable Structure Discovery in Causal DAG Models (ICML 2018 · similarity 0.89)
- Robust Learning of Fixed-Structure Bayesian Networks in Nearly-Linear Time (ICLR 2021 · similarity 0.89)
- Learning Gaussian DAG Models without Condition Number Bounds (ICML 2025 · similarity 0.89)
- A Fixed-Parameter Tractable Algorithm for Counting Markov Equivalence Classes with the Same Skeleton (AAAI 2024 · similarity 0.87)
- ProDAG: Projected Variational Inference for Directed Acyclic Graphs (NeurIPS 2025 · similarity 0.87)
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.