Research map / Learning theory and algorithms / Matrix and tensor methods
Estimation of Sparse Gaussian Graphical Models with Hidden Clustering Structure
Meixia Lin, Defeng Sun, Kim-Chuan Toh, Chengjing Wang
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 |
|---|---|---|---|
| add l1 penalty on pairwise differences of precision entries (hidden clustering) (ablation) | sparse Gaussian graphical model · AR(10) model, p = 5n | AUC: 0.68224 -> 0.87663 | Improved |
| add l1 penalty on pairwise differences of precision entries (ablation) | sparse Gaussian graphical model · Modular graph recovery, p = 0.5n | AUC: 0.59601 -> 0.65797 | Improved |
| add l1 penalty on pairwise differences of precision entries (ablation) | sparse Gaussian graphical model · Edge coloring model, p = 0.5n | AUC: 0.76952 -> 0.9224 | Improved |
| add l1 penalty on pairwise differences of precision entries (ablation) | sparse Gaussian graphical model · Edge coloring model, p = 50n | AUC: 0.85703 -> 0.86017 | No clear effect |
| add exact AR(k=10) structure constraints (ablation) | our unconstrained estimator · AR(10) model, p = 0.5n | AUC: 0.76864 -> 1 | Improved |
| add AR(k=20) constraints with overestimated k (ablation) | our unconstrained estimator · AR(10) model, p = 0.5n | AUC: 0.76864 -> 0.96337 | Improved |
| add linear constraints for modular structure (ablation) | our unconstrained estimator · Modular graph recovery, p = 10n | AUC: 0.85913 -> 0.87068 | Improved |
| add linear constraints for modular structure (ablation) | our unconstrained estimator · Modular graph recovery, p = 0.5n | AUC: 0.65797 -> 0.67838 | Improved |
| use two-phase algorithm (sGS-ADMM then pALM) instead of stand-alone sGS-ADMM | solver for model (P) · AR models, n = 1000, 47971 constraints, rho = 0.1 | time: 00:05:00 limit hit -> 00:36:47 | Improved |
| use two-phase algorithm instead of stand-alone sGS-ADMM | solver for model (P) · Modular graph recovery, n = 1000, nG = 100, rho = 0.1 | time: 02:02:43 -> 00:13:05 | Improved |
| add pALM Phase II after sGS-ADMM Phase I (ablation) | two-phase solver · modular graph recovery runtime | time | Improved |
| use proposed solvers instead of alternating linearization method | solver · unconstrained model (2), large n | time | Improved |
Closest papers by meaning
- Speeding Up Latent Variable Gaussian Graphical Model Estimation via Nonconvex Optimization (NeurIPS 2017 · similarity 0.90)
- Estimation of Graphical Models through Structured Norm Minimization (JMLR 2018 · similarity 0.88)
- A Fast and Scalable Joint Estimator for Integrating Additional Knowledge in Learning Multiple Related Sparse Gaussian Graphical Models (ICML 2018 · similarity 0.88)
- A Completely Tuning-Free and Robust Approach to Sparse Precision Matrix Estimation (ICML 2022 · similarity 0.88)
- Learning Local Neighborhoods of Non-Gaussian Graphical Models (AAAI 2025 · similarity 0.86)
- Learning Gaussian DAG Models without Condition Number Bounds (ICML 2025 · similarity 0.85)
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.
