Research map / Probabilistic and scientific modeling / Variational and generative sampling
High-dimensional Gaussian graphical models on network-linked data
Tianxi Li, Cheng Qian, Elizaveta Levina, Ji Zhu
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 network cohesion (GNC-lasso) instead of plain glasso | glasso · lattice network, t=1 | mean error ||M_hat-M||_inf: 1.037 -> 0.820 | Improved |
| use network cohesion (GNC-lasso) instead of plain glasso | glasso · coauthorship network, t=1 | mean error ||M_hat-M||_inf: 4.795 -> 0.942 | Improved |
| use the network instead of K-means oracle clustering | oracle cluster+glasso · coauthorship network, t=1 | mean error ||M_hat-M||_inf: 1.342 -> 0.942 | Improved |
| use GNC-lasso instead of cluster+glasso on lattice (estimation error) | oracle cluster+glasso · lattice network, t=0.1 | mean error ||M_hat-M||_inf: 0.493 -> 0.419 (t=0.1); 0.539 -> 0.669 (t=0.5) | Mixed |
| add oracle AUC tuning instead of 10-fold cross-validation (ablation) | GNC-lasso · lattice network, t=0.5 | mean error ||M_hat-M||_inf: 0.669 -> 0.520 | Improved |
| increase heterogeneity of the mean (t from 0.1 to 1) for glasso (ablation) | glasso · lattice network | mean error ||M_hat-M||_inf: 0.358 -> 1.037 | Hurt |
| increase heterogeneity of the mean (t from 0.1 to 1) for GNC-lasso (ablation) | GNC-lasso · lattice network | mean error ||M_hat-M||_inf: 0.419 -> 0.820 | Hurt |
| split lattice into 16 disconnected components (ablation) | oracle GNC-lasso · lattice network, t=0.5 | mean error ||M_hat-M||_inf: 0.520 -> 0.785 | Hurt |
| split lattice into 16 disconnected components (ablation) | glasso · lattice network, t=0.5 | mean error ||M_hat-M||_inf: 0.746 -> 2.942 | Hurt |
| split lattice into 16 components for CV-tuned GNC-lasso (ablation) | GNC-lasso · lattice network, t=0.5 | mean error ||M_hat-M||_inf: 0.669 -> 0.838 | Hurt |
| split coauthorship network into 4 components (ablation) | oracle GNC-lasso · coauthorship network, t=0.5 | mean error ||M_hat-M||_inf: 0.520 -> 0.958 | Hurt |
| reduce sample size from n=400 to n=100 (ablation) | GNC-lasso · lattice network, p=500, t=0.5 | mean error ||M_hat-M||_inf: 0.669 -> 0.874 | Hurt |
Closest papers by meaning
- High-Dimensional Inference for Cluster-Based Graphical Models (JMLR 2020 · similarity 0.89)
- Quantifying Network Similarity using Graph Cumulants (JMLR 2023 · similarity 0.88)
- Simultaneous Inference for Pairwise Graphical Models with Generalized Score Matching (JMLR 2020 · similarity 0.88)
- A Generic Family of Graphical Models: Diversity, Efficiency, and Heterogeneity (ICML 2025 · similarity 0.86)
- Learning Local Neighborhoods of Non-Gaussian Graphical Models (AAAI 2025 · similarity 0.86)
- Graph Structure Inference with BAM: Neural Dependency Processing via Bilinear Attention (NeurIPS 2024 · 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.
