Research map / Time series / Causal and generative time series
Low Tree-Rank Bayesian Vector Autoregression Models
Leo L Duan, Zeyu Yuwen, George Michailidis, Zhengwu Zhang
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 tree-rank (union of spanning trees) prior to shrinkage-only Bayesian VAR (ablation) | Bayesian VAR graph estimation · HCP fMRI test-retest | Jaccard index: 0.750 -> 0.951 | Improved |
| add continuous shrinkage to the trees-only model (ablation) | Bayesian VAR graph estimation · HCP fMRI test-retest | Jaccard index: 0.941 -> 0.951 | Improved |
| calibrate shrinkage-only prior to be sparser (alpha_eta 30, gamma_eta 0.0001) (ablation) | shrinkage-only Bayesian VAR · HCP fMRI test-retest | Jaccard index: 0.750 -> 0.632; 257 edges on test, 648 on retest | Hurt |
| use proposed tree-rank model instead of lasso | VAR graph estimation · HCP fMRI test-retest | Jaccard index: 0.737 -> 0.951 | Improved |
| use proposed tree-rank model instead of elastic net | VAR graph estimation · HCP fMRI test-retest | Jaccard index: 0.721 -> 0.951 | Improved |
| add tree-rank prior when ground truth is a random sparse graph (about 5% edges) | Bayesian VAR shrinkage-only · synthetic VAR, p=30 and 80, T 400-1200 | estimation error | Improved |
| use trees-only model when ground truth is low tree-rank 2 graph (ablation) | Bayesian VAR · synthetic VAR, tree-rank 2 | estimation error | No clear effect |
| replace generalized Pareto shrinkage with horseshoe prior (ablation) | shrinkage-only model · synthetic VAR | estimation error | No clear effect |
| relax tree-rank bound from m<=5 to m<=10 (ablation) | proposed model · synthetic small-world graph, p=80, T=1200 | graph recovery: m<=5 sparser than truth; m<=10 very close to ground truth | Improved |
| relax tree-rank bound from m<=10 to m<=30 (ablation) | proposed model · synthetic three-component graph (20,30,30 nodes), T=1200 | graph recovery: m<=10 sparser than truth; m<=30 matches | Improved |
Closest papers by meaning
- Generating Hypotheses of Dynamic Causal Graphs in Neuroscience: Leveraging Generative Factor Models of Observed Time Series (ICML 2025 · similarity 0.85)
- TNPAR: Topological Neural Poisson Auto-Regressive Model for Learning Granger Causal Structure from Event Sequences (AAAI 2024 · similarity 0.85)
- Identification of Latent Confounders via Investigating the Tensor Ranks of the Nonlinear Observations (ICML 2025 · similarity 0.84)
- Jacobian Regularizer-based Neural Granger Causality (ICML 2024 · similarity 0.84)
- Estimating Interventional Distributions with Uncertain Causal Graphs through Meta-Learning (NeurIPS 2025 · similarity 0.84)
- Discovering Mixtures of Structural Causal Models from Time Series Data (ICML 2024 · similarity 0.84)
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.
