Research map / Learning theory and algorithms / Kernels and regression theory
Scalable High-Dimensional Multivariate Linear Regression for Feature-Distributed Data
Shuo-Chieh Huang, Ruey S. Tsay
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 two-stage design (just-in-time stopping then second-stage RGA) instead of plain RGA with L_n=500 (ablation) | TSRGA · Simulation Spec 1, heavy-tailed predictors | parameter estimation error | Improved |
| use larger L_n=500 instead of oracle L_n equal to sum of |beta*| (ablation) | RGA · Simulation Spec 1 | convergence speed | Improved |
| update 50% instead of 25% of coordinates per node (ablation) | Hydra · Simulation Spec 1 | convergence rate | Improved |
| switch from independent to equicorrelated predictors (correlation 0.5) (ablation) | Hydra-type algorithms · Simulation Spec 2 | convergence rate | Hurt |
| increase number of processes from 16 to 64 (ablation) | TSRGA · Large-scale feature-distributed data (n=20000, p=1024) | wall-clock time: about 16 minutes -> less than 5 minutes | Improved |
| use TSRGA instead of iRRR and oracle least squares | multivariate low-rank regression · Simulation Spec 3, n=200 | parameter estimation error: 0.666 vs 0.929 (iRRR) and 0.851 (Oracle LS) | Improved |
| use TSRGA instead of iRRR and oracle least squares | multivariate low-rank regression · Simulation Spec 4, n=1200 | parameter estimation error: 0.751 vs 2.419 (Oracle LS) | Improved |
| use group-wise VAR instead of joint VAR (ablation) | VAR · S&P 500 financial prediction | RMSE | Improved |
| add reduced-rank constraint to VAR (ablation) | gVAR · S&P 500 financial prediction | RMSE | Mixed |
| add 10-K text features with low-rank estimation (TSRGA, iRRR) (ablation) | gVAR · S&P 500 volatility prediction | RMSE: 10% reduction in RMSE compared with gVAR and RR | Improved |
| add 10-K text features (ablation) | prediction of volume, beta, return · S&P 500 financial prediction | RMSE | No clear effect |
Closest papers by meaning
- Feature-distributed sparse regression: a screen-and-clean approach (NeurIPS 2016 · similarity 0.85)
- A convex optimization formulation for multivariate regression (NeurIPS 2020 · similarity 0.85)
- Structured Sparse Regression via Greedy Hard Thresholding (NeurIPS 2016 · similarity 0.84)
- DECOrrelated feature space partitioning for distributed sparse regression (NeurIPS 2016 · similarity 0.84)
- Solving Sparse \& High-Dimensional-Output Regression via Compression (NeurIPS 2024 · similarity 0.83)
- Distributed High-Dimensional Quantile Regression: Estimation Efficiency and Support Recovery (ICML 2024 · similarity 0.83)
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.
