Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Sketching in High-Dimensional Regression With Big Data Using Gaussian Scale Mixture Priors
Rajarshi Guhaniyogi, Aaron Wolfe Scheffler
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 |
|---|---|---|---|
| increase sketch size m from 100 to 500 (ablation) | CHS · Simulation 1 (Scenario 1, s=50) | avg accuracy (zero coefficients): 0.64 -> 0.95 | Improved |
| increase sketch size m from 100 to 500 (ablation) | CHS · Simulation 1 (Scenario 1, s=50) | avg accuracy (nonzero coefficients): 0.61 -> 0.86 | Improved |
| increase sketch size m from 100 to 500 at high sparsity (ablation) | CHS · Simulation 1 (Scenario 1, s=10) | avg accuracy (zero coefficients): 0.88 -> 0.98 | Improved |
| increase sketch size m from 100 to 500 (ablation) | CHS · Simulation 1 (Scenario 1, s=10) | computational efficiency: 2.83 -> 0.85 | Hurt |
| increase number of nonzero coefficients s from 10 to 50 (m=100) (ablation) | CHS · Simulation 1 (Scenario 1) | avg accuracy (zero coefficients): 0.88 -> 0.64 | Hurt |
| use correlated (compound) instead of independent features (ablation) | CHS · Simulation 1 | avg accuracy | No clear effect |
| sketch with Gaussian random projection instead of subsampling m points (PMCP) | CHS vs PMCP · Simulation 2 (Scenario 1, m=400, s=50) | MSPE x10^3: 33.19 -> 0.61 | Improved |
| increase m from 200 to 400 (ablation) | CHS · Simulation 2 (Scenario 1, s=50) | MSPE x10^3: 205.67 -> 0.61 | Improved |
| increase m from 200 to 400 (ablation) | CHS · Simulation 2 (Scenario 1, s=50) | coverage of nonzero coefficients: 0.68 -> 0.97 | Improved |
| use full-data MCP instead of sketched CHS | CHS · Simulation 2 (Scenario 1, m=400, s=10) | MSPE x10^3: 0.51 -> 0.02 | Improved |
| increase m from 300 to 800 (ablation) | CHS · Orthopedic fractures (SOF) | correlation of predicted and observed response: 0.87 (m=800) vs 0.91 for MCP | Improved |
| decrease m from 800 to 300 (ablation) | CHS · Orthopedic fractures (SOF) | 95% predictive interval length: 0.17 -> 0.48 | Hurt |
Closest papers by meaning
- Bayesian Sparse Gaussian Mixture Model for Clustering in High Dimensions (JMLR 2025 · similarity 0.88)
- Bayesian Data Sketching for Varying Coefficient Regression Models (JMLR 2025 · similarity 0.87)
- Compressed Gaussian Process for Manifold Regression (JMLR 2016 · similarity 0.86)
- Scalable Approximate MCMC Algorithms for the Horseshoe Prior (JMLR 2020 · similarity 0.86)
- Turbocharging Gaussian Process Inference with Approximate Sketch-and-Project (NeurIPS 2025 · similarity 0.85)
- Data Mixture Optimization: A Multi-fidelity Multi-scale Bayesian Framework (NeurIPS 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.
