Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Empirical Priors for Prediction in Sparse High-dimensional Linear Regression
Ryan Martin, Yiqi Tang
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 EB empirical prior instead of horseshoe | sparse linear regression prediction · Very-sparse simulations | MSPE: 0.86 vs 0.89 | No clear effect |
| use EB instead of horseshoe at moderate density | sparse linear regression prediction · Less-sparse simulations (p=125, Delta=0.2) | MSPE: 2.53 -> 1.43 | Improved |
| use EB instead of lasso | sparse linear regression prediction · Very-sparse simulations | MSPE: 1.18 -> 0.86 | Improved |
| use EB instead of adaptive lasso at large signal | sparse linear regression prediction · Very-sparse simulations (p=125, A=8, r=0.8) | MSPE: 8.99 -> 1.22 | Improved |
| use EB instead of HS in the hardest setting | sparse linear regression prediction · Less-sparse p=500, Delta=0.05, A=8, r=0.2 | MSPE: 6.44 -> 1.57 | Improved |
| use EB instead of HS for interval length | 95% prediction intervals · Less-sparse (p=125, Delta=0.2) | interval mean length: 5.59 -> 4.20 | Improved |
| use EB instead of HS for interval coverage | 95% prediction intervals · Less-sparse p=500, Delta=0.05, A=8, r=0.5 | coverage: 0.93 -> 0.88 | Hurt |
| use EB instead of HS for very-sparse interval length at p=500 | 95% prediction intervals · Very-sparse p=500 | interval mean length: 4.12 vs 4.10 | Mixed |
| use EB instead of HS for computation | posterior sampling · Simulations | run-time: about 20% of HS run-time | Improved |
| use discrete selection prior EB instead of continuous shrinkage HS when signals are tiny | NCI-60 gene B1 · NCI-60 multidrug, B1 | MSPE: 0.54 -> 0.73 | Hurt |
| use EB instead of lasso/HS on real gene responses | NCI-60 genes A2, A8, A12 · NCI-60 multidrug | MSPE: 0.93 vs 0.94 | No clear effect |
Closest papers by meaning
- A flexible empirical Bayes approach to multiple linear regression and connections with penalized regression (JMLR 2024 · similarity 0.93)
- Covariate-Powered Empirical Bayes Estimation (NeurIPS 2019 · similarity 0.88)
- Bayesian Model Selection with Graph Structured Sparsity (JMLR 2020 · similarity 0.88)
- Prior Specification for Bayesian Matrix Factorization via Prior Predictive Matching (JMLR 2023 · similarity 0.87)
- Sparse Variational Student-t Processes (AAAI 2024 · similarity 0.86)
- Bayesian Additive Distribution Regression (NeurIPS 2026 · 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.
