Research map / Learning theory and algorithms / Kernels and regression theory
Improved Shrinkage Prediction under a Spiked Covariance Structure
Trambak Banerjee, Gourab Mukherjee, Debashis Paul
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 coordinate-wise shrinkage factors (CASP versus CASP with all factors 1) (ablation) | CASP · Simulation Exp 1, Scenario 1 | REE: 1.00 -> 0.95 | Improved |
| use CASP instead of Bcv when the covariance is a spiked structure | CASP · Simulation Exp 1, Scenario 2 | REE: 4.24 -> 1.00 | Improved |
| add shrinkage when beta is large (beta=1.75) (ablation) | CASP · Simulation Exp 1, Scenario 2 | REE: 1.00 | No clear effect |
| use CASP instead of Bcv under heteroscedastic factor model | CASP · Simulation Exp 2, Scenario 1 | REE: 0.87 -> 0.91 | Hurt |
| use CASP instead of FactMLE/POET under heteroscedastic factor model | CASP · Simulation Exp 2, Scenario 1 | REE: 1.49 -> 0.91 | Improved |
| use CASP instead of Bcv when factor signal is weak (scenario 2) | CASP · Simulation Exp 2, Scenario 2 | REE: 1.23 -> 1.00 | Improved |
| use CASP when covariance is AR(1) rather than spiked | CASP · Simulation Exp 3, Scenario 1 | REE: 1.23 -> 0.94 | Improved |
| use CASP instead of Bcv under AR(1) covariance | CASP · Simulation Exp 3, Scenario 2 | REE: 4.27 -> 1.00 | Improved |
| use CASP instead of unshrunk previous-week sales | CASP · Grocery sales Coffee | loss ratio: 1.838 -> 0.999 | Improved |
| use CASP instead of unshrunk previous-week sales | CASP · Grocery sales Carb. Beverages | loss ratio: 8.885 -> 0.984 | Improved |
| add coordinate-wise shrinkage on real data (ablation) | CASP · Grocery sales | loss ratio: 1.002 | No clear effect |
Closest papers by meaning
- Prediction Risk for the Horseshoe Regression (JMLR 2019 · similarity 0.87)
- Bayesian Tensor Regression (JMLR 2017 · similarity 0.86)
- Interpolating Predictors in High-Dimensional Factor Regression (JMLR 2022 · similarity 0.85)
- Prediction-Powered Adaptive Shrinkage Estimation (ICML 2025 · similarity 0.85)
- A flexible empirical Bayes approach to multiple linear regression and connections with penalized regression (JMLR 2024 · similarity 0.85)
- TIC-TAC: A Framework For Improved Covariance Estimation In Deep Heteroscedastic Regression (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.
