Research map / Learning theory and algorithms / Matrix and tensor methods
A Statistical Perspective on Randomized Sketching for Ordinary Least-Squares
Garvesh Raskutti, Michael W. Mahoney
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 leverage-score sampling without re-scaling instead of with re-scaling | leverage-score sampling sketch · Synthetic least squares, large r (300-1000), nu=1 | relative prediction efficiency C_PE | Improved |
| use leverage-score sampling instead of uniform sampling | sketching for OLS · Synthetic least squares | relative prediction efficiency C_PE | Improved |
| add re-scaling to leverage-score sampling (ablation) | leverage-score sampling sketch · Synthetic least squares | mean-squared error | Hurt |
| use Gaussian or Hadamard projections instead of sampling when r is close to p | sketching for OLS · Synthetic least squares, small r (80-200), nu=1 | relative prediction efficiency C_PE | Improved |
| add shrinkage (theta=0.1) to leverage-score sampling versus no re-scaling (ablation) | S_NR · Synthetic least squares, small r, nu=1 | relative prediction efficiency C_PE | Improved |
| make leverage scores more non-uniform (nu=1 vs nu=10) | rescaled leverage sampling S_R · Synthetic least squares | relative prediction efficiency C_PE | Hurt |
| make leverage scores more non-uniform | Hadamard and Gaussian projections · Synthetic least squares | relative prediction efficiency C_PE | No clear effect |
| measure worst-case error instead of prediction efficiency | all six sketches · Synthetic least squares | relative error | Improved |
| use S_NR instead of S_R under worst-case error | leverage sampling · Synthetic least squares | worst-case relative error C_WC | Improved |
Closest papers by meaning
- Asymptotics for Sketching in Least Squares Regression (NeurIPS 2019 · similarity 0.92)
- Distributed Least Squares in Small Space via Sketching and Bias Reduction (NeurIPS 2024 · similarity 0.91)
- Matrix Product Sketching via Coordinated Sampling (ICLR 2025 · similarity 0.90)
- Fundamental Bias in Inverting Random Sampling Matrices with Application to Sub-sampled Newton (ICML 2025 · similarity 0.88)
- Sketching for Convex and Nonconvex Regularized Least Squares with Sharp Guarantees (ICLR 2025 · similarity 0.88)
- Optimal Sketching for Residual Error Estimation for Matrix and Vector Norms (ICLR 2024 · similarity 0.87)
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.
