Research map / Learning theory and algorithms / Matrix and tensor methods
Feature-distributed sparse regression: a screen-and-clean approach
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 sketching (sketch size n ~ s log d) in the cleaning stage instead of sending full Gram data | feature-distributed sparse regression (SCREEN-AND-CLEAN) · theoretical communication cost | communication cost (bits): O(mN^2) reduced to Õ(mNs) | Helped |
| Increase sketch size n so contraction factor shrinks (fewer iterations, more bits per round) (ablation) | SCREEN-AND-CLEAN cleaning stage · simulation N=1000, D=10000, s=10 | prediction error convergence per iteration | Helped |
| Screen features by SIS before cleaning (vs. using all D features) | SCREEN-AND-CLEAN · simulation D=3000, varying N | prediction error | No clear effect |
| Replace DECO with SCREEN-AND-CLEAN (SC) | feature-distributed sparse regression · simulation D=3000, varying N, Gaussian and correlated designs | prediction error | Helped |
| Spark implementation: increase number of machines from 1 to 16 | SCREEN-AND-CLEAN Spark implementation · 6000 x 200000 Gaussian, 20 nodes | running time (s): running time roughly halves per doubling of machines | Helped |
Closest papers by meaning
- Distributed Sparse Regression via Penalization (NeurIPS 2024 · similarity 0.89)
- Compressed and distributed least-squares regression: convergence rates with applications to federated learning (ICML 2025 · similarity 0.86)
- Better Locally Private Sparse Estimation Given Multiple Samples Per User (ICML 2024 · similarity 0.86)
- Implicit Regularization of Decentralized Gradient Descent for Sparse Regression (NeurIPS 2024 · similarity 0.86)
- Distributed High-Dimensional Quantile Regression: Estimation Efficiency and Support Recovery (ICML 2024 · similarity 0.85)
- Accelerated Methods with Compressed Communications for Distributed Optimization Problems Under Data Similarity (AAAI 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.