Research map / Learning theory and algorithms / Prediction and decision losses
Distributed Feature Screening via Componentwise Debiasing
Xingxiang Li, Runze Li, Zhiming Xia, Chen Xu
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 |
|---|---|---|---|
| replace simple averaging (SAS) with componentwise debiased aggregation (ACS) | distributed SIRS screening · Model (a), m=15, rho=0.8 | screened model size: 20 -> 8 (DC) | Improved |
| replace simple averaging with componentwise debiased aggregation | DC screening · Model (a), m=30, rho=0.8 | screened model size: 1500 -> 8 | Improved |
| replace simple averaging with componentwise debiased aggregation | DC screening · Model (e), m=60 | screened model size: 4800 -> 5 | Improved |
| replace simple averaging with componentwise debiased aggregation | DC screening · Model (f), m=100 | screened model size: 10000 -> 8 | Improved |
| add reinforcement with multiple data partitions (rACS, R=3) (ablation) | ACS with SIRS · Model (a), m=15, rho=0.8 | successful screening rate: .72 -> .94 | Improved |
| add reinforcement with multiple data partitions (rACS, R=3) (ablation) | ACS with SIRS · Model (a), m=30, rho=0.8 | screened model size: 39 -> 8 | Improved |
| add reinforcement with multiple data partitions (rACS) (ablation) | ACS with Kendall tau · Model (a), m=15 | computation time: 0.244 -> 0.732 | Hurt |
| use Pearson correlation under nonlinear model with ACS | ACS · Model (f), m=100 | screened model size: 8600 | No clear effect |
| use distributed ACS instead of centralized screening | Kendall tau screening · Model (a), m=15 | computation time (s): 81.42 -> 0.244 | Improved |
| use distributed ACS instead of centralized screening | DC screening · Model (f), m=100 | computation time (s): 213998 -> 1.923 | Improved |
| increase number of segments m | SAS · Superconductor data | number of features retained: 2479 | Hurt |
| increase number of segments m | ACS · Superconductor data | number of features retained | No clear effect |
Closest papers by meaning
- DECOrrelated feature space partitioning for distributed sparse regression (NeurIPS 2016 · similarity 0.85)
- Scalable Feature Selection via Distributed Diversity Maximization (AAAI 2017 · similarity 0.85)
- Doubly Distributed Supervised Learning and Inference with High-Dimensional Correlated Outcomes (JMLR 2020 · similarity 0.84)
- Distributed Statistical Inference under Heterogeneity (JMLR 2023 · similarity 0.83)
- Making Trillion Correlations Feasible in Feature Grouping and Selection (TPAMI 2016 · similarity 0.83)
- One-shot Distributed Ridge Regression in High Dimensions (ICML 2020 · similarity 0.83)
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.
