Research map / Learning theory and algorithms / Kernels and regression theory
Divide-and-Conquer for Debiased $l_1$-norm Support Vector Machine in Ultra-high Dimensions
Heng Lian, Zengyan Fan
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 debiasing before aggregation | distributed l1-SVM · Simulation, N=20000, p=5000 | l-infinity error | Improved |
| add debiasing before aggregation (no thresholding) | distributed l1-SVM · Simulation, N=20000, p=5000 | l2 error | Hurt |
| add hard thresholding after debiased aggregation (ablation) | debiased aggregated estimator · Simulation, N=20000, p=5000 | l2 and prediction error | Improved |
| add hard thresholding after debiased aggregation (ablation) | debiased aggregated estimator · Simulation, N=20000, p=5000 | l-infinity error | No clear effect |
| increase number of machines M from 1 to 30 | aggregated estimators · Simulation, N=20000, p=5000 | estimation and prediction error | Hurt |
| increase total sample size N from 10000 to 50000 with n=5000 per machine | thresholded estimator (TE) · Simulation, p=5000 | estimation error | Improved |
| use true Hessian instead of estimated Hessian in debiasing (ablation) | thresholded estimator · Simulation, p=5000 | estimation error | No clear effect |
| iterate: reuse aggregated estimator to re-estimate Hessian on each machine | debiased aggregated estimator · Simulation | estimation error | No clear effect |
| distribute computation instead of centralized estimation | l1-SVM · Simulation, N=10000 to 50000 | computation time: 4 to 29 hours -> about 2 hours | Improved |
| increase p for exchangeable correlation | inverse Hessian norm · Table 1, case 3 | L1 norm: 7.066 -> 7.399 | Hurt |
Closest papers by meaning
- An Error Bound for L1-norm Support Vector Machine Coefficients in Ultra-high Dimension (JMLR 2016 · similarity 0.90)
- Solving L1-regularized SVMs and Related Linear Programs: Revisiting the Effectiveness of Column and Constraint Generation (JMLR 2022 · similarity 0.89)
- On lp-Support Vector Machines and Multidimensional Kernels (JMLR 2020 · similarity 0.87)
- Faster Algorithms for Structured Linear and Kernel Support Vector Machines (ICLR 2025 · similarity 0.86)
- Deep Principal Support Vector Machines for Nonlinear Sufficient Dimension Reduction (ICML 2025 · similarity 0.85)
- Finite Smoothing Algorithm for High-Dimensional Support Vector Machines and Quantile 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.
