Research map / Learning theory and algorithms / Matrix and tensor methods
Dual Extrapolation for Sparse GLMs
Mathurin Massias, Samuel Vaiter, Alexandre Gramfort, Joseph Salmon
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 dual extrapolation to Gap Safe screening (ablation) | Gap Safe Rules · Lasso paths (leukemia, news20, rcv1) | computation time: up to a factor 3 | Improved |
| add dual extrapolation to the working set solver (ablation) | Celer w/o extr. · Lasso paths (coarse and fine grids) | computation time | Improved |
| use lower stopping tolerance with extrapolation (ablation) | Celer · Lasso paths | computation time | Improved |
| use working sets instead of screening on a coarse grid (ablation) | Celer / Blitz vs GSR · Lasso paths | computation time | Mixed |
| replace Blitz and scikit-learn with Celer | Celer · Lasso Finance, lambda_max/20 | computation time: Celer 5 / Blitz 25 / scikit-learn 470 s at eps 1e-2; Celer 10 vs Blitz 30 at 1e-6 | Improved |
| add dual extrapolation to cyclic CD screening and a dual-based working set (ablation) | Gap Safe / Celer · Logistic regression path news20 | computation time | Improved |
| add dual extrapolation to Prox-Newton working set solver (ablation) | Newton-Celer · Logistic regression rcv1 (eps 1e-6) | computation time | No clear effect |
| replace the QP-based coefficients with dual extrapolation coefficients (ablation) | dual extrapolation · Lasso leukemia, lambda_max/5 | computation time | Improved |
| use extrapolated dual point instead of rescaled residuals (ablation) | Lasso, logistic, multitask Lasso · news20, rcv1, MEG | computation time | Improved |
| use Celer working sets instead of Gap Safe rules (ablation) | Celer · Multitask Lasso MEG paths | computation time | Improved |
Closest papers by meaning
- Learning No-Regret Sparse Generalized Linear Models with Varying Observation(s) (ICLR 2024 · similarity 0.90)
- Scalable First-order Method for Certifying Optimal k-Sparse GLMs (ICML 2025 · similarity 0.89)
- Celer: a Fast Solver for the Lasso with Dual Extrapolation (ICML 2018 · similarity 0.88)
- Generalized Linear Model Regression under Distance-to-set Penalties (NeurIPS 2017 · similarity 0.87)
- Scaled Least Squares Estimator for GLMs in Large-Scale Problems (NeurIPS 2016 · similarity 0.87)
- Estimating Generalization Performance Along the Trajectory of Proximal SGD in Robust Regression (NeurIPS 2024 · 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.
