Research map / Learning theory and algorithms / Kernels and regression theory
Scalable Resampling in Massive Generalized Linear Models via Subsampled Residual Bootstrap
Indrila Ganguly, Srijan Sengupta, Sujit Ghosh
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 SRB with b=n^0.8 instead of full residual bootstrap (ablation) | SRB · Linear regression, n=1e5, p=200 | runtime (s): 4.60 -> 0.91 | Improved |
| Use SRB with b=n^0.8 instead of full residual bootstrap (ablation) | SRB · Linear regression, n=2e5, p=300 | runtime (s): 19.89 -> 4.91 | Improved |
| Use SRB with b=n^0.8 instead of full residual bootstrap (ablation) | SRB · Logistic regression, n=50000, p=200 | error rate (%): 0.84 -> 0.82 | No clear effect |
| Use SRB with b=n^0.8 instead of full residual bootstrap (ablation) | SRB · Logistic regression, n=50000, p=200 | runtime (s): 1.51 -> 0.51 | Improved |
| Use SRB with b=n^0.8 instead of full residual bootstrap (ablation) | SRB · Poisson regression, n=1e5, p=300 | runtime (s): 5.85 -> 2.50 | Improved |
| Reduce subsample size from b=n^0.6 to b=n^0.5 (ablation) | SRB · Linear regression, n=1e5, p=200 | error rate (%): 9.15 -> 16.86 | Hurt |
| Reduce subsample size from b=n^0.7 to b=n^0.5 (ablation) | SRB · Poisson regression, n=50000, p=200 | error rate (%): 0.97 -> 6.93 | Hurt |
| Reduce subsample size from b=n^0.8 to b=n^0.5 (ablation) | SRB · Logistic regression, n=50000, p=200 | error rate (%): 0.82 -> 0.90 | No clear effect |
| Increase subsample size b from n^0.6 to n^0.9 (ablation) | SRB · Linear regression, n=1e5, p=200 | runtime (s): 0.69 -> 1.61 | Hurt |
| Increase subsample size b from n^0.6 to n^0.9 (ablation) | SRB · Poisson regression, n=50000, p=200 | error rate (%): 2.28 -> 0.87 | Improved |
| Increase b from n^0.5 to n^0.7 (ablation) | SRB · Poisson regression, n=1e5, p=300 | error rate (%): 6.59 -> 0.84 | Improved |
| Use SRB on Forest Covertype with larger gamma (ablation) | SRB · Forest Covertype logistic regression q95 | q95 estimate | Improved |
Closest papers by meaning
- Small Resamples, Sharp Guarantees: Convergence Rates for Resampled Studentized Quantile Estimators (NeurIPS 2025 · similarity 0.86)
- Robust Inference for High-Dimensional Linear Models via Residual Randomization (ICML 2021 · similarity 0.86)
- Statistical Inference for High-dimensional Partially Linear Models via Debiased Rank Lasso (JMLR 2026 · similarity 0.86)
- Bootstrap in High Dimension with Low Computation (ICML 2023 · similarity 0.86)
- Learning No-Regret Sparse Generalized Linear Models with Varying Observation(s) (ICLR 2024 · similarity 0.86)
- Scale-invariant Optimal Sampling for Rare-events Data and Sparse Models (NeurIPS 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.
