Research map / Learning theory and algorithms / Sample complexity
More Efficient Estimation for Logistic Regression with Optimal Subsamples
HaiYing Wang
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 unweighted estimator combining pilot and second-step estimates instead of weighted OSMAC | OSMAC logistic regression · simulation, six covariate distributions | relative efficiency (MSE ratio) | Improved |
| use unweighted estimator instead of weighted with Lopt | OSMAC logistic regression · simulation T3 covariates | relative efficiency | Mixed |
| Poisson subsampling instead of subsampling with replacement | unweighted OSMAC estimator · simulation, six covariate distributions | relative efficiency: up to 250% as efficient as weighted estimator (nzNormal, Aopt) | Improved |
| use Aopt probabilities instead of Lopt | Poisson subsampling · SUSY, n0=200 | relative efficiency | Improved |
| use optimal probabilities instead of uniform subsampling | subsampling estimator · simulation | empirical MSE | Improved |
| increase second-step subsample size n with Lopt | unweighted/Poisson estimators · SUSY, n0=200 | relative efficiency | Improved |
| Poisson subsampling instead of replacement, data on hard drive | OSMAC · CPU time, N=10^7, Lopt | CPU seconds: 4384.99 -> 2699.13 | Improved |
| unweighted combination instead of weighted, data in RAM | Lopt, N=10^7 · CPU time in RAM | CPU seconds: 5.24 -> 3.71 | Improved |
| Aopt instead of Lopt probabilities, data on hard drive | OSMAC · CPU time | CPU seconds: 4384.99 -> 4369.04 | No clear effect |
| Aopt instead of Lopt probabilities, data in RAM | OSMAC · CPU time, N=10^7 | CPU seconds: 3.71 -> 34.98 | Hurt |
| subsampling instead of full data | Poisson subsampling · CPU time on hard drive, N=10^7 | CPU seconds: 138134.69 -> 2699.13 | Improved |
Closest papers by meaning
- Optimal Subsampling with Influence Functions (NeurIPS 2018 · similarity 0.87)
- A Provably Accurate Randomized Sampling Algorithm for Logistic Regression (AAAI 2024 · similarity 0.86)
- Scale-invariant Optimal Sampling for Rare-events Data and Sparse Models (NeurIPS 2024 · similarity 0.84)
- Data subsampling for Poisson regression with pth-root-link (NeurIPS 2024 · similarity 0.84)
- Pairwise Maximum Likelihood For Multi-Class Logistic Regression Model With Multiple Rare Classes (ICML 2025 · similarity 0.83)
- Towards a statistical theory of data selection under weak supervision (ICLR 2024 · similarity 0.82)
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.
