Research map / Learning theory and algorithms / Sample complexity
Nonuniform Negative Sampling and Log Odds Correction with Rare Events Data
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Increase full data size N with N_1^a doubling (10^3 to 10^6) | Full data MLE · Simulation, correct model | N_1^a tr(V_f): 5.41 -> 4.62 | No clear effect |
| Increase full data size N | Full data MLE · Simulation, correct model | N tr(V_f): 169.17 -> 18048.40 | Hurt |
| Increase full data size N | Full data MLE · Simulation, misspecified model | N_1^a tr(V_f): 30.99 -> 11.74 | Helped |
| Increase full data size N | Full data MLE · Simulation, correct model | tr(V_f): 0.169 -> 0.018 | Helped |
| Use optimal sampling with likelihood estimator (optLik) instead of other sampling methods | negative sampling · Simulated logistic regression | log MSE | Helped |
| Use likelihood estimator instead of IPW with optimal sampling | optW · Simulated logistic regression, small sampling rate | log MSE | Helped |
| Increase sampling rate rho from 0.002 to 0.02 | all sampling methods · Simulated logistic regression | log MSE | Helped |
| Misspecify the pilot estimator | optLik and LCC · Simulated logistic regression | log MSE | Hurt |
| Use nonuniform opt sampling instead of uniform at moderate rates [0.01, 0.05] | negative sampling · Internal CTR dataset | test AUC | Helped |
| Use nonuniform opt sampling instead of uniform at extremely small rates [0.001, 0.008] | negative sampling · Internal CTR dataset | test AUC | Mixed |
| Use likelihood estimator instead of IPW | negative sampling · Internal CTR dataset | test AUC | Helped |
| Tune truncation lower bound rho from very small to uniform sampling (ablation) | pi_rho opt sampling · Internal CTR dataset | AUC | Helped |
Closest papers by meaning
- Logistic Regression for Massive Data with Rare Events (ICML 2020 · similarity 0.88)
- Identification and Estimation for Nonignorable Missing Data: A Data Fusion Approach (ICML 2024 · similarity 0.84)
- Concentration and excess risk bounds for imbalanced classification with synthetic oversampling (NeurIPS 2025 · similarity 0.82)
- Boosting Test Performance with Importance Sampling--a Subpopulation Perspective (AAAI 2025 · similarity 0.81)
- Robust Sampling for Active Statistical Inference (NeurIPS 2025 · similarity 0.81)
- A Black-Box Debiasing Framework for Conditional Sampling (NeurIPS 2025 · similarity 0.81)
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.