Research map / Learning theory and algorithms / Sample complexity
Sharp Empirical Bernstein Inequalities for the Variance of Bounded Random Variables
Diego Martinez-Taboada, Aaditya Ramdas
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 empirical Bernstein variance confidence intervals instead of Maurer-Pontil self-bounding intervals | variance/std confidence intervals · batch setting, uniform(0,1), beta(2,6), beta(5,5) | confidence interval width | Improved |
| replace the variance confidence sequences in adaptive Neyman allocation with the proposed upper confidence sequence | adaptive Neyman allocation (Neopane et al., 2025) · ATE sequential estimation, uniform and beta potential outcomes | mean squared error of ATE | Improved |
| estimate variance of squared deviations empirically instead of upper bounding it by sigma squared (ablation) | variance confidence interval first-order width · asymptotic width, iid data | first-order width: 2 sigma sqrt(2 log(2/alpha)) -> sqrt(2 V[(X-mu)^2] log(1/alpha)) | Improved |
| express variance as E[X^2] - mu^2 and apply bounded-variable inequalities to both moments with a union bound (ablation) | variance confidence interval · theory and appendix comparison | interval tightness | Hurt |
| use alpha1 = alpha/log n split in lower interval for sharpness instead of equal split (ablation) | lower confidence interval · asymptotic sharpness | first-order width | Improved |
Closest papers by meaning
- Sharp Matrix Empirical Bernstein Inequalities (NeurIPS 2025 · similarity 0.91)
- Split-kl and PAC-Bayes-split-kl Inequalities for Ternary Random Variables (NeurIPS 2022 · similarity 0.88)
- Intrinsic-dimension empirical Bernstein inequalities for bounded self-adjoint operators (NeurIPS 2026 · similarity 0.88)
- Robust Statistical Estimators with Bounded Empirical Sensitivity (NeurIPS 2026 · similarity 0.84)
- Concentration Inequalities for General Functions of Heavy-Tailed Random Variables (ICML 2024 · similarity 0.84)
- On the Asymptotic Distribution of the Minimum Empirical Risk (ICML 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.
