Research map / Learning theory and algorithms / Sample complexity
V-statistics and Variance Estimation
Zhengze Zhou, Lucas Mentch, Giles Hooker
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 bias correction to the variance estimator (corrected-V instead of IJ) (ablation) | random forest variance estimation · MARS, p1, k=100, B=500 | variance ratio (ideal 1): 7.6985 -> 1.4962 | Improved |
| add bias correction to the variance estimator (ablation) | random forest variance estimation · MARS, p1, k=100, B=5000 | variance ratio (ideal 1): 1.7073 -> 1.0305 | Improved |
| add bias correction to the variance estimator (ablation) | random forest variance estimation · MARS, p1, k=100, B=2500 | 95% CI coverage (%): 99.2 -> 94.2 | Improved |
| add bias correction to the variance estimator (ablation) | random forest variance estimation · MARS, p1, k=500, B=500 | variance ratio (ideal 1): 10.7237 -> 4.5438 | Improved |
| increase number of trees B from 500 to 5000 | IJ original variance estimate · MARS, p1, k=100 | variance ratio (ideal 1): 7.6985 -> 1.7073 | Improved |
| increase subsample size k from 100 to 500 | corrected-V variance estimate · MARS, p1, B=500 | variance ratio (ideal 1): 1.4962 -> 4.5438 | Hurt |
| sample with replacement (V) instead of without (U), full subsample size | random forest · diabetes | RMSE: 58.7494 -> 55.4130 | Improved |
| sample with replacement (V) instead of without (U), full subsample size | random forest · boston | RMSE: 3.0588 -> 3.2944 | Hurt |
| reduce subsample proportion from 1 to 0.2 (U-statistics) | random forest · diabetes | RMSE: 58.7494 -> 54.6137 | Improved |
| reduce subsample proportion from 1 to 0.2 (U-statistics) | random forest · boston | RMSE: 3.0588 -> 3.9696 | Hurt |
| use Balanced Variance Estimation instead of Internal Variance Estimation | variance estimator · simulation, n=500, k=100 | variance bias | Improved |
| apply bias correction to U-statistics ensembles | corrected estimator · sampling without replacement | variance estimate | Hurt |
Closest papers by meaning
- Fluctuations, Bias, Variance & Ensemble of Learners: Exact Asymptotics for Convex Losses in High-Dimension (ICML 2022 · similarity 0.85)
- Learning Theory for Distribution Regression (JMLR 2016 · similarity 0.84)
- Scaling-up Empirical Risk Minimization: Optimization of Incomplete $U$-statistics (JMLR 2016 · similarity 0.84)
- Asymptotic Anytime-Valid Inference for U-statistics (NeurIPS 2026 · similarity 0.83)
- On the Consistency of Kernel Methods with Dependent Observations (ICML 2024 · similarity 0.83)
- On the Asymptotic Distribution of the Minimum Empirical Risk (ICML 2024 · similarity 0.83)
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.
