AI Paper Atlas

Research map / Learning theory and algorithms / Sample complexity

Understanding the Under-Coverage Bias in Uncertainty Estimation

Yu Bai, Song Mei, Huan Wang, Caiming Xiong

NeurIPS 2021 · Learning theory and algorithms › Sample complexity · approach: Optimal mean estimation and subsampling · 2 citations (OpenAlex)

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 changedWhereMeasuredEffect
increase model capacity from linear to MLP-3-512quantile regression · Community, alpha=0.9coverage (%): 88.63 -> 63.09Hurt
increase MLP width from 64 to 512MLP quantile regression · STAR, alpha=0.9coverage (%): 83.14 -> 69.71Hurt
increase MLP width from 64 to 512MLP quantile regression · MEPS_19, alpha=0.9coverage (%): 85.46 -> 78.55Hurt
increase MLP width from 64 to 512MLP quantile regression · Bike, alpha=0.9coverage (%): 88.75 -> 87.67Hurt
freeze representation layers of width-512 MLP, train last layer only (ablation)MLP-3-512 · Community, alpha=0.9coverage (%): 63.09 -> 87.85Improved
freeze representation layers of width-512 MLP, train last layer only (ablation)MLP-3-512 · MEPS_21, alpha=0.9coverage (%): 73.09 -> 89.15Improved
use linear model on datasets with small d/nquantile regression · Bike, MEPS_19, MEPS_20, MEPS_21coverage (%): 89.64, 90.09, 90.06, 89.99No clear effect
use linear model on dataset with highest d/n of about 1/16linear quantile regression · Community, alpha=0.9coverage (%): 88.63 vs 90.00Hurt
increase kappa=d/n from 0.01 to 0.5 (fewer training samples)linear quantile regression · MEPS_20 true labelscoverage (%): 89.83 -> 84.75Hurt
increase kappa=d/n from 0.01 to 0.5 (fewer training samples)linear quantile regression · MEPS_20 pseudo-labelscoverage (%): 90.05 -> 83.70Hurt
replace true labels by pseudo-labels from fitted linear modellinear quantile regression · MEPS_20coverage (%): 84.75 vs 83.70 at kappa 0.5No clear effect
increase quantile level alpha at fixed kappaquantile regression · simulated Gaussian linear modelcoverage (%)Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy