Understanding the Under-Coverage Bias in Uncertainty Estimation

Authors: Yu Bai, Song Mei, Huan Wang, Caiming Xiong

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on simulated and real data verify our theory and further illustrate the effect of various factors such as sample size and model capacity on the under-coverage bias in more practical setups.
Researcher Affiliation Collaboration Yu Bai Salesforce Research yu.bai@salesforce.com Song Mei UC Berkeley songmei@berkeley.edu Huan Wang Salesforce Research huan.wang@salesforce.com Caiming Xiong Salesforce Research cxiong@salesforce.com
Pseudocode No The paper does not contain any clearly labeled pseudocode or algorithm blocks.
Open Source Code No The paper does not include any explicit statements about releasing source code or provide links to a code repository for the methodology described.
Open Datasets Yes We take six real-world regression datasets: community and crimes (Community) [2], bike sharing (Bike) [1], Tennessee s student teacher achievement ratio (STAR) [6], as well as the medical expenditure survey number 19 (MEPS_19) [3], number 20 (MEPS_20) [4], and number 21 (MEPS_21) [5].
Dataset Splits Yes For each setting, we average over 8 random seeds where each seed determines the train-validation split, model initialization, and SGD batching.
Hardware Specification No The paper does not specify any particular hardware (e.g., GPU, CPU models, or memory) used for running the experiments.
Software Dependencies No The paper mentions using momentum SGD, but does not provide specific software names with version numbers (e.g., Python 3.x, TensorFlow 2.x, PyTorch 1.x) or specific solver versions.
Experiment Setup Yes We minimize the α-quantile loss (3) via momentum SGD with batch size 64.