Near-Optimal Confidence Sequences for Bounded Random Variables

Authors: Arun K Kuchibhotla, Qinqing Zheng

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

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct numerical experiments to verify our theoretical claims. Moreover, we apply the Bentkus confidence sequence to the pε, δq mean estimation problem and the best-arm identification problem.
Researcher Affiliation Collaboration 1 Department of Statistics and Data Science, Carnegie Mellon University. 2Facebook AI Research.
Pseudocode Yes Algorithm 1: Adaptive Stopping Algorithm; Algorithm 2: Best Arm Identification
Open Source Code No The paper does not contain any explicit statements or links indicating that source code for the described methodology is publicly available.
Open Datasets No The paper generates synthetic data for its experiments (e.g., "We generate samples Y1, Y2, . . . , Y20000 i.i.d Bernoullip0.1q" and "The data samples are i.i.d generated as Yi m 1 řm j 1 Uij, where Uij are i.i.d uniformly distributed in r0, 1s"). It does not use or provide access to any publicly available external datasets.
Dataset Splits No The paper describes sequential data generation and does not specify traditional training, validation, or test splits. It focuses on sequential stopping criteria.
Hardware Specification No The paper does not provide any specific details about the hardware used to conduct the experiments, such as CPU/GPU models or memory specifications.
Software Dependencies No The paper does not specify any software dependencies with version numbers.
Experiment Setup Yes We use δ 0.05 for all the experiments. For A-Bentkus, we fix the spacing parameter η 1.1, the stitching function hpkq pk 1q1.1ζp1.1q, and δ1 2δ{3, δ2 δ{3.