Sigmoid-FTRL: Design-Based Adaptive Neyman Allocation for AIPW Estimators

Fangyi Chen, Shu Ge, Jian Qian, Christopher Harshaw · 2025

Read Sigmoid-FTRL: Design-Based Adaptive Neyman Allocation for AIPW Estimators by Fangyi Chen, Shu Ge, Jian Qian, Christopher Harshaw (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-19905.txt.

Introduction

Fangyi Chen, Shu Ge, Jian Qian, and Christopher Harshaw design adaptive experiments in a design-based statistics setting. Outcomes and covariates are treated as fixed, and each arriving subject needs a treatment chance and a pair of predictors for an augmented inverse-probability estimator. The goal is to close the gap to an oracle that knew the best fixed choices in advance.

Their Sigmoid-FTRL method handles a non-convex problem by minimizing two convex regrets at once. They prove its Neyman regret shrinks at a rate that no adaptive design can beat under their conditions, so the method is minimax optimal. A central limit theorem and a safe variance estimator also let users build valid confidence intervals from the adaptive data.

Catalog record

CREATORFangyi Chen, Shu Ge, Jian Qian, Christopher Harshaw
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE599 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-19905

Browse the full library · About the Tower of Babel