Deeply-debiased off-policy interval estimation

Shi, Chengchun ORCID: 0000-0001-7773-2099, Wan, Runzhe, Chernozhukov, Victor and Song, Rui (2021) Deeply-debiased off-policy interval estimation. In: International Conference on Machine Learning, 2021-07-18 - 2021-07-24, Online. (In Press)

Text (Shi_deeply-biased-off-policy-interval-estimation--accepted) - Accepted Version Download (549kB)

• Author

Abstract

Off-policy evaluation learns a target policy’s value with a historical dataset generated by a different behavior policy. In addition to a point estimate, many applications would benefit significantly from having a confidence interval (CI) that quantifies the uncertainty of the point estimate. In this paper, we propose a novel deeply-debiasing procedure to construct an efficient, robust, and flexible CI on a target policy’s value. Our method is justified by theoretical results and numerical experiments. A Python implementation of the proposed procedure is available at https://github.com/RunzheStat/D2OPE.

Item Type:	Conference or Workshop Item (Paper)
Official URL:	https://icml.cc/
Additional Information:	© 2021 The Authors
Divisions:	Statistics
Subjects:	H Social Sciences > HA Statistics Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Date Deposited:	24 Jun 2021 08:03
Last Modified:	11 Sep 2025 00:06
URI:	http://eprints.lse.ac.uk/id/eprint/110920

Actions (login required)

View Item