EUVIMEDEUVIMED entdecken
Crossref

Reliable Off-Policy Evaluation for Reinforcement Learning

Jie Wang, Rui Gao, Hongyuan Zha

Operations Research · 2024-03 · 72 (2) · 699-716

DOI: 10.1287/opre.2022.2382 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Operations Research, 2024-03. Verfasst von Jie Wang und 2 weiteren Autor:innen. Die Arbeit wurde bisher 10-mal zitiert (Crossref) und nennt 42 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Off-policy evaluation is an important topic in reinforcement learning, which estimates the expected cumulative reward of a target policy using logged trajectory data generated from a different behavior policy, without execution of the target policy. It is imperative to quantify the uncertainty of the off-policy estimate before deployment of the target policy. Here we leverage methodologies from (Wasserstein) distributionally robust optimization to provide robust and optimistic cumulative reward estimates. With proper selection of the size of the distributional uncertainty set, these estimates serve as confidence bounds with nonasymptotic and asymptotic guarantees under stochastic or adversarial environments. We also generalize those results to batch reinforcement learning.

Zitieren

APA

Wang, J., Gao, R., & Zha, H. (2024). Reliable Off-Policy Evaluation for Reinforcement Learning. Operations Research, 72(2), 699-716. https://doi.org/10.1287/opre.2022.2382

Vancouver

Wang J, Gao R, Zha H. Reliable Off-Policy Evaluation for Reinforcement Learning. Operations Research. 2024;72(2):699-716. doi:10.1287/opre.2022.2382

BibTeX
@article{Wang2024,
  author = {Wang, Jie and Gao, Rui and Zha, Hongyuan},
  title = {Reliable Off-Policy Evaluation for Reinforcement Learning},
  journal = {Operations Research},
  year = {2024},
  volume = {72},
  number = {2},
  pages = {699--716},
  doi = {10.1287/opre.2022.2382}
}
Kostenlos starten →