State Relevance for Off-Policy Evaluation
Abstract
Importance sampling-based estimators for off-policy evaluation (OPE) are valued for their simplicity, unbiasedness, and reliance on relatively few assumptions. However, the variance of these estimators is often high, especially when trajectories are of different lengths. In this work, we introduce Omitting-States-Irrelevant-to-Return Importance Sampling (OSIRIS), an estimator which reduces variance by strategically omitting likelihood ratios associated with certain states. We formalize the conditions under which OSIRIS is unbiased and has lower variance than ordinary importance sampling, and we demonstrate these properties empirically.
Cite
Text
Shen et al. "State Relevance for Off-Policy Evaluation." International Conference on Machine Learning, 2021.Markdown
[Shen et al. "State Relevance for Off-Policy Evaluation." International Conference on Machine Learning, 2021.](https://mlanthology.org/icml/2021/shen2021icml-state/)BibTeX
@inproceedings{shen2021icml-state,
title = {{State Relevance for Off-Policy Evaluation}},
author = {Shen, Simon P and Ma, Yecheng and Gottesman, Omer and Doshi-Velez, Finale},
booktitle = {International Conference on Machine Learning},
year = {2021},
pages = {9537-9546},
volume = {139},
url = {https://mlanthology.org/icml/2021/shen2021icml-state/}
}