Off-Policy Evaluation for Large Action Spaces via Conjunct Effect Modeling

Yuta Saito, Qingyang Ren, Thorsten Joachims
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:29734-29759, 2023.

Abstract

We study off-policy evaluation (OPE) of contextual bandit policies for large discrete action spaces where conventional importance-weighting approaches suffer from excessive variance. To circumvent this variance issue, we propose a new estimator, called OffCEM, that is based on the conjunct effect model (CEM), a novel decomposition of the causal effect into a cluster effect and a residual effect. OffCEM applies importance weighting only to action clusters and addresses the residual causal effect through model-based reward estimation. We show that the proposed estimator is unbiased under a new assumption, called local correctness, which only requires that the residual-effect model preserves the relative expected reward differences of the actions within each cluster. To best leverage the CEM and local correctness, we also propose a new two-step procedure for performing model-based estimation that minimizes bias in the first step and variance in the second step. We find that the resulting OffCEM estimator substantially improves bias and variance compared to a range of conventional estimators. Experiments demonstrate that OffCEM provides substantial improvements in OPE especially in the presence of many actions.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-saito23b, title = {Off-Policy Evaluation for Large Action Spaces via Conjunct Effect Modeling}, author = {Saito, Yuta and Ren, Qingyang and Joachims, Thorsten}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {29734--29759}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/saito23b/saito23b.pdf}, url = {https://proceedings.mlr.press/v202/saito23b.html}, abstract = {We study off-policy evaluation (OPE) of contextual bandit policies for large discrete action spaces where conventional importance-weighting approaches suffer from excessive variance. To circumvent this variance issue, we propose a new estimator, called OffCEM, that is based on the conjunct effect model (CEM), a novel decomposition of the causal effect into a cluster effect and a residual effect. OffCEM applies importance weighting only to action clusters and addresses the residual causal effect through model-based reward estimation. We show that the proposed estimator is unbiased under a new assumption, called local correctness, which only requires that the residual-effect model preserves the relative expected reward differences of the actions within each cluster. To best leverage the CEM and local correctness, we also propose a new two-step procedure for performing model-based estimation that minimizes bias in the first step and variance in the second step. We find that the resulting OffCEM estimator substantially improves bias and variance compared to a range of conventional estimators. Experiments demonstrate that OffCEM provides substantial improvements in OPE especially in the presence of many actions.} }
Endnote
%0 Conference Paper %T Off-Policy Evaluation for Large Action Spaces via Conjunct Effect Modeling %A Yuta Saito %A Qingyang Ren %A Thorsten Joachims %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-saito23b %I PMLR %P 29734--29759 %U https://proceedings.mlr.press/v202/saito23b.html %V 202 %X We study off-policy evaluation (OPE) of contextual bandit policies for large discrete action spaces where conventional importance-weighting approaches suffer from excessive variance. To circumvent this variance issue, we propose a new estimator, called OffCEM, that is based on the conjunct effect model (CEM), a novel decomposition of the causal effect into a cluster effect and a residual effect. OffCEM applies importance weighting only to action clusters and addresses the residual causal effect through model-based reward estimation. We show that the proposed estimator is unbiased under a new assumption, called local correctness, which only requires that the residual-effect model preserves the relative expected reward differences of the actions within each cluster. To best leverage the CEM and local correctness, we also propose a new two-step procedure for performing model-based estimation that minimizes bias in the first step and variance in the second step. We find that the resulting OffCEM estimator substantially improves bias and variance compared to a range of conventional estimators. Experiments demonstrate that OffCEM provides substantial improvements in OPE especially in the presence of many actions.
APA
Saito, Y., Ren, Q. & Joachims, T.. (2023). Off-Policy Evaluation for Large Action Spaces via Conjunct Effect Modeling. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:29734-29759 Available from https://proceedings.mlr.press/v202/saito23b.html.

Related Material