Behaviour Policy Estimation in Off-Policy Policy Evaluation: Calibration Matters
Abstract
In this work, we consider the problem of estimating a behaviour policy for use in Off-Policy Policy Evaluation (OPE) when the true behaviour policy is unknown. Via a series of empirical studies, we demonstrate how accurate OPE is strongly dependent on the calibration of estimated behaviour policy models: how precisely the behaviour policy is estimated from data. We show how powerful parametric models such as neural networks can result in highly uncalibrated behaviour policy models on a real-world medical dataset, and illustrate how a simple, non-parametric, k-nearest neighbours model produces better calibrated behaviour policy estimates and can be used to obtain superior importance sampling-based OPE estimates.
Keywords
Cite
@article{arxiv.1807.01066,
title = {Behaviour Policy Estimation in Off-Policy Policy Evaluation: Calibration Matters},
author = {Aniruddh Raghu and Omer Gottesman and Yao Liu and Matthieu Komorowski and Aldo Faisal and Finale Doshi-Velez and Emma Brunskill},
journal= {arXiv preprint arXiv:1807.01066},
year = {2018}
}
Comments
Accepted to workshop on Machine Learning for Causal Inference, Counterfactual Prediction, and Autonomous Action at ICML 2018