Talk
Optimal and Adaptive Off-policy Evaluation in Contextual Bandits
Yu-Xiang Wang · Alekh Agarwal · Miroslav Dudik

Tue Aug 8th 11:06 -- 11:24 AM @ C4.5

We study the off-policy evaluation problem---estimating the value of a target policy using data collected by another policy---under the contextual bandit model. We consider the general (agnostic) setting without access to a consistent model of rewards and establish a minimax lower bound on the mean squared error (MSE). The bound is matched up to constants by the inverse propensity scoring (IPS) and doubly robust (DR) estimators. This highlights the difficulty of the agnostic contextual setting, in contrast with multi-armed bandits and contextual bandits with access to a consistent reward model, where IPS is suboptimal. We then propose the SWITCH estimator, which can use an existing reward model (not necessarily consistent) to achieve a better bias-variance tradeoff than IPS and DR. We prove an upper bound on its MSE and demonstrate its benefits empirically on a diverse collection of datasets, often outperforming prior work by orders of magnitude.

Author Information

Yu-Xiang Wang (Carnegie Mellon University / Amazon AWS)
Alekh Agarwal (Microsoft Research)
Miroslav Dudik (Microsoft Research)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors