Doubly Robust Policy Evaluation and Optimization

Miroslav Dudík, Dumitru Erhan, John Langford +1

We study sequential decision making in environments where rewards are only partially observed, but can be modeled as a function of observed contexts and the chosen action by the de…