NeurIPS 2020

Characterizing Optimal Mixed Policies: Where to Intervene and What to Observe

Meta Review

This paper consider a structural bandit problem where we wish to define a subset of variables to observe and intervene on, such that optimal performance can still be achieved. While the reviewers had some difficulty with some terms in the paper (such as 'simplicity,') this was to a large extent clarified in the rebuttal. The reviewers found this paper to be a nice addition to the growing 'structural bandits' literature.