The authors were unhappy about one review, but it was very low-confidence anyway. Except for this one, the reviews were positive, and the author feedback increased their confidence (and at least one score). Reviewers appreciated the rigorous approach to an important problem, establishing finite-sample bounds. One issue pointed out by two reviewers, namely the lack of any basic validation experiments, was addressed in the author feedback.