Off-Policy Evaluation of Slate Policies under Bayes Risk
This work provides a more accurate off-policy evaluation method for slate bandit problems, which is important for researchers and practitioners in recommender systems and online advertising to reliably assess new policies without costly A/B tests.
This paper addresses off-policy evaluation for slate bandits where the logging policy factorizes over slots. The authors introduce a new 'additive' estimator that, under Bayes risk, is guaranteed to have lower risk than the pseudoinverse (PI) estimator, with risk improvement growing linearly with the number of slots and the gap between arithmetic and harmonic means of slot-level divergences.
We study the problem of off-policy evaluation for slate bandits, for the typical case in which the logging policy factorizes over the slots of the slate. We slightly depart from the existing literature by taking Bayes risk as the criterion by which to evaluate estimators, and we analyze the family of 'additive' estimators that includes the pseudoinverse (PI) estimator of Swaminathan et al.\ (2017; arXiv:1605.04812). Using a control variate approach, we identify a new estimator in this family that is guaranteed to have lower risk than PI in the above class of problems. In particular, we show that the risk improvement over PI grows linearly with the number of slots, and linearly with the gap between the arithmetic and the harmonic mean of a set of slot-level divergences between the logging and the target policy. In the typical case of a uniform logging policy and a deterministic target policy, each divergence corresponds to slot size, showing that maximal gains can be obtained for slate problems with diverse numbers of actions per slot.