Matej Cief

Jan 19, 2024 | amazon.science | Jacek R. Golebiowski |Philipp Schmidt |Artur Bekasov |Matej Cief

Off-policy evaluation (OPE) methods allow us to compute the expected reward of a policy by using the logged data collected by a different policy. However, when the number of actions is large, or certain actions are under-explored by the logging policy, existing estimators based on inverse-propensity scoring (IPS) can have a high or even infinite variance.