Exponential Smoothing for Off-Policy Learning
This work addresses the problem of enhancing policy learning efficiency and reliability for researchers and practitioners in reinforcement learning and bandit settings, offering an incremental improvement with theoretical insights into regularization effects.
The paper tackles the problem of improving off-policy learning from logged bandit data by introducing a smooth regularization for the inverse propensity scoring estimator, deriving a tractable PAC-Bayes generalization bound that provides learning certificates and holds even without bounded importance weights. The results demonstrate favorable performance in learning tasks and challenge the belief that clipped IPS is always better, identifying cases where regularization might not be needed.
Off-policy learning (OPL) aims at finding improved policies from logged bandit data, often by minimizing the inverse propensity scoring (IPS) estimator of the risk. In this work, we investigate a smooth regularization for IPS, for which we derive a two-sided PAC-Bayes generalization bound. The bound is tractable, scalable, interpretable and provides learning certificates. In particular, it is also valid for standard IPS without making the assumption that the importance weights are bounded. We demonstrate the relevance of our approach and its favorable performance through a set of learning tasks. Since our bound holds for standard IPS, we are able to provide insight into when regularizing IPS is useful. Namely, we identify cases where regularization might not be needed. This goes against the belief that, in practice, clipped IPS often enjoys favorable performance than standard IPS in OPL.