Unified PAC-Bayesian Study of Pessimism for Offline Policy Learning with Regularized Importance Sampling

Off-policy learning (OPL) often involves minimizing a risk estimator based on importance weighting to correct bias from the logging policy used to collect data. However, this method can produce an estimator with a high variance. A common solution is to regularize the importance weights and learn the...

Ausführliche Beschreibung

Bibliographische Detailangaben

Veröffentlicht in:	arXiv.org 2024-06
Hauptverfasser:	Aouali, Imad, Victor-Emmanuel Brunel, Rohde, David, Korba, Anna
Format:	Artikel
Sprache:	eng
Schlagworte:	Bayesian analysis Importance sampling Regularization Weighting
Online-Zugang:	Volltext
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!