PUID: A Personalized Deconfounding Framework for Recommender Systems under Hidden Confounding

Zongyu Li

Recommender systems often rely on observational user-item interaction data, which is prone to selection bias due to users' selective interactions with items. While techniques such…