Optimal Regularization for Performative Learning

ArXi:2510.12249v2 Announce Type: replace In performative learning, the data distribution reacts to the deployed model - for example, because strategic users adapt their features to game it - which creates a complex dynamic than in classical supervised learning. One should thus not only optimize the model for the current data but also take into account that the model might steer the distribution in a new direction, without knowing the exact nature of the potential shift.