Optimizing Performative Learning
The Data Transitions seminars aim to make cutting-edge scientific research in the digital and data fields accessible to all. They are open to anyone with a curious mind, whether junior or senior, expert or novice.
The next seminar, "Optimizing Performative Learning" will be led by Edwige Cyffers (CNRS, Université Paris-Dauphine-PSL), on Thursday June 18th - 11:45 am (salle F107, Bâtiment Coriolis).
Performative learning (the model affects the data distribution it is meant to predict): introduction of this setting and presentation of two results on how to optimize performance in the long term, rather than only at a single time step.