Talk: training in performative setting

Today I talked about learning when data gradually reacts to the model deployed, the goal is to induce a favorable data distribution and a model that performs well on it. My slides are here.