Pinned post
New preprint w/ (co-first author) @royjamesadams and @suchisaria: "Evaluating Models Robustness Under Dataset Shift" arxiv.org/abs/2010.15100
How can we evaluate *ahead of time* whether or not a model's performance will generalize from training to deployment? 1/
How can we evaluate *ahead of time* whether or not a model's performance will generalize from training to deployment? 1/