Review for NeurIPS paper: Robust Federated Learning: The Case of Affine Distribution Shifts

Clarity: The paper is generally well-written. The authors do a very good job of discussing federated learning, robust optimization, and the interplay between the two. They also spend a lot of time in helping the reader understand the exact robust optimization setting being considered, which can be immensely helpful to the layman trying to understand the paper. The authors also do a good job of discussing many separate important theoretical areas, including minimax optimization, generalization, and distributional robustness. However, I wish that there was a bit more of a coherent flow as to why all three of these aspects are considered.