Does anti-aliasing smooth spatial loss surface?

The last two posts (1, 2) I published were on the topic of spatial robustness, i.e. the stability of a neural network prediction with respect to small translation and rotation. A curious observation from [1] states that first order adversary seem to ...