@PreetumNakkiran I agree. I will read it in detail. You should also check the domain adaptation theory literature as they might answer some of the conjectures. For example, this seminal one connects generalization to distributional divergences: https://t.c
@zacharylipton @mohitsharmaj29 @prfsanjeevarora It's unclear how much carries over from traditional learning theory—e.g., measures of model complexity? https://t.co/PBzwJhI6Ib but the theory of some fields hinges upon these ideas—for ex, domain adaptatio
そういえば今年のACMLのkeynote speakerの1人がBen-Davidさん（https://t.co/egnkuFYFjt の人）と昨日気づいてちょっと行けば良かったと後悔している