@pinkddle Geometric Deep learning in the usual formulation might not be super relevant to everyone, as in not everyone deals with data admitting group symmetries, but structurally equivariant architectures seem incredibly important for all sorts of cases
@taz_chu Some people manage, but... they're pretty rare
@Simmy16242170 is routinely making the rest of us look bad in that regard ππ she's too good
@ToKTeacher Yes, and they are just wrong.
Many animal species, some of which are almost as smart as humans, don't have language.
Yet they have thoughts.
@msciguy In the end, I could however write all my findings without the cat theory. But I abhor non-constructive mathematics, the chain of thought is much clearer with the category theoretic formalism
@msciguy Diagrams might seem silly as a reason by itself, but in practice it's been incredibly helpful for me to visually organize everything. Much much easier than walls of text and equations. Point (2) has actually lead me to discover structure in my problem I would not have otherwise