(*) Unless you make it part of it. https://en.wikipedia.org/wiki/The_Futur ... l_Congress During my PhD years I was pretty good at this game!
Derivations (Ti) or references (Si) are generally forgotten. They're considered irrelevant for actual thinking. Insofar extemporization is required, it becomes a Te exercise.
In terms of input (Ne) any new connections just get added on top of existing connections. It's all a big add onto existing structures. This is how metaphors become second nature. Hebbian theory-style. Insofar anything... any structure is detected, I can tell you which other structure is similar based on them having been attached to the same neurons. This explains my weird abuse of proverbs of idioms. Much as I try, I have a really hard time escaping this mode of thinking. I say things like "the waterfall that killed the camel" (that's 2 to 3 idioms combined) because of that.
As far as I'm concerned, there's no topological framework involved. As for PDE's the stuff taught in math class (like diff eq. 201), i.e. heat equations, parabolic, hyperbloc, whatever, ... is useless. Reality is way more complicated than that and I find no worthwhile lessons from ideal cases. Mathematical analysis is mostly dead to Ni. Ni is far more ghetto.
My sampling mode is a combination of statistical and visual. If you want to picture it, think of [PDEs as] a fuzzy cloud of nodes and flows complete with asymptotic behaviors. It's easy to have an [usually correct] opinion of how and where the system goes regardless of how many nodes. From that perspective, there's little difference between Lebesgue integrals over Riemann integrals. And thus I truly struggle with "analysis" making a point of how those two are theoretically different.
So, for Ni, latticework is really a good way of thinking about the world: mapping hundreds and thousands of observations onto a few (10-<100) different structures and declaring that "this is the way of the universe". The arrogance of this practice makes it possible to draw [arrogant] conclusions which are often annoyingly-right. Which is what Ni does.
Ghetto differentiation: Ti, Ne = logic with inspiration; Ni, Te = fuzzy logic with constraints.