You're reading the public-facing archive of the Category Theory Zulip server.
To join the server you need an invite. Anybody can get an invite by contacting Matteo Capucci at name dot surname at gmail dot com.
For all things related to this archive refer to the same person.
I learned about the concept of "homomorphism" first in the setting of groups. The idea is that they are maps that "preserve the structure". For example, if is a map between additive groups, then for to be a homomorphism of additive groups we require for all group elements . This condition on preservation of structure is very similar to what we require for functors, to my understanding.
However, I've also learned that continuous maps are the homomorphisms for topological spaces. I was trying to see how we can express that a function is continuous using a similar "structure preserving" requirement as for group homomorphisms.
Let be a sequence that converges to .
Then, a function is continuous at if , I think.
Now define a function that takes any convergent sequence and returns its limit. Letting be the set of real-valued convergent sequences, then .
Then . Under this notation, we have that is continuous at if .
This is similar to the structure-preserving requirement for groups; if we view as a function that takes two inputs, the structure preserving requirement becomes: . It seems like, from this perspective, a continuous functions really is "structure-preserving", where the structure is created by the operation of taking limits. If this is right, it seems cool!
As a sort of follow-up - Is there some way to view continuous functions between topological spaces as functors, from this perspective?
Edit: This was also making me wonder if one could try and define a topological space as a category where you can compute limits of sequences. This led me to discovering this interesting page: https://ncatlab.org/nlab/show/convergence+space . Neat!
"Homomorphism" and "homeomorphism" are different words...
Zhen Lin Low said:
"Homomorphism" and "homeomorphism" are different words...
Good point, thanks. I will fix it.
So your definition doesn't work for all topological spaces. The ones where it does work are https://ncatlab.org/nlab/show/subsequential+space
Max New said:
So your definition doesn't work for all topological spaces. The ones where it does work are https://ncatlab.org/nlab/show/subsequential+space
It would be great if you could be a little more specific (which definition? doesn't work how?). I don't really know much topology, and was basically just thinking about real numbers when I wrote this out. But thanks for the interesting link!
There are lots of topological spaces where the topology is not determined by the collection of sequences that converge.
This is why people invented nets... and I think the link here will help explain the issues.
When I teach analysis, I list 3 facts about sequences that are true for metric spaces but not general topological spaces. Later I show how all these facts can be "saved" if we use nets instead of sequences.
The link mentions just one of these facts, under "Motivation". That's the one were just talking about:
1) a function between metric spaces is continuous iff it maps convergent sequences to convergent sequences.
Here are the two other facts: for a metric space
2) a subset U is open iff given x U and a sequence converging to x, the sequence is eventually in U;
3) a subset K is compact iff every sequence in K has a convergent subsequence.
It's good to know how all three of these facts fail for general topological spaces, and how to save them using nets.
Ok, that makes a lot of sense! So in a metric space we have continuous functions interacting nicely with convergent seequences, but that's not the case in every topological space. And apparently this niceness can be recovered by using "nets", which are functions from directed sets (instead of from just the natural numbers).
That explains why my idea above won't work for all topological spaces, at least when working with sequences.
Thanks!
Right. Basically the problem is that for really "big" topological spaces, sequences just aren't long enough. They're infinitely long... but that's only the smallest infinity!
Right, our sequences only have countably many elements. But one could consider a map from the ordered real numbers, for example, to get an image with many more points (uncountably many).
Also, I seem to recall that directed sets can have more complicated shapes - some kinds of "branching", I think? I don't know if that is important to the "niceness" of nets.
Right. We need to let nets be longer than sequences... but while we're at it, we might as well let them be "wider".
That is, we index a net by a directed set, and a directed set doesn't need to be linearly ordered.
That lets us do tricks like this. Take as our directed set the collection of all open sets containing a point , ordered by reverse inclusion so iff . Then we can take a net indexed by this directed set!
This allows really slick proofs.
That's a really cool example! It sounds like that would let you talk about approaching a point from "all sides at once", or without committing to a single path of approach.
Exactly! In metric spaces we often say stuff like "choose a sequence such that ".
But with nets we can say "choose a net such that ", where is any open set containing .
So once you get used to them they are very easy to work with.
The point is that open neighborhoods, rather than distances, are how we measure "closeness" in a topological space.