aboutlogic: premises #11 | Category Theory: Algebra on Steroids?

Category Theory: Algebra on Steroids?

Show notes

Become an Omega or Aleph Member on Youtube and help us keep this going! 🙏

Or you can buy us a coffee here: https://buymeacoffee.com/aboutlogic

Further Reading & Resources: Get the HoTT Book for free (no advertisement): https://homotopytypetheory.org/book/ Thorsten Altenkirch: http://www.cs.nott.ac.uk/~psztxa/ Deniz Sarikaya: https://www.denizsarikaya.de/ Creative Production: Jan-Niklas Meyer: http://www.jammos.com/

Join the Discussion: Have questions or thoughts to share? Drop a comment below and engage in a discussion with fellow viewers and experts.

Show transcript

00:00:00: One of my slogans is to say a category theory can either learn it or you reinvented.

00:00:31: Visit or revisit this topic.

00:00:34: And I wanted to ask you, Thorsten what's all the fuss about this error diagram chasing and What's your perspective?

00:00:41: Is it just algebra on steroids Or what's category theory?

00:00:46: yes um i've given courses in the middle school about categories but actually a bit uh not really hardcore category theoretician.

00:01:01: But I think it's useful to know good about it.

00:01:06: And especially for somebody like me doing type theory, It is quite natural.

00:01:17: If you do this kind of thing then categories show up naturally actually even in programming... Actually everywhere!

00:01:26: One.

00:01:26: my slogan was A category theory can either learn it or you reinvented.

00:01:34: So that's as I've seen people who don't know category theory and then they ended up to have something which turns out to be very standard categorical notion, so sometimes is good to re-invent things You understand some better?

00:01:51: The main point about categories There are a way to do structural mathematics.

00:02:02: And what we mean by structural mathematics, it means that really... We do things in some more abstract ways instead of looking at concrete encodings and representations.

00:02:18: It proves in a more abstract way.

00:02:30: That's what I call structural mathematics.

00:02:35: Maybe let me start already there with some live questions.

00:02:40: So, would you say going from calculus to metric spaces?

00:02:45: To topological spaces it is more abstract.

00:02:50: You might find the reason isn't really the intricacies about the reals but just Stuff that is close to each other stays close, so topological reasoning suffices.

00:03:02: This would be already category theory for you?

00:03:05: No.

00:03:07: Category theory is a way to implement structural mathematics and to have the structure on which we can climb higher levels of abstraction.

00:03:18: Let me give them different concrete examples.

00:03:20: The last time we talked about arithmetic, and let's think about natural numbers.

00:03:28: And then you have the operation of addition... ...and they satisfy some laws that we can prove as the whole for example law of assertivity and zero is neutral if it proves this by induction.

00:03:43: That very concrete right?

00:03:46: So he can prove its a monoid.

00:03:48: We discover that multiplication is also monoid and the two together form a same ring.

00:03:55: And so on, we use algebra now to abstract from this very concrete case of the natural numbers... ...and say things more abstractly which are nice because they're more general.

00:04:13: we prove things not just for the actual numbers, but maybe proving things which hold in all semi-rings.

00:04:19: And it seems to be quite useful to do this step of abstraction.

00:04:27: But you don't need to stop there.

00:04:31: Once we have algebra We can look on how different algebraic structures are related two semi-rings and they are related in some way, so there's a homomorphism function which preserves the structure.

00:04:52: And we can look at different algebraic theories how the category of homomorphisms behave right?

00:05:06: This is that kind to climb up in the level of abstraction.

00:05:14: And now talk about algebraic theories and how they, different models of algebras are related.

00:05:24: but we can go further.

00:05:27: so okay no we have categories with some properties leading to higher categories, than the categories of categories.

00:05:43: So that's maybe I would say is a beauty category theory.

00:05:48: instead it supports abstraction and supports going up these levels And often there are some new mathematics discovered or taking place at this high level which interesting on its own actually.

00:06:09: But okay, let me maybe go back to just normal categories.

00:06:13: So what's the power of category theory?

00:06:19: Before we go into higher categories... Let us look at what is called one-categories or normal categories.

00:06:28: Then I think there are two things which come in my mind.

00:06:33: One is universal properties and another is duality some of the key features off of category theory.

00:06:42: And yeah, let me just

00:06:44: think we encountered this in group theory.

00:06:49: if you have your linear algebra course?

00:06:51: This is something You've seen.

00:06:55: it's with this universal property.

00:06:57: You'll just draw the two lines and your diagram until say It doesn't matter whether you go around these moffisms or or the

00:07:10: universe of property is a unicity.

00:07:14: So, you need an unicity feature right?

00:07:21: You say okay what does a product have two sets?

00:07:26: let's think about that.

00:07:27: I find obvious.

00:07:28: when we start doing categories theory it always good to first thing about the category of sets and then move on.

00:07:35: so in the category I mean, the product of two sets has projections from the product into A and to B. And what's a universal property?

00:07:48: is that for any set which looks like projection you find now it's important unique morphism in this product object which satisfies the property.

00:08:12: And by describing these concepts, as you say I mean products not only in the category of sets but also in this category groups and so on.

00:08:22: they are defined by universal properties .

00:08:27: A nice consequence is that any two objects satisfying universal properties.

00:08:35: So this is the way.

00:08:42: by defining properties, by defining objects by universal properties you get this uniqueness in a way.

00:08:53: You only have one thing which satisfies us and so that's just idea of universal properties which are very useful.

00:09:08: So because the category, you can turn all of these arrows around.

00:09:12: And if we do this... You get a dual-category and now look at every concept in the Dual Category.

00:09:24: What are products?

00:09:26: set op?

00:09:28: Set Op is the category of sets where all functions go backwards.

00:09:33: A function from A to B is an arrow from P to A. And what's the product in this setup?

00:09:39: Now there is a language already, it's called co-product.

00:09:43: So Co means that its same construction but in the dual category.

00:09:51: and what's a co-product?

00:09:53: in the category of sets as usually call disjoint union you can see two for one.

00:09:59: I mean we find products.

00:10:01: And

00:10:05: is it always true that going back and forth works?

00:10:10: Yes, because this is very simple.

00:10:11: You just turn things around.

00:10:13: so I mean... It's just a mirror!

00:10:15: The mirror of the mirror ends up where you started.

00:10:19: So we can also do with other concepts like for example quite interesting sometimes to discover something new.

00:10:26: So natural numbers in initial algebra The natural numbers.

00:10:34: algebra is a set with some zero and some successor.

00:10:39: And if you do this categorically, then the national number are called initial object in this category of algebras or at least one.

00:10:53: So into any algebra there's a unique homomorphism but and ask for the opposite.

00:11:03: So what is a terminal algebra in the natural numbers?

00:11:13: And this is what we call co-induction, so a co- inductive type.

00:11:18: So national numbers are an inductive types... ...and now you apply the mirror to get new concepts which are the co- Inductive Numbers.

00:11:28: That's quite fun!

00:11:29: You can compute with them.

00:11:30: They have the co-inductive numbers, they have an infinite element.

00:11:36: There are subinfinite elements and you can... ...they correspond in a way to viral loops.

00:11:45: so you have non-terminating viral loops And their corresponding number of iterations are represented by coincidental numbers.

00:11:54: You can do all sorts of arithmetic or you can define addition and multiplication for cool natural numbers and it's quite fun.

00:12:06: So I choose this in my, actually my type theory class.

00:12:10: i always played around with cool natural number.

00:12:14: And think we featured them shortly on the episode of Mike

00:12:18: right there?

00:12:19: We talked about that

00:12:23: to get the algorithm running people revisiting episodes.

00:12:26: so you got watch times.

00:12:28: Yeah,

00:12:30: maybe we should first watch

00:12:31: this episode.

00:12:34: Why do you think so?

00:12:40: But now I wanted to move to a different point namely usually when we think about categories or the way to thing about categories and then start studying them is in a way algebraics, generalized algebraic theories.

00:13:02: So if you know monoreeds like addition and zero-in-one categories are generalized version of monoreads where we don't have one carrier but the whole family of carriers which has amorphisms operation which is composition and then a neutral element of identity, but they don't live all in the same carrier.

00:13:29: They're living different carriers.

00:13:32: you can say okay actually categories are sort of monoids on steroids or monoids I call them sometimes because there's this notion that group would eat.

00:13:51: And so if you forget about the group and go on only back to the monoid, then you have set a category is actually a monoidoid.

00:14:00: So that's one way two view categories.

00:14:03: This is very algebraic point of view.

00:14:07: but there is another which we already mentioned this diagram chasing here that categories are often explained or used by by drawing diagrams or commuting diagrams.

00:14:21: So maybe just a remark before we switch to this perspective?

00:14:25: I mean, in some sense there is huge tradition of mathematics and ancient times it's about the integers.

00:14:38: but then we realized that was always about rings.

00:14:41: everything i really want know isn't the ring structure.

00:14:48: It was always about the group, and maybe this is where at least in linear algebra we stop.

00:14:53: We say learn groups and you get everything from there.

00:14:57: but category theory went further.

00:14:59: You are mentioning monoids which may be in some linear algebra classes mentioned sometimes don't even mention them or go beyond that.

00:15:09: Historically speaking it's a super power of Grotendieg right?

00:15:13: That what he was famous for really analyzing what went into the proof and it was never the whole structure.

00:15:21: It's a little less than everything that was there, really naming the property that is essential rather then all the bloatware around and reducing to that... ...is huge milestone!

00:15:35: And also sometimes why doesn't look so impressive?

00:15:39: because afterwards its trivial so-to speak you just distilled exactly when needed.

00:15:46: So your theorem often looks like, yeah this just follows.

00:15:49: But to get these notions is a very constructive act and I mean there's the scrutiny quote i think There was paper by Friedmann about it that By solving The problem Is not really going into the details of the problem but raising the sea.

00:16:07: Yes

00:16:09: And Really Get the conceptual apparatus drops out for free, so to speak.

00:16:14: Because you

00:16:15: just... It's interesting that we manage linear algebra because linear algebra can be also reconstructed using string diagrams and this is actually a very categorical notion of monoidal categories with some extra structure And it's really nice!

00:16:42: completely different angle, and it's very useful to do this quantum theory.

00:16:52: Yeah!

00:16:52: This was exactly why I mentioned that not because it wasn't my first math lecture with where things started from there.

00:17:00: Okay but i interrupted you... You wanted

00:17:03: to talk about diagrams so actually intuitive or the nice way to write equations, but actually there's a bit more.

00:17:17: Namely that if you draw diagram so have three levels of zero-dimensional objects which are the points and then you have morphisms with one dimensional object And then we have commuting diagrams.

00:17:34: You say okay I wrote this diagram let's do it.

00:17:37: It means you can view it as a field triangle, not just the boundary of a triangle but there's a filled triangle.

00:17:47: And this is how we actually derive or explain laws categories in purely geometric way.

00:18:00: and nice thing that works... which actually works for higher categories.

00:18:11: So let me first tell you why would you be interested in higher categories, I mentioned already.

00:18:16: Categories of categories.

00:18:17: so just category of all categories.

00:18:23: now that is a bit often an argument here.

00:18:26: So classically the category of categories are just a category but if hot, then you stop accepting this.

00:18:39: So a hot homotopythal theory gives you lots of things in particular for it's very well with categorical theory It is almost unroydable to do both together I find.

00:18:53: but In HOT these universal properties are actually unique solutions.

00:19:00: in classical category theory, you see products are unique up to isomorphisms.

00:19:05: We always mumble this unique up-to-isomorphism.

00:19:09: that's like oh money part my home or something.

00:19:12: it's like we have this mantra up to Isomorphism.

00:19:15: and when you learn hot?

00:19:17: You realize that you can drop this.

00:19:19: You can say yeah, products a unit!

00:19:20: That's the consequence of univalence.

00:19:24: so uh...that's quite cool.

00:19:26: but then you realise hang on Here is what seems to be a problem, but it's actually feature.

00:19:35: The category of categories... This is actually a category.

00:19:39: Now yeah this is the problem because categories have what we call home sets.

00:19:45: so there are simple equality on morphisms and in the category two objects.

00:19:57: So equality of objects is actually isomorphism.

00:20:01: And yeah, it's not a simple property because you have the choice of isomorphisms.

00:20:14: so what was happening?

00:20:17: Is that the category of categories turns out to be no category anymore but higher categories where we had identify these two one-category or and so on.

00:20:28: And the question is now how can you define these higher categories?

00:20:32: and if your start with your algebra toolkit, I mean... You start writing down the laws of a two-category.

00:20:42: Then say okay let's go on to three category and it gets more complicated.

00:20:51: basically there are so much boilerplate!

00:20:56: But it turns out that the geometric view works quite nicely.

00:21:02: And what we say is, in a geometric view... We think of higher categories as sort of higher topological spaces.

00:21:12: so we have points, we have lines and triangles.

00:21:15: then at next level you can put diagrams together.

00:21:24: there is a commuting cube or tetrahedra of diagrams.

00:21:36: So you have then like higher equalities, the qualities between diagrams and then you can move on an go up all dimensions And it turns out that this view of categories which comes from homotropic theory From simple sets, that works for all dimensions.

00:21:58: So you can easily write down what do we want?

00:22:01: I mean first of all you have... You compose triangles and compose two lines to get a triangle And the third line is filled with this triangle.

00:22:16: Then when you move up then maybe okay get a tetrahedron and one side missing.

00:22:21: So this is also called a horn, or higher horn.

00:22:24: And you get unique filler for these and so on.

00:22:28: You can do it in all dimensions with an analytic notion of what higher categories are?

00:22:37: I think we shortly talked about this... I don't want premises to end with Ulrich because there's then the simplicial view at the cubicle models where those objects are slightly different

00:22:53: How you glue them together

00:22:54: for higher?

00:22:55: There's actually a slight complication, which I'm just looking at or it's not a complication.

00:23:01: It is another feature that the classical theory of higher categories works with simplicial sets and we work with semi-simplical types.

00:23:14: so this long story.

00:23:16: but In a way, the classical view makes a difference between the identities and composition.

00:23:24: But in a way you could say identity is just an empty composition.

00:23:28: so why it's that different?

00:23:29: And if your golden hot then this doesn't actually work!

00:23:34: This hot stops from this asymmetry of composition and identity... It forces to think about something longer And you end up with the notion which is in a way conceptually better because we make both identity and composition are sort of weak, have weak properties that not strictly identified.

00:24:02: That seems to be more natural for me but it's actually also your only thing which works hot.

00:24:11: Is there any good moment about this?

00:24:15: switch in perspective between types and categories because the theories are so related with each other.

00:24:20: or do you think this needs another?

00:24:21: Yeah,

00:24:25: that's a good question answer.

00:24:28: So some people things at.

00:24:29: category theory is sort of alternative.

00:24:32: to talk theory alternative foundation.

00:24:35: right

00:24:37: I Think twenty five years ago This was the leading narrative.

00:24:41: right.

00:24:41: there's really its own foundation.

00:24:44: taking algebra series?

00:24:46: Yes, but I mean okay.

00:24:49: And I personally don't agree with this because i think that category theory is just gives you a structure as it gives your way to structure knowledge But he doesn't tell you in the video what was right!

00:25:02: This sort of like thing also has this beautiful idea of a symmetric Cartesian closed category, which is Cartesian-closed and this dual is also Cartesian Closed.

00:25:18: It's beautiful but it actually equatially inconsistent.

00:25:22: so how do you know that This not useful construction looks very nice from the point of Category Theory?

00:25:29: So I would say You need some other sort more intuitive foundations Of constructions where you can explain.

00:25:40: And to me, this is type theory.

00:25:42: Type theories are really based on it's basically functional programming and since we can understand... We have good intuition for programing so that we can extend to types theory.

00:26:03: So I think you could easily develop a very natural intuition for what constructions are possible in talk theory.

00:26:14: Whereas category theory, I don't see this.

00:26:18: It's a good way to organize your thoughts but it is not the source of you ideas and at least...I do think that its' good for us!

00:26:30: But two together they're sort-of perfect.

00:26:34: as So you use the type theory as an intuitive foundation, and then you structure your developments using category theory.

00:26:48: And very easily.

00:26:49: I mean this thing is like category theory it's a bit like heroin addiction.

00:26:56: You always want more.

00:26:58: so start with normal categories but realize that these are not good enough that you really need to go higher categories and so on.

00:27:06: You very easily get seduced into this world of higher structures, which obviously has its own beauty.

00:27:16: The learning curve is also very problematic because it normally always starts very slow... Oh there are things in their arrows or these other morphisms.

00:27:26: three minutes later they say the co-products Zimi quasi prime ideal.

00:27:35: I mean, you shouldn't do this...I think you can do it in a pedestrian speed.

00:27:41: There's an issue that I have as mathematicians and they often tell me too many things And then i get lost and don't even understand the basic thing.

00:27:52: So I think its good to approach a bit pedestrian and slowly introduce one concept after another.

00:28:00: Yeah, but we were with the metaphor of the addicting nature of category theory and in contrast type theory is harmless.

00:28:09: No I mean it's not called dependent types for nothing no?

00:28:18: That's true!

00:28:20: Great.

00:28:21: so let us call that a day here as this pedestrian level at some point will build on it.

00:28:28: besides Cartesian product There was nothing mentioned without a definition.

00:28:33: We could mention exponentials as well or function types, that's very nice and beautiful if you look at the definition of functions in a categorical way but it is already really nice.

00:28:47: Okay great then.

00:28:49: thank-you all for listening.

00:28:51: we will follow up on this topic.

00:28:52: I guess we aren't done with our introduction into category theory.

00:28:57: these are great first steps.

00:29:00: So as always feel free to comment ask question Correct us if there was some slip and we said something.

00:29:07: We didn't intended to say

00:29:09: Something wrong.

00:29:11: Yeah, yeah that wasn't what

00:29:12: I mean.

00:29:13: you say sometimes They are

00:29:14: only wrong.

00:29:15: yes

00:29:16: No no but we want to test the attention of our audience

00:29:20: Exactly.

00:29:22: That's also how i would like to frame it.

00:29:24: so whether let see where they use spotted all our little arrows.

00:29:28: Technique is breaking down, so it's now the moment.

00:29:30: Yeah yeah my life is doing funny tricks on me.

00:29:35: Okay lovely then see you all around next week for our

00:29:39: guest.

00:29:40: bye-bye.

New comment

Your name or nickname, will be shown publicly
At least 10 characters long
By submitting your comment you agree that the content of the field "Name or nickname" will be stored and shown publicly next to your comment. Using your real name is optional.