6

I am reading texts about (co)ends, and everywhere I see a lack of examples. I am not an expert in this area, and without examples it is difficult for me to use my intuition to grasp the idea. MacLane and Loregian mention some examples in passing, in particular, the so-called "geometric realization" $$ \int^n (Sn)\cdot\Delta n, $$ and the "integral against measure" $$ \int_X f(x) d\mu, $$ but they don't give details, and I don't understand what is meant here.

Can anybody enlighten me,

What are typical examples of (co)ends?

Not necessarily the examples that MacLane and Loregian mention, but just examples with accurate formulations: "if we take this bifunctor, then this construction will be the (co)end"... Of course, the more examples, the better. Thank you.

Edit. From people's comments here I see that I have to clarify that by examples I mean examples for non-specialists, namely, the constructions from other fields of mathematics that could be interpreted as (co)ends. The examples that I see up to now in the texts are methodical, they explain to specialists the details of the definition, and do not rouse interest of non-specialists.

At the same time, what could rouse this interest, are not examples of (co)ends: as an illustration, in Example 1.4.5 Fosco Loregian mentions the Stokes theorem, but, as far as I understand the construction he considers there is a cowedge, not a coend. So it remains unclear why (co)ends are important.

Gregory Arone gave a link to a thread where some informative examples are discussed. It will take me some time to analyze this, but I hope, people will add some more examples here if there indeed are some.

David White
  • 29,779
  • 3
    This question asked for intuition rather than examples, but the answers include some nice examples. – Gregory Arone Apr 17 '21 at 16:47
  • @GregoryArone thank you! It's strange that I did not see this... – Sergei Akbarov Apr 17 '21 at 16:54
  • 5
    what do you mean "some" examples? There are many :D – fosco Apr 17 '21 at 17:49
  • @fosco after Gregory's comment I indeed see several examples, which Todd Trimble and Anton Fetisov describe with some details (and I need some time to analyse this). Could you, please, explain in detail what you mean by $$ \int_X f(x) d\mu $$ in your text? Which bifunctor do you consider, and which construction is its end (interpreted as this integral)? – Sergei Akbarov Apr 17 '21 at 18:00
  • @PhilippeGaucher I mentioned this text in my question. Perhaps I did not explain this clearly, the examples that I see there are examples for people who already have intuition. To a person who is not a specialist, these examples tell not much. What could be indeed useful for understanding all this are the examples that show which constructions in other fields of mathematics can be interpreted as (co)ends. I don't see such examples. As an illustration, in Example 1.4.5 Fosco mentions the Stokes theorem. But, as far as I understand the construction he describes there is a cowedge, not a coend. – Sergei Akbarov Apr 17 '21 at 19:48
  • Maybe, I don't see something important. Excuse me in this case. – Sergei Akbarov Apr 17 '21 at 19:50
  • 1
    It seems you're trying to deduce that integrals in analysis are examples of (co)ends; this is not true, as far as I know. Instead, one can draw a connection between the two concepts at an informal level: both an integral and a coend depend from the same variable twice, once covariantly and once contravariantly; for both integrals (of nice functions) and coends, the order in which you perform an integral in many variables does not matter; an integral is a certain limit of sums, and a (co)end is a (co)limit of... well, this analogy breaks down a little bit, but you get the idea. (cntd') – fosco Apr 17 '21 at 20:38
  • 1
    (/2) Now, coming to your question: examples of coends. There's an entire book of examples :) natural transformations can be obtained as an end. (Pointwise) Kan extensions are ends (on the right) or coends (on the left). Algebraic topology is chock-full of coends (singular complexes, nerves of categories...). Weighted co/limits can be appreciated in full when they are expressed as co/ends. Profunctors, operads... I'm basically listing the chapters of the book! In case you need something specific, just ask, but I have no clue how to find more examples than those 300+ pages :D – fosco Apr 17 '21 at 20:42
  • 1
    Finally, let's talk about Stokes' theorem. The way in which one can define the Stokes' pairing as a cowedge towards $\mathbb R$ is very interesting. You're right that's not a coend, just a cowedge. But cowedges have coends, and the coend of the map $\int$ defined in 1.4.5 is an interesting object, it is linked to a certain de Rham cohomology group. If I remember well, I have left this as a riddle for the reader to find out. – fosco Apr 17 '21 at 20:44
  • To really conclude, it also appears to me (correct me if I'm wrong) that you want to get an intuition about co/ends, without knowing category theory. This is not the way to go; intuition comes after a long time of no-intuition, and the no-intuition interzone is unavoidable to gain said intuition. All the more because, being familiar with the fundamentals of category theory, co/ends are simple objects: they are just "yet another initial/terminal object of a certain category". Nothing more. – fosco Apr 17 '21 at 20:52
  • @fosco regarding this intergal: $$ \int_X f(x) d\mu. $$ I think, in the next edition of this text, you should explain to the reader that this is not a special example of a (co)end, but a heuristic consideration. – Sergei Akbarov Apr 17 '21 at 21:24
  • 1
    @fosco I would say, on the contrary: usually intuition preceeds studying. For example, to understand the idea of category you should just look at several examples: the classes of sets, of topological spaces, of vector spaces, etc. Similarly, to understand the idea of monoidal category you should look at the constructions of "products" in these classes. There is no necessity for thorough and long studying of category theory if your aim is just to understand what it is about. Moreover, for most people it is impossible to start to study something without understanding what it is about. – Sergei Akbarov Apr 18 '21 at 04:27
  • I see Mathematics as a path to enlightenment; and very much like satori you can't attain it if you think about it, and you can't attain it if you don't. Finding your way out of this conundrum is your daily task as a mathematician. Anyway, there's still something I don't understand about your question: each chapter of that book is an entire class of motivating examples. What else would you want? Examples of category-theoretic constructions outside category theory? – fosco Apr 18 '21 at 07:31
  • Regarding $\int_X f dx$: it seems to me you didn't read the second-to-last line of page i: "At the cost of pushing this analogy further than permitted... etc etc". I have no idea how to make more clear the heuristic nature of that consideration. Anyway, I'm wiling to help you: I just fail to see what could possibly constitute a satisfying answer. – fosco Apr 18 '21 at 07:35
  • @SergeiAkbarov Sorry. Yes I had not checked the link in your post. It's why I had removed my useless comment before you answer it (I guess that my comment was still in your cache). – Philippe Gaucher Apr 18 '21 at 09:22
  • @fosco for me this phrase is too vague (perhaps, because English is not my native): "At the cost of pushing this analogy further than permitted... etc etc". Don't think that I did not like your book, I find it very interesting, better than other texts on this topic. My only advice is that it would be good to clarify these details to non-specialists. If you'll give (perhaps, from the very beginning) a list of examples that can be understood by different people, not only by specialists in category theory, this will attract much more readers (and will make your book much more popular). – Sergei Akbarov Apr 18 '21 at 09:47
  • @PhilippeGaucher it's ok. – Sergei Akbarov Apr 18 '21 at 09:50
  • 2
    I would like to refer you to my question "Toy examples of Kan extensions" (https://math.stackexchange.com/questions/1108266/toy-examples-for-kan-extensions) which got many good answers. Since left Kan extensions can be written as coends, this also gives you several examples of coends. – Martin Brandenburg Apr 21 '21 at 17:29
  • @MartinBrandenburg thank you, Martin! – Sergei Akbarov Apr 21 '21 at 20:53

1 Answers1

4

From what I understand, the OP wants a list of examples of (co)ends, where the notion is something familiar that doesn't require a background in category theory to understand. Something like this list of examples of adjoint functors, including free functors, functors of the form $-\otimes X$, induction $Ind_H^G$ from a subgroup $H<G$, and the discrete topology (left adjoint to the forgetful functor from $Top$ to $Set$). Since this question has no answer, I'll pull together a list, most of which have already appeared in the links in the comments to the OP.

  1. Let $R$ be a ring, and $A$ (resp. $B$) be a right (resp. left) $R$-module. Then $A\otimes_R B = \int^R A\otimes B$ is a coend, fundamental to algebra. See Example 3.2.12 of Loregian's book.

  2. The geometric realization functor from topological spaces to simplicial sets, left adjoint to the singularization functor, is a coend $\int^{n\in \Delta} \Delta^n \times X_n$. Similarly, the categorical realization of a simplicial set, arising from $\tau_1: sSet\leftrightarrows Cat: N_i$, is a co-end, as is the coherent nerve. See Sec 3.1.2 and Examples 3.2.5, 3.2.6, 3.2.7 of Loregian's book. This is also a left Kan extension.

  3. Given two functors $F,G:C\to D$, the collection of natural transformations $Nat(F,G)$ is an end, built from pieces of the form $Hom_C(Fc,Gc)$ over all $c\in C$. As a special case, the global section functor of a sheaf $\mathcal{F}$ is $\Gamma(X,\mathcal{F}) = \int_U \mathcal{F}(U)$ over a particular collection of $U$. See also Theorem 1.4.1 in Loregian's book.

  4. Viewing a metric space as a category $X$ enriched over $\mathbb{R}_+$, for any $F:X\to \mathbb{R}_+$, the supremum of $F$ is the end $\sup F = \int_{x\in X} F(x)$.

  5. Stokes' Theorem has already been mentioned in the comments, and appears as Example 1.4.5 of Loregian's book.

  6. Reconstructing a $G$-space $X$ from its fixed points is a co-end $\int^{H\in Orb(G)} X^H \times G/H \cong X$. Similarly, one can apply co-ends to extensions in Galois theory; see Exercise 1.8 of Loregian's book.

  7. Let $V$ be a finite dimensional $k$-vector space and let $V^\vee$ be the dual vector space of linear transformations $V\to k$. Then the co-end reconstructs $k$ from the vector spaces, as $\int^V V^{\vee} \otimes_k V \cong k$. This is Example 2.3.12 of Loregian's book.

  8. The Fourier transform is a co-end. See Exercise 5.19 of Loregian's book.

  9. As Loregian points out in a comment, "(Pointwise) Kan extensions are ends (on the right) or coends (on the left)" - see Chapter 2 of Loregian's book. So, the Dold-Kan correspondence from simplicial abelian groups to non-negatively graded chain complexes, is a left Kan extension, hence a co-end. See Example 3.2.10 of Loregian's book. Similarly, subdivision functors in simplicial contexts.

  10. Building on the above, there are many examples of Kan extensions, including limits of functors (as right Kan extensions, hence ends), colimits of functors, direct sum, kernels and cokernels, extending presheaves (Example 2.7 in the link), etc. As the link says "any list is necessarily wildly incomplete."

  11. The convolution product is a co-end; see Loregian Prop 6.2.1. Also, it's a left Kan extension. The composition product for symmetric sequences (used to define operads) is a co-end; see Chapter 6 of Loregian, and exercise 6.8.

  12. Homotopy Kan extensions are examples of homotopy (co)ends. These see application in physics, including extended quantum field theories and path integral quantization.

  13. The induced representation functor, coming from a group homomorphism $f: H\to G$ (or, an inclusion $H\leq G$), is a left Kan extension. Also, taking the connected components of a simplicial object, or its Cech nerve.

  14. Weighted (co)limits are examples of (co)ends; see Loregian chapter 4 and Sec 2.4.2. Here's a concrete example. If $R$ is a ring and $f$ is a map of chain complexes of $R$-modules, then the mapping cone of $f$ is the co-end $C(f) \cong \int^i W(i)\otimes f(i)$ where $W$ is the collection of weights in this weighted colimit. The mapping cone is fundamental to homological algebra. See Example 4.2.1 of Loregian's book.

Obviously, there are many, many more examples. But, I've run out of steam for now. I may add more later. Ends and co-ends are also useful for what they can prove, e.g., the Bousfield-Kan formulas for homotopy (co)limits.

David White
  • 29,779
  • David, I didn't understand, in which sense a metric space is a category enriched over ${\mathbb R}_+$? – Sergei Akbarov Jan 08 '24 at 17:02
  • David, yes, I would be grateful if you could add other examples. – Sergei Akbarov Jan 08 '24 at 17:04
  • It's a classical example in category theory, spelled out here: https://golem.ph.utexas.edu/category/2023/05/metric_spaces_as_enriched_categories_ii.html among other places. The composition rule gives you the triangle inequality. – David White Jan 08 '24 at 17:10
  • 1
    Let me add one more to this list: the espace étalé of a (pre)sheaf $\mathscr F$ on a topological space $X$ is the coend $\int^U U \times \mathscr F(U)$ (taken in topological spaces over $X$). This makes pretty clear the adjunction between (pre)sheaves and local homeomorphisms, and the unit gives the sheafification functor. (Meanwhile, most textbooks make some construction where they first define the underlying set, then awkwardly endow it with a topology, and then prove everything by hand.) – R. van Dobben de Bruyn Jan 08 '24 at 22:13