Want to take part in these discussions? Sign in if you have an account, or apply for one below
Vanilla 1.1.10 is a product of Lussumo. More Information: Documentation, Community Support.
I am hereby moving ancient query-box discussion out of the entry to here:
+– {: .query} Mike: I just made this up, but it seems right. With a $\le$-version of cyclic orders we could maybe recover monotone functions more directly. Whatever the definition of monotone function is, it should let us recover $\Lambda$ as below.
Toby: Doubting a $\le$ version and believing that reflecting linear orders is quite natural, I've phrased this entirely in those terms. Also, fixed (in my opinion) the definition of strict monotone function. (It's interesting that this suggests that $\Delta$ should be seen as a category of linearly ordered sets; being finite sets, this is equivalent even constructively.) =–
+– {: .query} Mike and Toby: If anyone wants to have a go at a cyclic analogue of a total order, be our guest. =–
more such query-box discussion, hereby moved out of the entry to here:
+– {: .query} Todd: Guys, I may be confused here, but I’m having a problem: it seems that the empty relation on the 1-element set 1 is a cyclic order (in fact the only one on 1), and therefore (under the definition that monotone functions are those which reflect the cyclic order) that 1 is terminal. But then the category of finite (nonempty) cyclically ordered sets and monotone sets would have to be contractible. How does this square with the homotopy type of Connes’ cyclic category (which I’m told is the same as that of $S^1$)?
This is the point that I stumbled on in the blog discussion on the cyclic category almost exactly two years ago, when I attempted to give my own description of $\Lambda$ (but using a cyclic analogue of total orders, which you two have challenged readers to give). You can see my attempted description here, and my retraction here.
Toby: The problem is that the $0$-cycle (unlike the $[0]$-simplex) should not be a point but instead have a nontrivial loop, is that right? And so it's not terminal, since a map to it from $[n]$ really consists of $n + 1$ binary choices (whether to be a degeneracy or to map to the nontrivial loop).
Todd: I’m sorry – I must be making some embarrassingly stupid mistake – but I still don’t follow. What loop? For the 1-element set {t}, there are only two ternary relations. The relation consisting of the point (t, t, t) cannot be a cyclic order by irreflexivity. So the only cyclic order is the empty one (which seems to be vacuously allowed by the axioms). And it seems that the condition that the unique map f: C –> {t} is monotone [by Mike’s definition], for any cyclically ordered set C, is vacuously satisfied.
Trust me – I want you guys to be right, because I think that the definition of $\Lambda$ I attempted two years ago is (in classical logic) equivalent to yours (although I should check that carefully) – and I was disappointed when I thought it failed.
Toby: It's also possible that I'm making an embarrassingly stupid mistake, but if you're making one, then it's probably missing my word ’should’. I agree with you that there is a unique cyclic order (as defined here) on the $1$-element set and that this gives the terminal cyclically ordered set. What I'm not certain about is that I understand what the cycle category is supposed to be, but is it right that the $1$-element cycle (the $0$-cycle) should have a nontrivial loop? Because if so, then I understand what your objection is!
Todd: Part of the problem is that I don’t have a deep feeling myself for Connes’ $\Lambda$. But David Ben-Zvi (I believe) mentioned on the blog that $\Lambda$ has the homotopy type of a circle, meaning I guess that the classifying space of the nerve $B N \Lambda$ has this homotopy type. But the classifying space of a category with a terminal object is contractible. That’s the point where I got worried about my own construction, and it’s a worry for me here as well: terminal objects would seem to be fatal.
Toby: OK, so neither of us understands it well enough to be sure. Still, if I get the description as an ordered graph, then I see that there is a problem with maps to the $0$-cycle.
The good news is that, looking at your definition of a cyclically reflexive notion of cyclic order, I can no longer imagine why I might have thought that such a thing would not work. Indeed, applying de Morgan duality to the axioms above for an irreflexive version, we immediately get a reflexive version that is (classically, and even constructively for finite sets) equivalent to your definition. (It's still true that the irreflexive version is probably better from a constructive perspective, much as is true for linear/total orders, because cyclic totality fails for the reflexive ternary order relation on the unit circle in the located Dedekind complex plane, but that's not what was concerning me before, nor is it relevant for finite sets.)
Mike Shulman: Probably this whole discussion should be taking place at cycle category.
According to Berger-Moerdijk’s characterization of the cycle category (Example 2.7), I think the $0$-cycle is not terminal. They describe it as the “total category” of a certain “crossed $\Delta$-group” which is a presheaf $n\mapsto G_n$ of groups on $\Delta$ with certain extra structure; in this case the relevant presheaf sends each set $[n] = \{0,1,\dots,n\}$ to $C_n$, the cyclic group on $n$ letters. The total category of a crossed $\Delta$-group has the same objects as $\Delta$, and the morphisms $[m]\to [n]$ are pairs $(\alpha,g)$ where $\alpha:[m]\to [n]$ in $\Delta$ and $g\in G_m$. Thus, in particular, the morphisms $[m]\to [0]$ in $\Lambda$ can be identified with elements $g\in C_m$, so there is more than one of them.
I think I have just lost whatever geometric intuition I used to think I had for the cycle category.
David Corfield: I think I understood it once as “$\Lambda$ is the category whose objects $[n]$ are circles marked by the sets of $n$ th roots of unity. Maps are degree 1 mappings between circles which send marked points to marked points, and are increasing. This is quite like $\Delta$ but where you are allowed to cycle round the domain before you choose an order-preserving map.”
It’s good to see how the morphisms are counted.
So $\Lambda$ is a kind of product between $\Delta$ and the groupoid which is a union of one copy of each finite cyclic group. =–
1 to 3 of 3