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.
Created a new page generalized graph based on the definition given in Hackney, Robertson and Yau’s recent book, which appears to be influenced at least in part by the paper of Kock cited on the page. As far as I can tell none of the other things on the nlab (e.g. quiver or graph or their associated sub-entries about pseudographs and so forth) deal with the case of the “exceptional cell.” If the notion I describe on this page is already somewhere on the nlab, I’d be happy to know that and get rid of the page I made.
I’m not really following. Where is the output and input data in the definition? Which part is vertices and which part is edges?
Ah you know what. I need to beef up this definition a bit, since the “directedness” doesn’t show up in this generality.
But the vertices are given by the partition indices, and the flags being a given part of the partition are the ones that “attach” to that vertex.
Should have worked on it quite a bit more before posting a link to it here. Doing that now.
What role does the infinite set $\mathcal{S}$ play?
Also, I wouldn’t call those “properties” of a generalized graph; those are structures.
Are all the cells supposed to be nonempty, including the exceptional cell? (Which would be usual when speaking of a partition, but I’m just checking to make sure.)
I feel as though I’d have something more to say later about this page.
The cells don’t need to be non-empty. In fact there is, for instance, an empty graph. Perhaps I’m saying things incorrectly. I’m trying to not stray too far from Hackney, Robertson and Yau’s work. The ultimate point of all this is that one can then define infinity properads, which generalize infinity categories and infinity operads, except that the indexing category, rather than being the simplex category or the tree category, is a certain category of generalized graphs.
Oh, and the set $\mathcal{S}$ is just the set of names of half edges. One could easily use any infinite set.
But honestly, looking back at it, the set $\mathcal{S}$ really isn’t necessary. That was just something I ported over directly from the book, but in retrospect there’s no reason for it.
Right. My only thought here is that maybe this $\mathcal{S}$ is similar to wanting to have a countable stock of names of variables. (In operadic or string-diagrammy pictures, it’s sometimes useful to translate between diagrams and and more traditional syntactic terms where external edges are named by free variables and internal edges by bound variables. But taken on their own terms, the string diagrams don’t need such labels; the edges can be left to name themselves.)
I’m not really grasping yet the reason for the terminology “flags”, “cells”, etc., although I’m familiar with such terms from other contexts. Maybe the reason will emerge later, but I’d like to put a pin through that.
In my impatience, I feel an urge to rewrite the definition in terms more familiar to me (which is something I can do in the privacy of my own home), but I do encourage avoiding term usages like “partition” here that actually clash with usage at e.g. partition. Part of the data seems just to be a map $s: F \to V + 1$ which maps a flag to the vertex it lies over, except that exceptional flags map to the summand $1$. (Leaving out mention of the involutions.) If I were any good at graphics, I’d try to help out by drawing up some pictures.
Or maybe I do sort of get it? A flag is supposed to conjure perhaps a picture of a node and half-open edge sticking out from it? So that if two flags are paired by the involution, you’re supposed to think of sticking them together (identifying the interiors of the half-open edges) to get a full-fledged edge with two endpoints. Or at least so in the unexceptional case.
Since $F_\epsilon$ is finite, the condition $\iota F_\epsilon \subseteq F_\epsilon$ is equivalent to $\iota F_\epsilon = F_\epsilon$, right? So now the structure induced on the finite set $F_\epsilon$ is an involution $\iota$, together with a further free involution $\pi$ on the $\iota$-fixed points. That seems to me to be the same as giving one free involution, $\iota'$ say, defined by $\iota'(x) = \iota(x)$ if $\iota(x)\neq x$ and $\iota'(x) = \pi(x)$ if $\iota(x)=x$, together with a subset of $F_\epsilon$ closed under $\iota'$ (the $\iota$-fixed points).
It also seems to me that we could get rid of the other $\iota$-fixed-points by adding new flags in $F_\epsilon$ to be their images; the previous $F_\epsilon$ would then simply be the “loops” therein. Then the entire involution $\iota$ would be free, and so we could simply pass to its quotient to obtain a set $E$ of edges, with a function assigning to each edge the 1- or 2-element set consisting of its vertices.
In other words, it looks to me as though a “generalized graph” is just
Is this true? If so, is there a reason this way of presenting it is less useful? I find it much easier to visualize. (And in any case, I don’t understand the role of the “exceptional legs”.)
Yeah, so there are definitely other ways to define this type of stuff. I’m not personally attached to this way of doing it. Basically I want to write up these basic things so I can extend the dictionary of segal spaces giving infinity categories and segal dendroidal spaces giving infinity operads to segal graphical spaces giving infinity properads.
Do we have by the way some discussion (somewhere in the nLab) of Joachim Kock’s nice work on trees and polynomial endofunctors? That could help illuminate some of what is going on here…
In response to Mike’s comments also, I am way out of my depth in being able to respond to these questions. I trust your analysis of the redundancies in the definition. I can give a reason for the necessity of exceptional edges, but am less clear on exceptional legs myself.
Ultimately, one is going to want to have a “graphical category” (which is frustratingly denoted $\Gamma$ in HRY) in which the morphisms are basically graph substitution. That way functors $\Gamma\to Set$ or $\Gamma\to sSet$ will define properads or simplicial properads. One definitely needs an exceptional edge here when one gets around to defining things like inner face maps and substitutions products and so forth.
An “exceptional leg” is basically just a half edge with no vertices (I think of it as either being the tail or the head of an arrow). I guess the main idea is that this is useful in gluing generalized graphs, since you need to glue them along things, so if you want to do this as a pushout, you need be able to pick out legs.
Could you say again how this is unifying (if it is?) simplicial sets with dendroidal sets etc.? Maybe it would be good to have a remark on this in the entry.
Oh sure. The idea is that there is a series of inclusions $\Delta\hookrightarrow \Omega\hookrightarrow \Gamma$, where $\Gamma$ is basically the category of wheel-free graphs with morphisms generated by certain face and degeneracy maps (which restrict to the usual ones on $\Delta$ and $\Omega$). Then there are three categories: simplicial sets, dendroidal sets and graphical sets with the relevant adjunctions between them. The analogy goes much further. For instance, you can define inner horns in graphical sets that restrict to inner horns in simplicial sets (and the inner horns of dendroidal sets defined by Cisinski and Moerdijk). By enriching everything, one can construct simplicial model category structures which present: $\infty$-categories, $\infty$-operads, and $\infty$-properads. The point of properads of course is that they are less general than PROPs (hence easier to work with) but that pretty much every interesting example of a PROP is in fact a properad (e.g. bialgebras, bialgebras with (co)modules, Lie bialgebras).
Great, thanks. Maybe you could find the time to add this paragraph at a suitable place on the $n$Lab and cross-link with model structure on dendroidal sets? That would be useful
Yes definitely. I’m still trying to figure out how exactly to structure all of this. I think the next step is to create a page for the graphical category, and that’s where the above paragraph would probably go. There are some slight intimations of that kind of structure on the wheeled graph page, under the part where I define all the different types of wheeled graphs, but I’ll me more explicit on this other page. The only issue with putting together the graphical category page is that I have to explain the morphisms in it, which basically come from graph substitution (but only certain kinds of graph substitution), which makes it a bit unwieldy to write out.
pretty much every interesting example of a PROP is in fact a properad (e.g. bialgebras, bialgebras with (co)modules, Lie bialgebras).
I assume you mean something like “most small props that present ’theories’ can in fact also be presented by a properad”. I don’t think a prop can “be” a properad since it has more operations, except in the sense you don’t mean that every prop has an underlying properad. Also I think there are lots of (colored) props that are not in any sense properads, e.g. the underlying (colored) prop of a symmetric monoidal category.
(None of which detracts from your real point, of course, I’m just being pedantic (-: except that I do care about props of the latter sort myself.)
1 to 20 of 20