This is an embarrassingly simple question, but I was not able to find a definitive answer from literature search.

Suppose one has some collection of functions $f_1: X \to Y_1, \dots, f_n: X \to Y_n$ on a common domain $X$. Then one can form the function $(f_1,\dots,f_n): X \to Y_1 \times \dots \times Y_n$ in the usual fashion:

$$ (f_1,\dots,f_n)(x) := (f_1(x),\dots,f_n(x)).$$

My question is: what does one call the function $(f_1,\dots,f_n)$? I had (without thinking much about it) used to call it the "direct sum" of $f_1,\dots,f_n$, before realising that this actually had no relation with the usual meaning of direct sum. In category theory, one might call $(f_1,\dots,f_n)$ the "product" of $f_1,\dots,f_n$, but this could get confusing if $f_1,\dots,f_n$ already take values in some ring, so that there is also a pointwise product. Is there some other commonly accepted term for describing the function $(f_1,\dots,f_n)$? It doesn't sound quite right grammatically to refer to it as the "tuple" of $f_1,\dots,f_n$, and "concatenation" or "join" don't quite seem to fit either.

  • 9
    In the case $n = 2$, I would call it the pairing. Similarly, one has "tripling", "quadrupling", and so in general I would call it the tupling of the list $f_1, \ldots, f_n$. – Todd Trimble Oct 5 '15 at 21:54
  • 5
    One option might be to call this the "diagonal" of the Cartesian product? I don't know though if such a name is standard outside of optimization where this "diagonal" arises as a part of the so-called "product-space trick" – Suvrit Oct 5 '15 at 22:38
  • 4
    Suvrit's comment is to some extent supported by Engelking's General topology, which is very precise about terminology. It uses the term "diagonal of mappings" and the notation $\bigtriangleup_{i=1}^n f_i$. (the definition is at the bottom of page 79 in the 1989 edition) – Niels J. Diepeveen Oct 6 '15 at 0:16
  • 4
    it IS the product (object) of the (objects) $(f_i : X \to Y_i)$ in the comma category $X \backslash YourCat$. So you could perfectly well call your map the (comma) product of those maps, and leave out "comma" when it gets tiresome. – Jesse C. McKeown Oct 6 '15 at 1:22
  • 7
    As $X_1\times\cdots \times X_n$ is called cartesian product I use the same name for $f=(f_1,\ldots,f_n)$. – Jochen Wengenroth Oct 6 '15 at 6:27
up vote 15 down vote accepted

I was encouraged to make my comment an answer:

In the case $n = 2$, I would call it the pairing. Similarly, one has "tripling", "quadrupling", and so in general one might call it the ($n$-)tupling of the list $f_1, \ldots, f_n$. And indeed that is what the nLab calls it: see here.

Whatever this should be called, I would not call it the cartesian product of $f_1, \ldots, f_n$. The product is a functor $\mathcal{C}^n \to \mathcal{C}$ whose value at a morphism $(f_1: X \to Y_1, \ldots, f_n: X \to Y_n)$ of $\mathcal{C}^n$ is rather the morphism $f_1 \times \ldots \times f_n: X \times \ldots \times X \to Y_1 \times \ldots \times Y_n$ of $\mathcal{C}$, and it's the latter that I would call the product of $f_1, \ldots, f_n$.

I will see if I can track down further citations for $n$-tupling.

  • 1
    If the maps happen to go into a field, "pairing" has the minor misfortune of being a term one might also use to describe an inner product of $f_1$ and $f_2$. To make matters worse, come to think of it, an inner product is often denoted $(f_1, f_2)$. Fortunately, there's an easy fix: just refer to $(f_1, f_2) \colon X \to Y_1 \times Y_2$ as the tupling even when you're only tupling two things (and hope nobody adopts this terminology for multilinear functionals, I guess). – Vectornaut Oct 6 '15 at 18:47
  • 1
    @Vectornaut I'd think if you referred to the pairing $(f_1, f_2): X \to Y_1 \times Y_2$ or $\langle f_1, f_2 \rangle: X \to Y_1 \times Y_2$, then context should usually be sufficient to disambiguate this sense of pairing from bilinear pairing, since a product $Y_1 \times Y_2$ of two rings is never a field. :-) – Todd Trimble Oct 6 '15 at 19:27
  • tupling is nice; though one could perhaps also use 'stringing' but that may strike the wrong chord.... It is also interesting to note that syntactically, 'tupling a list' as noted above, literally means 'parenthesizing' it --- the $f_i$ and the commas are already there! – Suvrit Oct 6 '15 at 22:37
  • 2
    I'm accepting this answer as being (a) one with a little bit of precedent (from the n-lab), (b) a term which does not require specialised knowledge in order to understand, and (c) the most popular of the alternatives suggested here. I also like the fine distinction between the tuple $(x \mapsto f_1(x),\dots,x \mapsto f_n(x))$ of $f_1,\dots,f_n$ and the tupling $x \mapsto (f_1(x),\dots,f_n(x))$ of $f_1,\dots,f_n$; these two concepts are canonically and naturally equivalent and it is a fairly safe abuse of notation to call them both $(f_1,\dots,f_n)$, and the similar names support this. – Terry Tao Oct 8 '15 at 16:45

You could call $$\mathbf r(t)=\langle x(t), y(t), z(t)\rangle$$ the vector function (or vector?) of $(x,y,z)$.

  • I thought of mentioned this name but then I saw this mathworld.wolfram.com/VectorFunction.html and decided to pass...but I like this name nevertheless – Suvrit Oct 6 '15 at 18:42
  • I wanted to say something like "vector-valued function with components $f_i$" but this only really works for appropriate $Y_i$. – JP McCarthy Oct 6 '15 at 18:56

Perhaps you can take inspiration from computer science.

In "APL: an interactive approach", Gilman and Rose talk about the catenation and lamination operators for joining together appropriately shaped arguments which (in my view) are multidimensional arrays. I think 'catenating' functions is appropriate for this usage. (Gilman, Leonard, and Allen J. Rose. APL: an interactive approach. Krieger Publishing Co., Inc., 1992. I have the second edition from an earlier year at home, with catenation on page 138, I think.)

If you decide that you prefer something else, be bold and define a term. Although I like Brendan McKay's suggestion of 'assemblage', for some reason 'amalgam' appeals to me. However, it is important to know how you end up using the name and the construct: there may be other things that better deserve those names. Further, be sure to point out similar but different uses (e.g. 'amalgamated product'). (You know that already, but a reminder shouldn't hurt.) You might even give a nod to history and note how such things were named and used in early vector calculus and linear algebra texts.

Gerhard "Can Say You Were Ill-advised" Paseman, 2015.10.06

A reasonable case could be made for calling your tuple-function a multi-span in whichever category $C$ your $\{f_i\}_1^n$ inhabit. When $n=2$, this reduces to the usual span given by roof-diagrams which look like this:

$$Y_1 \stackrel{f_1}{\gets} X \stackrel{f_2}{\to} Y_2$$

In this happy $n=2$ case you can actually compose these guys provided $C$ admits pullbacks; in this case, you have a bicategory of spans in $C$. For larger $n$, things are much less structured (as far as I know).

The obvious dual gadgets (unfortunately, but not unexpectedly, called multi-cospans) have actually appeared in work by Grandis, pdf here.

Your Answer

 
discard

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.