Not signed in (Sign In)

Not signed in

Want to take part in these discussions? Sign in if you have an account, or apply for one below

  • Sign in using OpenID

Discussion Tag Cloud

Vanilla 1.1.10 is a product of Lussumo. More Information: Documentation, Community Support.

Welcome to nForum
If you want to take part in these discussions either sign in now (if you have an account), apply for one now (if you don't).
    • CommentRowNumber1.
    • CommentAuthormaxsnew
    • CommentTimeSep 26th 2017
    • (edited Sep 26th 2017)

    I asked this question on mathoverflow but I think I should have just posted here in the first place.

    A TT-multicategory in the sense of Crutwell-Shulman, where TT is a monad on a virtual double category CC is a monoid in the “horizontal kleisli category”, i.e., an object of objects OO, a horizontal arrow of arrows A:OTOA : O \to T O with composition and identity cells. Is there a good way to define a co-presheaf on such a generalized multicategory?

    This seems to give a natural notion of TT-presheaf as just a bimodule of the monad with a terminal object, i.e. a horizontal arrow P:1TOP : 1 \to T O with a composite P;APP;A \Rightarrow P that is compatible with composition and identity. If you look at what this means for specific cases like T=T = free symmetric monoidal category monad on the virtual equipment of cats, functors and profunctors this looks like the definition of presheaf you would come up with, with P:1TOP : 1 \to T O giving an abstract notion of map from a list of objects of OO to PP and so can be used to define universal properties like the product, and it looks like you can use the language of cartesian cells to define the right notion of representability.

    On the other had, just taking the dual doesn’t look like the right thing. What I’m hoping would happen is that if I try to define the universal property of a coproduct using such a copresheaf I would be “forced” to make it a distributive coproduct. However, if we say a copresheaf is a bimodule Q:OT1Q : O \to T 1, this doesn’t look right because for our example TT, this would give us an abstract notion of maps Q nAQ^n \to A for each AOA \in O, but it seems to me that the right notion of copresheaf (based on the type theory) would be to give an abstract notion of A 1,...,Q,...A nBA_1,...,Q,...A_n \to B.

    In particular for A,BOA,B \in O if I try to define a copresheaf A+BA+B by

    (A+B) nC=Π m+l=nA m,B lC(A+B)^n \to C = \Pi_{m+l=n} A^m,B^l \to C

    this only gives me a distributive coproduct-like behavior when A+BA+B repeated is the only thing in the domain, and it looks like for a representing object

    (A+B),CD=(A,CD)×(B,CD)(A+B),C \to D = (A,C \to D)\times (B,C \to D)

    will not in general be true.

    • CommentRowNumber2.
    • CommentAuthorMike Shulman
    • CommentTimeSep 27th 2017

    I saw your question on MO, but don’t have an answer for it; this is something I’ve also pondered. See for instance Definition 4.2.11 in my categorical logic notes.

    • CommentRowNumber3.
    • CommentAuthormaxsnew
    • CommentTimeSep 27th 2017
    • (edited Sep 27th 2017)

    Thanks Mike, I’ll take a look. What is the reason for the name “Hopf left adjoint”, i.e., the Hopf part?

    • CommentRowNumber4.
    • CommentAuthorMike Shulman
    • CommentTimeSep 27th 2017

    It’s a reference to a Hopf adjunction, which I believe is named because it induces a Hopf monad, which is named by analogy with a Hopf algebra.

    • CommentRowNumber5.
    • CommentAuthormaxsnew
    • CommentTimeOct 2nd 2017

    Thanks Mike, this is exactly the kind of thing I was looking for.