Jason (jcreed) wrote,

Aha! The notion of relative monads in "Monads Need Not Be Endofunctors" (which, by the way, is a fantastic title: provocative without being goofy) generalize Hughes's notion of arrows without making me memorize a huge bunch of fiddly unmotivated axioms. (Don't even get me started on Freyd categories...) Instead, arrows are relative monads on the yoneda embedding! This only requires me to remember a (slight variant on) a small bunch of fiddly axioms that I already memorized, namely the monad axioms.

The proof that it is a great title: I only heard about this by skimming a tiny workshop announcement I got in my email this morning and seeing the title, and figured I just had to read it.


The thing that surprises me just a little is the blatant asymmetry of definition 2: a relative adjunction between the functor J : JC and the category D is given by two functors L : JD and R : DC such that there is a natural isomorphism C(JX, RY) = D(LX, Y).

What happens if we instead have a "doubly-relative" adjunction that has
J : JC
L : JD
R : KC
M : KD
such that there is a natural isomorphism
C(JX, RY) = D(LX, MY)
Surely we ought to get some relative comonads out of this, too, at the very least.


Oh, oops, Arrows are actually strong relative monads on the yoneda embedding. That explains all the fiddly axioms about products and projections and stuff I guess!
Tags: monads, papers, relatives

  • Post a new comment


    Anonymous comments are disabled in this journal

    default userpic

    Your reply will be screened

    Your IP address will be recorded 

  • 1 comment