From mboxrd@z Thu Jan 1 00:00:00 1970 X-Msuck: nntp://news.gmane.io/gmane.science.mathematics.categories/1668 Path: news.gmane.org!not-for-mail From: "Mamuka Jibladze" Newsgroups: gmane.science.mathematics.categories Subject: Re: coinduction Date: Sat, 28 Oct 2000 21:39:57 +0200 Message-ID: <000901c04116$db094e40$c2036882@ucl.ac.be> NNTP-Posting-Host: main.gmane.org Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Trace: ger.gmane.org 1241018007 32333 80.91.229.2 (29 Apr 2009 15:13:27 GMT) X-Complaints-To: usenet@ger.gmane.org NNTP-Posting-Date: Wed, 29 Apr 2009 15:13:27 +0000 (UTC) To: Original-X-From: rrosebru@mta.ca Sat Oct 28 17:38:48 2000 -0400 Return-Path: Original-Received: (from Majordom@localhost) by mailserv.mta.ca (8.11.1/8.11.1) id e9SKAHF08131 for categories-list; Sat, 28 Oct 2000 17:10:17 -0300 (ADT) X-Authentication-Warning: mailserv.mta.ca: Majordom set sender to cat-dist@mta.ca using -f X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 5.00.2919.6600 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2919.6600 Original-Sender: cat-dist@mta.ca Precedence: bulk X-Keywords: X-UID: 26 Original-Lines: 81 Xref: news.gmane.org gmane.science.mathematics.categories:1668 Archived-At: Hi all, although I cannot say much on coinduction itself, I would like to mention one possible point of view on "universal coalgebra" commented on by Dusko at the end of his message: > PS i have a great appreciation for jan rutten's work on > coinduction, but it seems to me that the title of *universal > coalgebra* is a bit exaggerated. universal algebra does not > dualize. by design, it is meant to be model theory of algebraic > theories. what would be the "coalgebraic theories", leading to > coalgebras as their models? what would their functorial > semantics look like? (implicitly, these questions are already > in manes' book.) > > and then, moving from coalgebraic theories to comonads, like > from algebraic theories to monads, should presumably yield an > HSP-like characterization for the categories of (comonad) > coalgebras. but coalgebras for a comonad over a topos usually > form just another topos. > > i may be missing something essential here, but then i do stand > to be corrected. I would say that, although there is no straightforward dualization, there might be a "hidden" one. In my opinion, comonads can sometimes play the role of "strengtheners" or "enrichers" for monads, so that while monads can be viewed as a "substrate for structure", comonads are "substrate for semantics". Rather than trying to give sense to this vague phrase, let me give three illustrations. -1- S. Kobayashi, Monad as modality. Theoret. Comput. Sci. 175 (1997), no. 1, 29--74. The starting point in that paper is the monad semantics by Moggi. It seems that for computer-scientific purposes strength of the monad is too restrictive a requirement, and the author modifies the semantics by assuming strength of the monad *only*with* respect*to*a*comonad*. This means that the monad and comonad distribute in such a way that the monad descends to a strong monad on the category of coalgebras over the comonad. Under the Curry-Howard correspondence this yields an interesting intuitionistic version of the modal system S4. A realizability interpretation is also given. -2- Freyd, Yetter, Lawvere, Kock, Reyes and others have studied "atoms", or infinitesimal objects - objects I in a cartesian closed category with the property that the exponentiation functor I->(_) has a right adjoint. This induces a monad, which cannot be strong in a nontrivial way, just as in the previous example. There is a "largest possible" subcategory over which it is strong, namely the category of I-discrete objects, i.e. those objects X for which the adjunction unit from X to I->X is iso. In good cases the inclusion of this subcategory is comonadic. As Bill pointed out, one might view Cantor's idea of set theory as a case of this: the universe of sets is the largest - necessarily "discrete" - part of "analysis" over which the latter can be "enriched" (well, this is awfully inaccurate, sorry -- consider it as a metaphor). -3- In differential homotopical algebra, there was developed a notion of torsor (principal homogeneous object) which works in a non-cartesian monoidal category. This notion requires not just a (right) action of a monoid M on an object P, but also a "cartesianizer" in form of a left coaction of a comonoid C on P which commutes with the action, and is principal in an appropriate sense. It turns out that the monoid and the comonoid enter in an entirely symmetric way. In the monoidal category of differential graded modules over a commutative ring, principal C-M-objects are classified by s.c. twisting cochains from C to M. Moreover in that category, for a monoid M there is a universal choice of P and C given by the Eilenberg-Mac Lane bar construction, and for a comonoid C there is a universal choice of P and M, given by the Adams' cobar construction. I would be grateful for a reference to a version of these for general monads/comonads. Regards, Mamuka