Discussion of Homotopy Type Theory and Univalent Foundations
 help / color / mirror / Atom feed
From: Peter LeFanu Lumsdaine <p.l.lu...@gmail.com>
To: Steve Awodey <awo...@cmu.edu>
Cc: Michael Shulman <shu...@sandiego.edu>,
	Thierry Coquand <Thierry...@cse.gu.se>,
	 homotopy Type Theory <homotopyt...@googlegroups.com>
Subject: Re: [HoTT] Semantics of higher inductive types
Date: Wed, 7 Jun 2017 11:40:11 +0200	[thread overview]
Message-ID: <CAAkwb-k1GnZKkrEoszNQQ0zyA3HSFh1LXmKVp-+3juRk5s8taQ@mail.gmail.com> (raw)
In-Reply-To: <CAAkwb-nogDyVRwnvDXUphBQb0ttL-TuHfSahnrSMhFnuunc7hw@mail.gmail.com>

[-- Attachment #1: Type: text/plain, Size: 4309 bytes --]

On Thu, Jun 1, 2017 at 6:08 PM, Peter LeFanu Lumsdaine <
p.l.lu...@gmail.com> wrote:

> On Thu, Jun 1, 2017 at 6:56 PM, Steve Awodey <awo...@cmu.edu> wrote:
> >
> > you mean the propositional truncation or suspension operations might
> lead to cardinals outside of a Grothendieck Universe?
> Exactly, yes.  There’s no reason I know of to think they *need* to, but
> with the construction of Mike’s and my paper, they do.  And adding stronger
> conditions on the cardinal used won’t help.  The problem is that one takes
> a fibrant replacement to go from the “pre-suspension” to the suspension
> (more precisely: a (TC,F) factorisation, to go from the universal family of
> pre-suspensions to the universal family of suspensions); and fibrant
> replacement blows up the fibers to be the size of the *base* of the
> family.  So the pre-suspension is small, but the suspension — although
> essentially small — ends up as large as the universe one’s using.

I realise I was a bit unclear here: it’s only suspension that I meant to
suggest is problematic, not propositional truncation.  The latter seems a
bit easier to do by ad hoc constructions; e.g. the construction below does
it in simplicial sets, and I think a similar thing may work also in cubical
sets.  (I don’t claim originality for this construction; I don’t think I
learned it from anywhere, but I do recall discussing it with people who
were already aware of it or something similar (I think at least Mike,
Thierry, and Simon Huber, at various times?), so I think multiple people
may have noticed it independently.)

So suspension (or more generally pushouts/coequalisers) is what would make
a really good test case for any proposed general approach — it’s the
simplest HIT which as far as I know hasn’t been modelled without a size
blowup in any infinite-dimensional model except cubical sets, under any of
the approaches to modelling HIT’s proposed so far.  (Am I right in
remembering that this has been given for cubical sets?  I can’t find it in
any of the writeups, but I seem to recall hearing it presented at

Construction of propositional truncation without size blowup in simplicial

(1)  Given a fibration Y —> X, define |Y| —> X as follows:

an element of |Y|_n consists of an n-simplex x : Δ[n] —> X, together with a
“partial lift of x into Y, defined at least on all vertices”, i.e. a
subpresheaf S ≤ Δ[n] containing all vertices, and a map y : S —> Y such
that the evident square commutes;

reindexing acts by taking pullbacks/inverse images of the domain of the
partial lift (i.e. the usual composition of a partial map with a total map).

(2) There’s an evident map Y —> |Y| over X; and the operation sending Y to
Y —> |Y| —> X is (coherently) stable up to isomorphism under pullback in X.

(3) In general, a fibration is a proposition in the type-theoretic sense
iff it’s orthogonal to the boundary inclusions δ[n] —> Δ[n] for all n > 0.
 (Non-trivial but not too hard to check.)

(4) The map |Y| —> X is a fibration, and a proposition.  (Straightforward,
given (3), by concretely constructing the required liftings.)

(5) The evident map Y —> |Y| over X is a cell complex constructed from
boundary inclusions δ[n] —> Δ[n] with n > 0.

To see this: take the filtration of |Y| by subobjects Y_n, where the
non-degenerate simplices of Y_n are those whose “missing” simplices are all
of dimension ≤n.  Then Y_0 = Y, and the non-degenerate simplices of Y_{n+1}
that are not in Y_n are all {n+1}-cells with boundary in Y_n, so the
inclusion Y_n —> Y_{n+1} may be seen as gluing on many copies of δ[n+1] —>

(6) The map Y —> |Y| is orthogonal to all propositional fibrations, stably
in X.  (Orthogonality is immediate from (3) and (5); stability is then by

(7) Let V be either the universe of “well-ordered α-small fibrations”, or
the universe of “displayed α-small fibrations”, for α any infinite regular
cardinal.  Then V carries an operation representing the construction of
(1), and modelling propositional truncation.  (Lengthy to spell out in
full, but straightforward given (2), (6).)


[-- Attachment #2: Type: text/html, Size: 5150 bytes --]

  parent reply	other threads:[~2017-06-07  9:40 UTC|newest]

Thread overview: 25+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-05-25 18:25 Michael Shulman
2017-05-26  0:17 ` [HoTT] " Emily Riehl
2017-06-01 14:23 ` Thierry Coquand
2017-06-01 14:43   ` Michael Shulman
2017-06-01 15:30   ` Steve Awodey
2017-06-01 15:38     ` Michael Shulman
2017-06-01 15:56       ` Steve Awodey
2017-06-01 16:08         ` Peter LeFanu Lumsdaine
2017-06-06  9:19           ` Andrew Swan
2017-06-06 10:03             ` Andrew Swan
2017-06-06 13:35               ` Michael Shulman
2017-06-06 16:22                 ` Andrew Swan
2017-06-06 19:36                   ` Michael Shulman
2017-06-06 20:59                     ` Andrew Swan
2017-06-07  9:40           ` Peter LeFanu Lumsdaine [this message]
2017-06-07  9:57             ` Thierry Coquand
     [not found]             ` <ed7ad345-85e4-4536-86d7-a57fbe3313fe@googlegroups.com>
2017-06-07 23:06               ` Michael Shulman
2017-06-08  6:35                 ` Andrew Swan
2018-09-14 11:15               ` Thierry Coquand
2018-09-14 14:16                 ` Andrew Swan
2018-10-01 13:02                   ` Thierry Coquand
2018-11-10 15:52                     ` Anders Mörtberg
2018-11-10 18:21                       ` Gabriel Scherer
2017-06-08  4:57     ` CARLOS MANUEL MANZUETA
2018-11-12 12:30       ` Ali Caglayan

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=CAAkwb-k1GnZKkrEoszNQQ0zyA3HSFh1LXmKVp-+3juRk5s8taQ@mail.gmail.com \
    --to="p.l.lu..."@gmail.com \
    --cc="Thierry..."@cse.gu.se \
    --cc="awo..."@cmu.edu \
    --cc="homotopyt..."@googlegroups.com \
    --cc="shu..."@sandiego.edu \


* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).