New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Apartness in sets with decidable equality (axiom removal) #227
Comments
PS: Another way to bring out the essence of the issue is to point out that the equivalence |
PPS: Another proposition defined as a negation and in use by us is the relation
We'll have to change that, too. It's easy, because |
Here is the branch I will use: |
I began to work on apartness the last week : https://github.com/cathlelay/UniMath/blob/dev/UniMath/Dedekind/Apartness.v |
I see you use On Mon, Nov 2, 2015 at 9:07 AM, cathlelay notifications@github.com wrote:
|
I see that the HoTT formalization avoids this problem. Here is their definition of decidability:
It follows that |
Here's how to say what an apartness relation is, in a nutshell: it's an axiom-free propositional relation that is equivalent, by a proof using consistent axioms, such as |
One more point is that if the user needs a complement P for |
On 11/02/2015 11:09 AM, Daniel R. Grayson wrote:
Ok, that sounds reasonable! |
I believe that's the same as "apartness is a propostional relation which cannot be shown different from ≠ (no matter what axioms we have, including none, as long as they are consistent)." |
Maybe. One way I can imagine proving that apartness differing from inequality is inconsistent would be by proving |
Consider the natural problem of proving that
Σ i, i≠0
andΣ i, i>0
areequivalent sets. The former set is the complement of
0
innat
, as encapsulatedin the following definition from PartC:
One easily defines functions in both directions:
, but then one has to prove that the two composite functions are homotopic to
identity functions, and there one uses the lemmas that show that
i≠0
andi>0
are propositions. The proof thati≠0
is a proposition involves anaxiom, because
i≠0
is defined asi=0 -> empty
. The axiomfunextempty
provides the needed statement:
Use of that axiom makes its way into the definition of the homotopy
q∘p ~ idfun _
, rendering it incomputable. Such homotopies are provided fromany weak equivalence by these functions:
Those homotopies, in turn, make their way into definitions of other maps
and equivalences, rendering them incomputable. For example, consider
weqfp
,a useful theorem:
, where one of the two maps in the equivalence it provides incorporates one of
those potentially incomputable homotopies:
The route for the propagation of incomputability passes through the following
basic lemma:
, which gets used in the proof of
, where
recompl
is defined this way:From there incomputability is transferred into
and thus into
This is unfortunate, because any
stn m
has an axiom-free propositionequivalent to
x≠y
, given by the evident recursive algorithm on nat, whichcould have been used instead to define the complement of a set! Yesterday I
started rewriting
weqdnicoprod
, but this morning I realized there is a betterway, which preserves all the theorems already proved.
I would like to fix this by postponing the decision to appeal to
funextempty
:We redefine
as
Using
funextempty
, one can show thatP x'
andx=x'
are equivalent.A suitable name for
P
would be "apartness from x". Similarly, we would redefine, any of whose applications would involve
funextempty
, asor as
, and redefine
as
A suitable name for such a
P
would be "apartness on X". We could even makethe following definition:
Similarly, the
P
inisdecprop
above could be called "a complement of X",and we could adopt this definition:
We could then also, more simply, define
isisolated
andisdeceq
in terms ofisdecprop
.Notice that under this proposal, even though theorems are seemingly weakened,
we lose nothing, because a user of the new definitions is always free to appeal
to
funextempty
if necessary. In that way, it's just like postponing the useof LEM until it's needed, by refraining from adopting it as axiom, and
insisting it be used as a hypothesis.
Let me know what you think. I'd like to start working on it immediately, since
it seems necessary.
The text was updated successfully, but these errors were encountered: