Zulip Chat Archive
Stream: Is there code for X?
Topic: Sum of Inf is Inf of sum
Pierre-Alexandre Bazin (Feb 01 2022 at 13:40):
example {X Y : Type} (a : X → ℝ) (b : Y → ℝ) :
((⨅ x, a x) + (⨅ y, b y) = ⨅ (x : X) (y : Y), a x + b y) :=
begin library_search end
do we have a lemma stating this ?
Pierre-Alexandre Bazin (Feb 01 2022 at 14:56):
I asked the same question on the discord, and it seems it's false if X or Y are empty. So the lemma would have to include X or Y non empty in hypotheses, and maybe it's also wrong if the inf is -infinity (i don't know how it behaves in Lean in this case)
Chris B (Feb 01 2022 at 15:38):
Looks like https://leanprover-community.github.io/mathlib_docs/data/pi.html#pi.add_def
Chris B (Feb 01 2022 at 15:40):
The has_add (f i)
assumption would function as the nonempty assumption I guess.
Chris B (Feb 01 2022 at 15:41):
Well, I guess not now that I think about it.
Anne Baanen (Feb 01 2022 at 15:41):
This is about the infimum operator ⨅
, not the dependent function Π
:)
Chris B (Feb 01 2022 at 15:42):
I've been had.
Vincent Beffara (Feb 01 2022 at 15:49):
Inf
of a set that is empty or not bounded below gives (if I understand the code) 0
so the equality is also wrong if one of the infs on the left is -infinity and the other is not.
Stuart Presnell (Feb 01 2022 at 15:51):
If you add the required hypotheses to the lemma does library_search
find anything?
Pierre-Alexandre Bazin (Feb 01 2022 at 16:06):
still nothing found for
example {X Y : Type} (a : X → ℝ≥0) (b : Y → ℝ≥0) (x : X) (y : Y) :
((⨅ x, a x) + (⨅ y, b y) = ⨅ (x : X) (y : Y), a x + b y) :=
begin library_search end
Eric Wieser (Feb 01 2022 at 16:28):
That's false when Y
is empty and X
is not, isn't it? In that case, it's infi a + 0 = 0
, which is not true for all a
Pierre-Alexandre Bazin (Feb 01 2022 at 16:34):
I added (x : X) and (y : Y) in hypotheses so that X and Y are non empty
Floris van Doorn (Feb 01 2022 at 16:40):
I don't think we have it, but docs#ennreal.infi_add_infi is related. Especially the proof is interesting, because your statement is proven as part of it (for ennreal
).
Floris van Doorn (Feb 01 2022 at 16:43):
You can probably mimic that proof to get the version for real
using the dual versions of docs#csupr_add and docs#add_csupr (which seem to be missing).
You are right that you have to think about boundedness: the infimum of a set of reals that is not bounded below is defined to be 0.
Last updated: Dec 20 2023 at 11:08 UTC