Skip to content

Add variate transport #62

New issue

Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.

#

By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? # to your account

Jump to bottom
Merged
merged 70 commits into from
Jun 19, 2022
Merged

Add variate transport #62

merged 70 commits into from
Jun 19, 2022

Conversation

oschulz
Copy link
Collaborator

@oschulz oschulz commented Jun 15, 2022

No description provided.

@oschulz oschulz requested a review from cscherrer June 15, 2022 14:06
@oschulz oschulz marked this pull request as draft June 15, 2022 14:06
@github-actions
Copy link

Package name latest stable
MeasureTheory.jl

@github-actions
Copy link

Package name latest stable
MeasureTheory.jl

@codecov
Copy link

codecov bot commented Jun 15, 2022 •
edited
Loading

Codecov Report

Merging #62 (a8faa66) into master (bf7eae6) will increase coverage by 8.53%.
The diff coverage is 78.57%.

@@            Coverage Diff             @@
##           master      #62      +/-   ##
==========================================
+ Coverage   39.02%   47.55%   +8.53%     
==========================================
  Files          34       37       +3     
  Lines         756      942     +186     
==========================================
+ Hits          295      448     +153     
- Misses        461      494      +33
Impacted Files Coverage Δ
src/MeasureBase.jl 60.00% <ø> (ø)
src/insupport.jl 0.00% <0.00%> (ø)
src/transport.jl 63.88% <63.88%> (ø)
src/combinators/transformedmeasure.jl 71.42% <80.64%> (+71.42%) ⬆️
src/getdof.jl 87.50% <87.50%> (ø)
src/interface.jl 96.29% <94.11%> (-3.71%) ⬇️
src/combinators/power.jl 84.90% <100.00%> (+12.40%) ⬆️
src/combinators/weighted.jl 81.81% <100.00%> (+2.87%) ⬆️
src/primitives/dirac.jl 70.58% <100.00%> (+9.04%) ⬆️
src/primitives/lebesgue.jl 80.95% <100.00%> (+4.48%) ⬆️
... and 5 more

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update bf7eae6...a8faa66. Read the comment docs.

cscherrer
cscherrer reviewed Jun 15, 2022
View reviewed changes
Copy link
Collaborator

@cscherrer cscherrer left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks @oschulz ! I have some questions, may have a better basis for review after I understand some things better

@github-actions
Copy link

Package name latest stable
MeasureTheory.jl

@github-actions
Copy link

Package name latest stable
MeasureTheory.jl

@github-actions
Copy link

Package name latest stable
MeasureTheory.jl

@github-actions
Copy link

Package name latest stable
MeasureTheory.jl

@github-actions
Copy link

Package name latest stable
MeasureTheory.jl

@oschulz
Copy link
Collaborator Author

oschulz commented Jun 19, 2022 •
edited
Loading

If we want to stress the transport aspect, I'd suggest transport_map(nu, mu) or gettransport(nu, mu) (with transport_def(nu, mu, x) for the internal function, and nu the target and mu the source).

@oschulz
Copy link
Collaborator Author

oschulz commented Jun 19, 2022

Ok, new list for bikeshedding:

  • transformation_to(nu, mu)
  • transformation_map(nu, mu)
  • transport_to(nu, mu)
  • transport_map(nu, mu)

I like the ones with _to better, but I'm Ok with all of them.

@cscherrer , @mschauer , any picks?

@cscherrer
Copy link
Collaborator

I think it's important to try to distinguish between a measure transformation and a support transformation (that can be pushed forward or pulled back for some measure). And the to/from thing could be confusing.

If the concept of a transport is meaningful when the spaces are different, I like transport_to

@oschulz
Copy link
Collaborator Author

oschulz commented Jun 19, 2022

I think it's important to try to distinguish between a measure transformation and a support transformation (that can be pushed forward or pulled back for some measure). And the to/from thing could be confusing.

If the concept of a transport is meaningful when the spaces are different, I like transport_to

If I read https://en.m.wikipedia.org/wiki/Transportation_theory_(mathematics) correctly, the spaces can be different. Correct, @mschauer?

We do not guarantee that we return an optimal transport, though (we shovel the earth correctly but not necessarily with minimum effort).

@mschauer
Copy link
Member

Exactly, we also don’t promise an optimal transport, right? :-)

@oschulz
Copy link
Collaborator Author

oschulz commented Jun 19, 2022

Exactly, we also don’t promise an optimal transport, right? :-)

Indeed no, just a "correct transport, not an optimal one.

@cscherrer
Copy link
Collaborator

@mschauer if we change to transport_to do you think we're good to go?

I'm going to be out for a while, so once you're both happy with this please go ahead and merge, and tag a release

@mschauer
Copy link
Member

I should not tag releases on the phone but I am happy with the PR

@oschulz
Copy link
Collaborator Author

oschulz commented Jun 19, 2022

I should not tag releases on the phone but I am happy with the PR

Ok, let me do the rename, then we'll tag?

@oschulz
Copy link
Collaborator Author

oschulz commented Jun 19, 2022

Ok, renamed, let's see if CI goes through.

@mschauer Ok like this?

@github-actions
Copy link

Package name latest stable
MeasureTheory.jl

@oschulz
Copy link
Collaborator Author

oschulz commented Jun 19, 2022

Ah, stop, need to rename vartransform_origin to

@oschulz
Copy link
Collaborator Author

oschulz commented Jun 19, 2022

Ok rename complete.

@github-actions
Copy link

Package name latest stable
MeasureTheory.jl

@oschulz
Copy link
Collaborator Author

oschulz commented Jun 19, 2022

I should not tag releases on the phone but I am happy with the PR

I can do the squash-merge, increase version number and tag, if there are no objections?

@oschulz
Copy link
Collaborator Author

oschulz commented Jun 19, 2022 •
edited
Loading

This will be v0.11.0, not v0.10.1, right?

@oschulz
Copy link
Collaborator Author

oschulz commented Jun 19, 2022

This will be v0.11.0, not v0.10.1, right?

Ah, yes - has to be, since it moves out StdNormal.

@github-actions
Copy link

Package name latest stable
MeasureTheory.jl

@oschulz oschulz merged commit 648376d into JuliaMath:master Jun 19, 2022
@oschulz oschulz deleted the transforms branch June 19, 2022 21:27
@oschulz
Copy link
Collaborator Author

oschulz commented Jun 19, 2022

Released. :-)

@oschulz oschulz changed the title Add vartransform Add variate transport Jun 20, 2022
@oschulz
Copy link
Collaborator Author

oschulz commented Jun 20, 2022 •
edited
Loading

I apologize, I overlooked one name - we still have the VarTransform type. Should I make a v0.12 (we could name it to TransportFunction? It's not a type users will interact with much, but maybe best to fix that naming inconsistency now instead of later?

@oschulz
Copy link
Collaborator Author

oschulz commented Jun 20, 2022

I did a grep for var ... we also forgot about test_vartransform, NoTransport, checked_var, NoVarCheck and a few internal functions during the rename (oops).

@oschulz oschulz mentioned this pull request Jun 20, 2022
Merged
@oschulz
Copy link
Collaborator Author

oschulz commented Jun 20, 2022

#63 should fix it - grep shows no remaining "...var..." where it doesn't belong.

# for free to join this conversation on GitHub. Already have an account? # to comment
Reviewers

@cscherrer cscherrer cscherrer approved these changes

@mschauer mschauer mschauer approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@oschulz @cscherrer @mschauer