Reorganise sinkhorn_stabilized and sinkhorn_stabilized_epsscaling
#84
Conversation
Pull Request Test Coverage Report for Build 888723622
💛 - Coveralls |
There was a problem hiding this comment.
Thanks, these are some good improvements! I planned to update these functions as well and add some tests 👍 There are still some additional improvements that I have in mind but I think they can be postponed to other PRs. I only made some minor suggestions and I think we should fix the sequence of the epsilons right away (it's incorrect currently).
Can you also update the tests to the new format?
Co-authored-by: David Widmann <devmotion@users.noreply.github.com>
Co-authored-by: David Widmann <devmotion@users.noreply.github.com>
zsteve
force-pushed
the
sinkhorn_stabilized
branch
from
d71a2f1 to
a72be90
Compare
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
test/entropic.jl
Outdated
| # compute optimal transport map (Julia implementation + POT) | ||
| eps = 0.001 | ||
| γ = sinkhorn_stabilized(μ, ν, C, eps; maxiter=5_000) | ||
| γ_pot = POT.sinkhorn(μ, ν, C, eps; method="sinkhorn_stabilized", numItermax=5_000) |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
| γ_pot = POT.sinkhorn(μ, ν, C, eps; method="sinkhorn_stabilized", numItermax=5_000) | |
| γ_pot = POT.sinkhorn( | |
| μ, ν, C, eps; method="sinkhorn_stabilized", numItermax=5_000 | |
| ) |
test/entropic.jl
Outdated
| # compute optimal transport map (Julia implementation + POT) | ||
| eps = 0.001 | ||
| γ = sinkhorn_stabilized_epsscaling(μ, ν, C, eps; k=5, maxiter=5_000) | ||
| γ_pot = POT.sinkhorn(μ, ν, C, eps; method="sinkhorn_stabilized", numItermax=5_000) |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
| γ_pot = POT.sinkhorn(μ, ν, C, eps; method="sinkhorn_stabilized", numItermax=5_000) | |
| γ_pot = POT.sinkhorn( | |
| μ, ν, C, eps; method="sinkhorn_stabilized", numItermax=5_000 | |
| ) |
Codecov Report
@@ Coverage Diff @@
## master #84 +/- ##
==========================================
+ Coverage 88.36% 96.12% +7.76%
==========================================
Files 1 1
Lines 275 284 +9
==========================================
+ Hits 243 273 +30
+ Misses 32 11 -21
Continue to review full report at Codecov.
|
Pull Request Test Coverage Report for Build 887585953Warning: This coverage report may be inaccurate.This pull request's base commit is no longer the HEAD commit of its target branch. This means it includes changes from outside the original pull request, including, potentially, unrelated coverage changes.
Details
💛 - Coveralls |
Pull Request Test Coverage Report for Build 887559779Warning: This coverage report may be inaccurate.This pull request's base commit is no longer the HEAD commit of its target branch. This means it includes changes from outside the original pull request, including, potentially, unrelated coverage changes.
Details
💛 - Coveralls |
This PR updates
sinkhorn_stabilizedandsinkhorn_stabilized_epsscalingto use the same calling convention as the othersinkhornfunctions, the convergence criterion, as well as a general refactor.I think it should be doable to allow batch computations similar to #67 , but those can be added in a separate PR once both this and #67 have been merged.