[WIP] Translation Invariant Sinkhorn for UOT #668
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #668 +/- ##
==========================================
- Coverage 96.85% 96.80% -0.05%
==========================================
Files 88 88
Lines 17910 18002 +92
==========================================
+ Hits 17346 17427 +81
- Misses 564 575 +11 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Types of changes
This PR aims to add the Translation Invariant Sinkhorn algorithm proposed in Faster Unbalanced Optimal Transport: Translation invariant Sinkhorn and 1-D Frank-Wolfe
to solve the Unbalanced OT problem.
sinkhorn_unbalanced_translation_invariantinunbalanced.pyplot_conv_sinkhorn_ti.pyto demonstrate the convergence benefitstest_unbalanced.pyMotivation and context / Related issue
This version of Sinkhorn converges faster than the classical Sinkhorn algorithm for UOT.
How has this been tested (if it applies)
The function can be called by using the method "sinkhorn_translation_invariant" in
sinkhorn_unbalanced, so I added this method in the itertools of tests oftest_unbalanced.py.PR checklist