[MRG] CO-Optimal Transport solver #447
Merged
Conversation
This file contains 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
This reverts commit f3d36b2.
Codecov Report
Additional details and impacted files@@ Coverage Diff @@
## master #447 +/- ##
==========================================
+ Coverage 94.85% 94.92% +0.06%
==========================================
Files 30 31 +1
Lines 6770 6879 +109
==========================================
+ Hits 6422 6530 +108
- Misses 348 349 +1 |
agramfort
approved these changes
Mar 18, 2023
|
no idea what's happening with the doc here... |
|
CrcleCI ususally runs in 20 minutes but sometimes when it is busy it is very slow and the doc buid is canceled (here not during an example from this PR). I cannot ask for a new run but the dame doc is running and building on other PR right now.... |
rflamary
reviewed
Mar 21, 2023
README.md
Outdated
| @@ -303,6 +303,8 @@ Dictionary Learning](https://arxiv.org/pdf/2102.06555.pdf), International Confer | |||
|
|
|||
| [46] Bonet, C., Berg, P., Courty, N., Septier, F., Drumetz, L., & Pham, M. T. (2023). [Spherical Sliced-Wasserstein](https://openreview.net/forum?id=jXQ0ipgMdU). International Conference on Learning Representations. | |||
|
|
|||
| [47] Chowdhury, S., & Mémoli, F. (2019). [The gromov–wasserstein distance between networks and stable network invariants](https://academic.oup.com/imaiai/article/8/4/757/5627736). Information and Inference: A Journal of the IMA, 8(4), 757-787. | |||
| [47] Redko, I., Vayer, T., Flamary, R., and Courty, N. (2020). [CO-Optimal Transport](https://proceedings.neurips.cc/paper/2020/file/cc384c68ad503482fb24e6d1e3b512ae-Paper.pdf). Advances in Neural Information Processing Systems, 33. | |||
There was a problem hiding this comment.
add your ref at the end and use the same number in the doc of the function
| # Main function | ||
|
|
||
| if method_sinkhorn not in ["sinkhorn", "sinkhorn_log"]: | ||
| raise ValueError( |
There was a problem hiding this comment.
make tets that check that the error is raised
| eps_samp, eps_feat = epsilon, epsilon | ||
| else: | ||
| if len(epsilon) != 2: | ||
| raise ValueError("Epsilon must be either a scalar or an indexable object of length 2.") |
| alpha_samp, alpha_feat = alpha, alpha | ||
| else: | ||
| if len(alpha) != 2: | ||
| raise ValueError("Alpha must be either a scalar or an indexable object of length 2.") |
| ny_feat, type_as=Y)) # shape nx_feat, ny_feat | ||
| else: | ||
| pi_samp, pi_feat = warmstart["pi_sample"], warmstart["pi_feature"] | ||
| duals_samp, duals_feat = warmstart["duals_sample"], warmstart["duals_feature"] |
There was a problem hiding this comment.
also add a warmstart in one of teh tests
| break | ||
|
|
||
| if verbose: | ||
| print( |
| (vx_samp, vx_feat, vy_samp, vy_feat, gradX, gradY)) | ||
|
|
||
| if log: | ||
| return coot, dict_log |
rflamary
changed the title
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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
Implementation of CO-Optimal Transport (COOT) [47] in the new file
coot.py, containing two methods:co_optimal_transport: returns the sample and feature couplings.co_optimal_transport2: returns the COOT distance.Both methods can handle:
ot.lp.emdsolver will be used.ot.sinkhornsolver will be used.The COOT distance outputed by
co_optimal_transport2is also sub-differentiable with respect to the input matrices and marginal distributions.Motivation and context / Related issue
CO-Optimal Transport is not yet available on POT.
How has this been tested (if it applies)
test_coot.py.PR checklist