[MRG] Implementation of two news algorithms: SaGroW and PoGroW.

[Hv0nnus]

Add two new algorithms to solve Gromov Wasserstein: Sampled Gromov Wasserstein and Pointwise Gromov Wasserstein. Based on Sampled Gromov Wasserstein (Machine Learning Journal 2021) (https://hal.archives-ouvertes.fr/hal-03232509/document).

Types of changes

• Docs change / refactoring / dependency upgrade
• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)

Motivation and context / Related issue

How has this been tested (if it applies)

It has been tested on some GW problems, with various values of hyper-parameters.

Checklist

• The documentation is up-to-date with the changes I made. (I change few lines in README.md, comment the code and document it and provide some examples in examples/gromov/plot_gromov.py.)
• I have read the CONTRIBUTING document.
• All tests passed, and additional code has been covered with new tests. (I do not have changed the other tests, the deletions are only a typo correction: "constratints" to "constraints")

[codecov]

Codecov Report

Merging #275 (10df3f6) into master (96bf1a4) will decrease coverage by 0.27%.
The diff coverage is 85.60%.

@@            Coverage Diff             @@
##           master     #275      +/-   ##
==========================================
- Coverage   93.34%   93.06%   -0.28%     
==========================================
  Files          17       17              
  Lines        3547     3679     +132     
==========================================
+ Hits         3311     3424     +113     
- Misses        236      255      +19     

[rflamary]

Hell @Hv0nnus and thank you very much for this PR.

I did a code review to ensure that the integration in POT and its documentation is seamless, there are a few things to do but I'm confident we can merge soon.

[Hv0nnus]

After further check, I think the bug is not due to any of the modification of this branch but rather a bug on the test function "test_mapping_transport_class" on the file test_da.py. This bug occurs randomly, as shown on the .PNG. This bug should be probably not be solve on this branch and an issue should be raised. I test my branch and the main one, both have the same bug.

[rflamary]

Yes I agree, could you please open an issus with this? I think is should be corrected by setting the seed in the test but it should also be looked at because there is clearly a fail condition on this mapping estimation.