Initial commit for implementation of Affinity Propagation

[Yuan-Ru-Lin]

Implement interface of Affinity Propagation to MLJ. The implementation mimics that of KMedoids and directly uses the existing MMI.predict

[ablaom]

@Yuan-Ru-Lin I see this is a draft. Please go ahead and ping me when you are ready for a review. No hurry.

[Yuan-Ru-Lin]

The code is ready to be reviewed. I basically just made the model Static, turned MMI.transform and MMI.fit into MMI.predict and added doc.

Two dependencies are added: LinearAlgebra and StatsBase for we need diagind, triu! and median to fill the diagonal part of the similarity matrix with the median similarity of all pairs. I am not familiar with handling compat so it may need your attention there.

Oh I should mention that I use GLMakie in the doc since it makes visualization easier. I'm not sure where I should add that as a dependency.

[Yuan-Ru-Lin]

@ablaom Forgot to ping you!

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 16 lines in your changes missing coverage. Please review.

Project coverage is 79.34%. Comparing base (3d8548b) to head (7d0453c).
Report is 3 commits behind head on master.

Files with missing lines	Patch %	Lines
src/MLJClusteringInterface.jl	0.00%	16 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           master      #28       +/-   ##
===========================================
- Coverage   96.10%   79.34%   -16.76%     
===========================================
  Files           1        1               
  Lines          77       92       +15     
===========================================
- Hits           74       73        -1     
- Misses          3       19       +16     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[ablaom]

@Yuan-Ru-Lin This is looking great. Thanks for taking the time to wrap your head around the MLJ API and the docstring requirements.

Can you add some basic test that include:

• check the report contains what you think it does
• generic interface test (just add AffinityPropagation here.
• test that the preference=nothing fallback has expected behaviour.

You can bump Julia support in the Project to 1.10 (hopefully fixes your fail) and, if you know how, do the same bump in .github/workflows/ci.yaml - otherwise I can do it.

[Yuan-Ru-Lin]

Hi @ablaom, I have

• added tests for the model
• bumped supported Julia version from 1.6 to 1.10 in both Project.toml and ci.yml
• edit docs as you suggested.

[Yuan-Ru-Lin]

@ablaom Morning from PST! The PR is ready to be reviewed again.

[ablaom]

Thanks again @Yuan-Ru-Lin for this valuable contribution.