Add KMedoids

[juliohm]

No description provided.

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.24%. Comparing base (1de5b52) to head (8ad1323).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #298      +/-   ##
==========================================
+ Coverage   98.17%   98.24%   +0.07%     
==========================================
  Files          48       49       +1     
  Lines        1367     1424      +57     
==========================================
+ Hits         1342     1399      +57     
  Misses         25       25              

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[eliascarv]

I think the only point that we don't have a simple way to resolve would be the type of dists = fill(Inf, nobs), as we don't have a practical way of knowing what the type returned by the parwise of the columns would be.

So I think starting it as Float64 is the simplest and safest solution, just like it is in the PR.

[juliohm]

Perhaps we can evaluate the distance with a single row to get the type. Will try when I'm back to the computer. Em qui., 27 de mar. de 2025, 19:10, Elias Carvalho ***@***.***> escreveu:

…

***@***.**** approved this pull request. — Reply to this email directly, view it on GitHub <#298 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAZQW3LHTWXJQSPRRCM5EYD2WRZM5AVCNFSM6AAAAABZ53TJECVHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZDOMRTGQ3TQMZQGM> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[juliohm]

There is still an issue regarding unitful columns with different units, but that should be addressed in TableDistances.jl. If distances are computed with different units, and these units can't be added, we will get an error. Perhaps TableDistances.jl should remove the units to make it work, or warn users that columns must have compatible units.

[juliohm]

I am going to merge assuming that the last review was an approval with comment 👍🏽

[eliascarv]

Amazing PR @juliohm!

[juliohm]

Quick correction: the StdFeats transform applied in the beginning of KMedoids takes care of the units since it applies the ZScore for Continuous variables. I pushed a fix to master renaming the resulting column from cluster to label and will do the same downstream for consistency.