Tidyclust

[kbodwin]

trafficstars

This is my writeup with some details about clustering (unsupervised learning) and how I envision that fitting into the tidymodels framework.

[alexpghayes]

Just wanted to comment that for unsupervised methods there is often both a forward and a backward transformation. Less so with clustering, but for many PCA-like tools. I previously brought this up at little in tidymodels/recipes#264.

Another possible consideration before starting to prototype things is the difference between inductive/transductive models, or methods that can be applied to a new dataset versus those that cannot.

[michaelgaunt404]

Heyo! really excited to see this developing. I've recently fallen into the world of unsupervised clustering (via some gnar text projects) and have been having a hard time understanding the literature and not being able to find certain methods in tidymodels.

[brshallo]

Very cool! Looking forward to developments here and this functionality coming to tidymodels!

I posted a toy solution on SO for validating kmeans cluster partition stability on a holdout set: https://stackoverflow.com/a/68845111/9059865 . (For anyone stumbling onto this thread and looking for something simple in the interim before {celery} 😊 gets implemented in tidymodels.)

[EmilHvitfeldt]

Closing this as we already this this :)