This is about scikit-learn sklearn.metrics.pairwise and not about kmedoids.
The metric is computed correctly, it's is just a warning that in the future versions of scikit-learn the standardization coefficient V will not be automatically computed (so your code will not work anymore, an error will be issued), you will have to give the V as an argument.

Thanks for your fast reply.

Yes, strictly speaking it is about sklearn.metrics.pairwise and not kmedoids but how will I be able to pass V as an argument or can V be computed internally by Kmedoids?

Ah, you are right.
There are two solutions that I see

• you can use the precomputed metric and to compute the distance matrix beforehand.
• you can give a callable function into metric argument in KMedoids and the callable function would be something like lambda (X,Y): pairwise_distances(X, Y, metric='seuclidean', V=V)

you can give a callable function into metric argument in KMedoids and the callable function would be something like

I think we should handle this in scikit-learn-extra. So that passing metric='seuclidean' wouldn't raise a warning.

I would be very happy if scikit-learn-extra could take care of that instead of having to do some gimmicks myself.

Thank you so much!