Possible to specify custom distance matrix? about k-means-constrained

[joshlk]

No it does not. But if you want to hack your own solution you can try changing this line:

k-means-constrained/k_means_constrained/k_means_constrained_.py

Line 402 in eda4074

D = euclidean_distances(X, C, squared=False)

[amks1]

Thanks. Sorry if this is a dumb question, but are the cluster centers chosen from the set of input points?

[bibblybobblyben]

Hi - thanks for your work building this package. I have also come across the question of trying to implement a custom distance metric in the clustering so have opened a PR to address it. Happy to be corrected if this is not a desirable change in the main package.