You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As reported by @Witiko in #20097 (reply in thread), the centroids' initialization in BallTree is tight to the case of the euclidean distance regardless of the DistanceMetric set indirectly by the user provided metric argument.
The initialisation of those centroids should depends on the effective DistanceMetric. Apart from the euclidean case, are there closed form solutions for computing them?
The text was updated successfully, but these errors were encountered:
As reported by @Witiko in #20097 (reply in thread), the centroids' initialization in
BallTree
is tight to the case of the euclidean distance regardless of theDistanceMetric
set indirectly by the user providedmetric
argument.scikit-learn/sklearn/neighbors/_ball_tree.pyx
Lines 71 to 104 in 7b71511
The initialisation of those centroids should depends on the effective
DistanceMetric
. Apart from the euclidean case, are there closed form solutions for computing them?The text was updated successfully, but these errors were encountered: