I’m finding this takes a long time on my data set of about ~1 million data points over 11 classes. I think it takes at least 15–20 minutes, maybe longer. I tried Condensed Nearest as well as Near Miss. Is there some other implementation or method that could work?