Compare With HDBSCAN
Core-SG is most meaningful when compared against repeated HDBSCAN runs for
several k values.
from hdbscan import HDBSCAN
from core_sg import CoreSGClusterer
clusterer = CoreSGClusterer(k_max=30)
clusterer.fit(X, k=10)
hdb = HDBSCAN(min_cluster_size=10, min_samples=10)
hdb.fit(X)
core_labels = clusterer.labels_
hdbscan_labels = hdb.labels_
Small differences can occur because Core-SG reconstructs hierarchy artifacts through its support graph and HDBSCAN adapter boundary.