Repeated Multi-k Workloads
The intended workload is:
clusterer = CoreSGClusterer(k_max=50)
for k in range(50, 1, -1):
clusterer.fit(X, k=k)
For CoreSGClusterer, repeated calls to fit(X, k=...) after the first
call reuse core_sg_ and should be interpreted as extraction cost, not a
full rebuild.
This is the workload where Core-SG is expected to differ most from running
HDBSCAN independently for each k.