Reusing One Estimator Across Multiple k Values

The estimator workflow is designed around reuse:

clusterer = CoreSGClusterer(k_max=30)

for k in [25, 20, 15, 10]:
    clusterer.fit(X, k=k)
    labels = clusterer.labels_

Build once, extract several k values

CoreSGClusterer(k_max=30)

↓

fit(X, k=25): build core_sg_ and expose labels_ for k=25

↓

fit(X, k=20): reuse core_sg_ and expose labels_ for k=20

↓

fit(X, k=15): reuse core_sg_ and expose labels_ for k=15

↓

fit(X, k=10): reuse core_sg_ and expose labels_ for k=10

Only the first fit builds the reusable support graph. Later fits update the estimator outputs for the requested k.

Interpretation

k_max defines the largest neighborhood value available from this estimator. The first fit(...) builds the internal support graph. Every later k extraction reuses the same core_sg_ object and recomputes the current MST and hierarchy outputs for that target value.

The attributes labels_, probabilities_, cluster_persistence_, and the current tree artifacts always refer to the most recent extraction.

Use clusterer.core_sg_ only for advanced inspection of fit-time artifacts.