Choosing k_max

k_max defines the reusable support graph capacity. It must be large enough for the largest k that will be extracted later.

clusterer = CoreSGClusterer(k_max=30)

Practical Guidance

Choose k_max as the largest neighborhood value in your planned analysis. For example, if you want to compare k values 30, 20, and 10, fit with k_max=30.

A larger k_max can increase build cost because the reusable support graph must cover more neighborhood information. A smaller k_max limits what can be extracted later.

Repeated multi-k workflows are where Core-SG is most useful. If you only need one value, plain HDBSCAN may be simpler.