EasyDeepLearn

Interview: 'how would you tune HDBSCAN's min⁡clustersize\operatorname{min}_{\mathrm{cluster}}\mathrm{size}?'

hard

Answer

  • (1) Business constraint: what's the smallest meaningful group in the domain?
  • (2) Sensitivity analysis: sweep min⁡clustersize\operatorname{min}_{\mathrm{cluster}}\mathrm{size} ∈ [5, 10, 20, 50, 100] → measure silhouette + noise-fraction + number-of-clusters.
  • (3) Stability: bootstrap Jaccard between runs.
  • (4) Choose the min⁡clustersize\operatorname{min}_{\mathrm{cluster}}\mathrm{size} that gives stable + interpretable + right number of clusters.
  • (5) Also tune min⁡samples\operatorname{min}_{\mathrm{samples}} (density conservativeness) and clusterselectionmethod\mathrm{cluster}_{\mathrm{selection}}\mathrm{method} ('leaf' vs 'eom').
  • Iterate.
Check yourself — multiple choice
  • Random
  • Business minimum group size + sweep [5..100] + silhouette + noise fraction + stability bootstrap; also tune min⁡samples  +  clusterselectionmethod\operatorname{min}_{\mathrm{samples}}\; + \;\mathrm{cluster}_{\mathrm{selection}}\mathrm{method} (leaf/eom)
  • Just default
  • Not tunable

HDBSCAN tuning: min⁡clustersize  +  min⁡samples  +  selection\operatorname{min}_{\mathrm{cluster}}\mathrm{size}\; + \;\operatorname{min}_{\mathrm{samples}}\; + \;\mathrm{selection} method; stability sweep.

#interview#clustering

Practise Unsupervised Learning

214 interview questions in this topic.

Read the lessonFlashcardsTake the quiz

Related questions