Introduction

The HDBSCAN estimator implementation from scikit-learn-contrib/hdbscan has been adopted, modified and refactored to conform to scikit-learn API and is now merged into the hdbscan feature branch. There are still several changes to be made both before and after merging hdbscan-->main, and the goal of this issue is to serve as a tracker for the remaining changes, as well as to host meta discussion regarding the estimator as a whole as needed.

In particular, I would encourage discussion regarding:

1. What other tasks may be relevant/necessary for HDBSCAN overall.
2. What tasks should be promoted from follow-up work to mandatory work before merging into main.

To do for merger into main

Mandatory work before consideration for final merger

• CLN Cleaned cluster/_hdbscan/_linkage.pyx #24857
• CLN Cleaned cluster/_hdbscan/_reachability.pyx #24701
• Clean _hdbscan/_tree.pyx
  • CLN Update cluster/_hdbscan/_tree.pyx style and syntax #25768
  • ENH _hdbscan HIERARCHY data type introduction #25826
  • CLN Cleaned TreeUnionFind in _hdbscan/_tree.pyx #25827
  • HDBSCAN _tree.pyx overhaul
    • CLN _hdbscan/_tree.pyx algorithms refactor #26011
    • CLN BFS Style Improvement #26096
    • CLN HDBSCAN _tree.pyx::do_labelling refactor #26101
• ENH Extends outlier encoding scheme to HDBSCAN.dbscan_clustering #24698
• DOC Adds HDBSCAN.dbscan_clustering section to plot_hdbscan.py #25538
• MAINT Sync hdbscan branch with main #25134

Follow-up after merger into main

Discussion regarding follow-up tasks has been moved to #26801

CC: @thomasjpfan @jjerphan @glemaitre