Skip to content

Releases: scikit-learn-contrib/hdbscan

Bugfixes for DBCV index

28 Jan 19:00
Compare
Choose a tag to compare
Pre-release
0.8.5

Version bump for bug fix release.

Validity Index

05 Jan 00:24
Compare
Choose a tag to compare
Validity Index Pre-release
Pre-release

We introduce the density based clustering validity index from Density Based Clustering Validation by Moulavi, Jaskowiak, Campello, Zimek and Sander, as well as some minor bug fixes, and the option to match the reference implementation of HDBSCAN*

Patch release

25 Nov 17:12
Compare
Choose a tag to compare
Patch release Pre-release
Pre-release
0.8.3

Version bump for patch release.

Fix for issue #62

15 Sep 13:04
Compare
Choose a tag to compare
Fix for issue #62 Pre-release
Pre-release
0.8.2

Bump version for issue #62 that was fixed earlier but didn't make 0.8.1

Minor fixes

18 Aug 19:46
Compare
Choose a tag to compare
Minor fixes Pre-release
Pre-release
Merge pull request #51 from whitewhim2718/fix-typos

Some small spelling/grammar fixes in "How HDBSCAN Works notebook

Cluster strengths

31 May 22:03
Compare
Choose a tag to compare
Cluster strengths Pre-release
Pre-release

Support for cluster strengths and general bug fixes.

Plotting refresh

29 Apr 23:45
Compare
Choose a tag to compare
Plotting refresh Pre-release
Pre-release

A visual refresh of the plotting routines, along with a series of bug fixes for various corner cases.

Working on Linux Again

29 Feb 01:32
Compare
Choose a tag to compare
Pre-release
0.7.2

Roll back _hdbscan_linkage to match 0.6.4 and remove some dead code.

Bugfix labelling

26 Feb 03:39
Compare
Choose a tag to compare
Bugfix labelling Pre-release
Pre-release

A very minor performance regression to get this fixed, but it eliminates segfaults on Linux, so is necessary.

Fixing Boruvka Issues and Dimensional Scaling

22 Feb 15:45
Compare
Choose a tag to compare

Boruvka scales poorly with dimension; by approximating the minimal spanning tree we can achieve far better scaling with dimension at the cost of slightly less accurate clustering. In testing the loss of accuracy seems very small, so I'm pushing it out with that defaulted to on.