search

TutteInstitute / fast_hdbscan

A fast multi-core implementation of HDBSCAN for low dimensional Euclidean spaces
BSD 2-Clause "Simplified" License
78 stars 8 forks source link

Define 'low dimensional' in README.txt #2

Closed bmschmidt closed 1 year ago

bmschmidt commented 1 year ago

Exciting project, thanks for sharing! I wonder if in the readme.txt you could put some broad numbers on what you mean by this being optimized for 'low dimensional data.' Sometimes I see people use that to mean 2-4 dimensions, sometimes up to 128 or beyond--would help in thinking about where I might use this.

lmcinnes commented 1 year ago

Good point. I have clarified the README with some more precise numbers. Right now it is built with KD-Trees so only really handles KD-Tree style workloads. In the future I may add other indexes that support higher dimensions better.

bmschmidt commented 1 year ago

Thank you!