Open maawoo opened 4 months ago
In comparison to the new version of the calculation, the old version seems to be more performant and works:
I can do a PR if there are no other plans for improvement or ideas for fixing the newer version.
Ping @omad
Hi all, first off, great job on the rust implementation of
geomedian
. It's impressively fast! 🙂 I did a quick test between the implementation in version0.2.4
of the package and theadd-rust-geomedian-impl
branch. Shape of my test dataset is(time: 5, y: 1311, x: 3704)
and the average absolute differences for each variable are quite small, however, thecount
variable stands out and has not been computed as expected (see second screenshot):Is this a known issue because the implementation has not been finished yet?