Turns out this has major performance issues on linux on the latest sklearn versions since auto makes some bad choices. There should be a noticeable improvement in step times on aiarena ladder
The algorithm choice of kd_tree has been well tested in Eris (despite not using sharpy it does use this very same function), and fairly certain Zoe has used this option for some time too.
The default algorithm option for
DBSCAN
insklearn
isauto
https://scikit-learn.org/stable/modules/generated/sklearn.cluster.DBSCAN.htmlTurns out this has major performance issues on linux on the latest
sklearn
versions sinceauto
makes some bad choices. There should be a noticeable improvement in step times on aiarena ladderThe algorithm choice of
kd_tree
has been well tested in Eris (despite not using sharpy it does use this very same function), and fairly certain Zoe has used this option for some time too.