geometry

Current behavior:

root [0] TFile f("root://eospublic.cern.ch//eos/root-eos/benchmark/Run2012B_SingleMu.root")
Error in <TFile::TFile>: file /eos/root-eos/benchmark/Run2012B_SingleMu.root does not exist
(TFile &) Name: /eos/root-eos/benchmark/Run2012B_SingleMu.root Title:

Desirable behavior:

root [0] TFile f("root://eospublic.cern.ch//eos/root-eos/benchmark/Run2012B_Single

Right now converted meshes approximate curves with a given resolution. Ideally this resolution could be configured as users might want more detail on the converted mesh.

Since mesh converting is currently based on the bezier converter this simply boils down to setting the correct values in the bezier datablock before the to_mesh() method is called. Note that setting this property by the user doe

Add it here: https://github.com/georust/geo/blob/1d3b32970a49746de6ce2400305725edd21f7c15/geo/src/lib.rs#L124-L130

Where should the content be modified?

geomstats/learning/knn.py

What should be modified?

We should come up with a "bridge" to scikit-learn, i.e. a centralized way to pass geomstats' Riemannian geodesic distances to scikit-learn learning methods. Not only for the K-NN algorithm.

Additional information

Example snippet that is not working:

import geomstats.datasets.utils as dat