Nicolas Guigui

Do you have a code snippet? We have noticed this before if the matrices are not positive definite, and this is why we have PR #1362, I would like to...

What about writing a `divide` function for pytorch and tf with similar behavior as np? It is similar to using `where` to avoid dividing by 0, but unlike `where` it...

Hi Niels, sorry for the inconvenience. We have not released a new version of geomstats on pypi yet so reinstalling with pip won't get you the update of #1026. However...

Page 30 of [this paper](https://arxiv.org/pdf/physics/9806030.pdf) describes the different distances and how to define them. We need to implement each RiemannianMetric that defines them, in some cases it is just Frobenius...

The theory indeed should work in general, but the implementation is for diagonal metric only because of [`reshape_metric_matrix`](https://github.com/geomstats/geomstats/blob/78a5778b5d5ce85225fd97e765d43047fb4526d1/geomstats/geometry/invariant_metric.py#L64C9-L64C31) that is used to easily compute the inner product: https://github.com/geomstats/geomstats/blob/78a5778b5d5ce85225fd97e765d43047fb4526d1/geomstats/geometry/invariant_metric.py#L105