Mateusz Baran

Thanks for asking! I think `Manifolds.jl` isn't quite ready yet. A few things we need to consider before using it in things like `Optim.jl`: 1. Stiefel manifold isn't done yet...

Thanks, we are trying to make this package as good as we can :slightly_smiling_face: . On my computer it takes 24 seconds to load Manifolds, that's way too much for...

We extend their `VariateForm` and `ValueSupport` as well as derive from `Distribution`: https://github.com/JuliaNLSolvers/Manifolds.jl/blob/master/src/DistributionsBase.jl .

Right, that sounds reasonable. I'll add separating `ManifoldsBase.jl` to my TODO list :slightly_smiling_face: .

I've looked at this again and essentially the last missing piece is gradient calculation. `Optim.jl` currently just projects Euclidean gradient to the tangent space but it doesn't work for some...

Cool idea but currently there may not be enough content for a homepage I think. Maybe we could also put there some examples and how-tos?

Some parts of FunManifolds.jl are usable but it's far from being polished, so these three are the only ones that I'd advertise as ready for general usage. And maybe we...

Things like `det_local_metric` are not sensitive to change of basis so it would make sense to add a method that doesn't take a basis but overall this seems like a...

I personally wouldn't do any restructuring around bases without solving that ManifoldsBase issue first, at least the part relevant to bases.

I think reworking this is a good idea. We probably should replace `exp` with default retraction. One more thing, it isn't guaranteed that any arbitrary basis `B` would be a...