Manifolds.jl

Manifolds.jl copied to clipboard

Docs tutorial, "How to work with Rotations"

21

Hi @mateuszbaran , @kellertuer, Please find a new tutorial and comment as necessary, a new tutorial on `SpecialOrthogonal(2)`. My idea here was to collect the most important functions and type...

dehann

Multilinear operators

25

I've started some work to solve https://github.com/JuliaManifolds/ManifoldsBase.jl/issues/45 . It's far from complete but the general idea is to have a `AbstractTensorField` type that represent arbitrary partial functions between vector bundles,...

mateuszbaran

Inconsistency in invariant metrics

2

Currently invariant metrics on groups are somewhat inconsistent: `inner` continues dispatch with traits without unwrapping the manifold while `exp` and `log` unwrap and run trait dispatch from scratch. This should...

mateuszbaran

local_jacobian dimension bug.

2

I am currently taking a look at our differential stuff – since we nearly got AD in the embedding working already I stumbled upon our Jacobins and noticed, that they...

kellertuer

It would be nice to have some information-geometric metrics. * [ ] [Multivariate Normal Distributions Parametrized as a Riemannian Symmetric Space](https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.61.4046&rep=rep1&type=pdf)

mateuszbaran

An Interface and first steps towards a Riemannian Gradient and a Riemannian Hessian

29

This thread -- as a continuation of the discussion started in #27 -- shall collect ideas and approaches to the Riemannian Gradient (for example employing tools like the `egrad2rgrad` from...

kellertuer

Described here: https://www.aimsciences.org/article/doi/10.3934/jgm.2020031

mateuszbaran

Elementwise distance

2

A function for calculating elementwise distance between points might be useful. Something like this for example: ```julia function elementwise_distance(M::AbstractPowerManifold, p, q) dists = similar(p, number_eltype(p), power_dimensions(M)) return elementwise_distance!(M, dists, p,...

mateuszbaran

Provide default basis for `local_metric` functions and make the basis parameter hence optional

5

Currently `local_metric(M, p, B)` requires a basis (similarly `det_local_metric`, `inverse_local_metric` and `jacobian_local_metric`). It would be nice (and not too difficult) to provide a reasonable default for this last parameter, e.g....

kellertuer

Taking AD seriously

10

I think I would like to get into a proper way of handling AD on manifolds. I know we have quite some issues open here (#17, #42, #44, #88, #329)...

kellertuer