Support for kernel mean embeddings and kernels on distributions

[IsakFalk]

My work involves embedding distributions in RKHS using kernel mean embeddings or equivalently kernels on distributions Ch. 2, especially the generalized RBF kernel.

Are there any plans of extending this library to this case (or other structured domains for that matter)?

In the near future I'll be implement algorithms that uses this and if this sounds interesting I'd be happy to clean up and contribute the resulting code!

Thanks for the hard work :)

[trthatcher]

Hey, thank you!

Unfortunately I am not planning on extending this library any further - my attention has shifted to other things so I will be archiving this repo soon. However, Théo Galy-Fajou has taken some of this code as a basis and rebuilt things in a way that is more amenable to auto-differentiation and non-numeric kernels. His project is the go-forward solution... I would open up this issue over in his project and discuss with him:

https://github.com/theogf/KernelFunctions.jl

Looks exciting though! I'm interested to see what you come up with.