Should we "implement" RowVecs and ColVecs from KernelFunctions.jl?

[davibarreira]

Once we make the discretemeasure function available to users, they will have to use:

using KernelFunctions
# rows correspond to samples
μ = discretemeasure(RowVecs(rand(7,3)), normalize!(rand(10),1))

# columns correspond to samples, each with equal probability
ν = discretemeasure(ColVecs(rand(3,12)))

So I think it would be nice to have ColVecs and RowVecs available inside our package.

[devmotion]

I opened an issue in KernelFunctions: JuliaGaussianProcesses/KernelFunctions.jl#394 I think we should be able to refactor them into a separate package quite quickly if we push it a bit 🙂