Closed ParadaCarleton closed 1 year ago
Distributions is not needed for the example, a bit more minimal and clearer:
julia> PDMat(z)
3×3 PDMat{Float64, Matrix{Float64}}:
1.87205 3.20828 -0.993003
3.20828 5.97114 -1.88158
-0.993003 -1.88158 0.597444
I'm not sure if it's an actual issue with PDMats though, the main problem seems to be
julia> using LinearAlgebra
julia> cholesky(z)
Cholesky{Float64, Matrix{Float64}}
U factor:
3×3 UpperTriangular{Float64, Matrix{Float64}}:
1.36823 2.34484 -0.725759
⋅ 0.687651 -0.261456
⋅ ⋅ 0.0485701
As an example:
This makes it impossible to use DimensionalData.jl or other array wrappers (such as CuArrays) together with PDMats.jl.