David Widmann
David Widmann
Just ask if anything is unclear or doesn't work 🙂
You have to use a `Gibbs` sampler if the variables don't have a common Gaussian prior.
I don't think this should be done in this PR. In general, I'm not sure if it's a good idea at all to define addition for `Cholesky` at all but...
Maybe the same error as in https://github.com/invenia/KeyedDistributions.jl/issues/24?
Changing https://github.com/JuliaStats/PDMats.jl/blob/15fecad14e700e024d359733d101c992707e981b/src/pdiagmat.jl#L4 to `struct PDiagMat{T
> PDMats is not at fault given it can also happen without PDMats. Also the other (well, actually same) issue in KeyedDistributions is caused by the change of the type...
> This seems like a strange error. Doesn't help but I've seen these quite often. > I tried to reproduce with my own type, but couldn't: A major difference is...
The usual approach for fixing these errors is defining an rrule or a projector with CR, as Will discussed in the linked PR.
> I started with `det(PDMats(x))` because it is throwing the same error as the more complicated example involving matrix division. Probably one should restrict `rrule(::typeof(det), ...)` in the same way...
> Regardless of AD, I think it would be useful to add definitions of `det(::AbstractPDMat) = ....` since otherwise these calls will fall back to the generic definitions based on...