Michiel Stock

> Last year, some work was started on `kron!` ([JuliaLang/julia#31069](https://github.com/JuliaLang/julia/pull/31069)) that computed the Kronecker product in-place. > > * Should this implementation by used? If so, should we wait for...

> > This is good idea. It is simple to either > > > > * update the vec trick, such that vector products with Kronecker products of two Kronecker...

Ok, I will summarize the discussion (as I see it) - we need a `collectin!(C, K)` function that collects GeneralizedKroneckerProducts in `C`, this should work for all kinds of Kronecker...

The question might be what is fixed and what you might want to compute the derivative of. I originally conceived Kronecker to work with systems as `f(K * w)` where...

Hmmm, I don't really know this product but would it not possible to directly reproduce it using block matrices and the dot syntax, for example: ```julia A = [rand(2,2) for...

I am sorry, it slipped through the cracks. Could you quickly resolve the conflict and then I run the tests and merge.

The documentation indeed does not seem to match the package anymore. I could make the basic examples work (noting that you only need to use `rand` instead of `randsample`). I...

The issue is that Distributions seems to use PDMmat for the covariance matrix. It would seem useful to be able to use the Matrix normal then. To avoid type piracy,...

It might, but the problem is that `M⊗1`would be a `Matrix` rather than a `KroneckerProduct`, which would be a type instability. I have implemented this in #138

HI, I am afraid I haven't experimented with CUDA.jl and Kronecker.jl. I assume they would work with some minimal adjustments?