Kai Xu

Opned https://github.com/JuliaLang/julia/issues/33361.

PS: the GPU support of `pairwise` is based the branch of this PR https://github.com/JuliaStats/Distances.jl/pull/142

I tried it on macOS and it works. I simply download Visual Studio then build & run this project with the command described in the README file.

> @datnamer > > > What's needed for GPU support? > > I am not sure what you mean. In doubt, I'd say generic containers everywhere instead of hardcoded Float64...

Yes I believe this is one way to say it.

I'm probably not having the best practice here but I overwrote the signature so that I can use `rand(rng)` for a vector of RNGs at https://github.com/TuringLang/AdvancedHMC.jl/blob/master/src/utilities.jl#L5. The intension is that...

Yes, a vector of RNGs is a separate RNG per chain. The `rand_coupled` is a seperate concern and only used here: https://github.com/TuringLang/AdvancedHMC.jl/blob/692e646f66307e4d35627b57904e7fe04bb35234/src/trajectory.jl#L302 Basically, in vectorized mode, there are also some...

I guess in principle we can implement extend the jitter operation to support different noise and implement a specific RNG for Halton sequence. But I feel it might be simpler...

Riemann HMC (https://github.com/TuringLang/AdvancedHMC.jl/issues/6) is on the TODO list but is of low priorty at the moment. If you are interested in implementing it, please let me know.

Thanks for the pointer. There is already an [issue](https://github.com/TuringLang/AdvancedHMC.jl/issues/19) for this work. I plan to take a closer look on a few newer HMC variants and implement them later this...