Seth Axen

I think for Julia PPLs it would be best for each model to be defined in a Julia script that would include 1. any necessary package imports 2. any necessary...

FWIW, there's another ℝⁿ⁻¹ → ℝⁿ implementation in [HMCUtilities.jl](https://github.com/salilab/HMCUtilities.jl/blob/master/src/constraint.jl#L540-L654).

Fair point. My ultimate goal is to use directional/angular distributions in Turing, at least the two in Distributions, but ideally arbitrary ones. If there's a better way to do that...

Well, it all depends on what we want the transformation for. My assumption due to my need is that we are seeking a transformation `f: x ↦ y` for `x...

I don't think it's terribly relevant, but on the off-chance it's useful, here's some early work on adding maps to Manifolds.jl, ultimately with the goal of pushing forward/pulling back objects...

> Looks good to me! I’m not sure why some CIs are broken. Oh, this actually does not currently work. As noted in the top comment, I'm not even certain...

> I don't know if this will have any negative impact on sampling. After further thought, yes, this is detrimental to sampling with NUTS, because it makes it very unlikely...

As demonstrated in https://discourse.julialang.org/t/case-study-speeding-up-a-logistic-regression-with-rhs-prior-turing-vs-numpyro-any-tricks-im-missing/87681/34, this impropriety also introduces post-warmup numerical error. A quick-and-dirty hack to get this working for Turing users might be to edit https://github.com/TuringLang/Bijectors.jl/blob/b20471252c01dd4832e06aa80045046483f3804e/src/bijectors/corr.jl#L83-L95 to add ```julia for...

Maybe part of what is going on here is that the output type of `^` is tightly linked to how close to an integer the power is. e.g. ```julia julia>...

Actually, I'm not convinced anymore that this is a bug. This works ```julia julia> j′vp(fdm, p -> complex(x)^real(p), complex(seed), complex(3.0)) (487.58111835736804 - 4.440892098500626e-14im,) julia> j′vp(fdm, p -> complex(x)^imag(p), complex(seed), complex(3.0))...