David Widmann
David Widmann
It seems Distances uses `pairwise!(out, metric, a[, b])` whereas the docstring in StatsAPI and the implementation in StatsBase use `pairwise!(f, dest, x[, y])`.
https://github.com/JuliaStats/Distances.jl/blob/44036b573ec85287022f4368c4e1e279698bd031/src/mahalanobis.jl#L23 and probably also `dot_percol` could benefit from the 3-argument method `dot(x, A, y)`. However, since this version was only introduced in Julia 1.4, Compat >= 3.2.0 would be required...
The total variation distance for probability measures on discrete spaces can be defined as ``||p-q|| = ||p-q||₁ / 2`` and hence is implemented as the cityblock distance divided by 2...
Uses `xlogy` and `xlog1py` now which take care of the special case `p == 1` automatically. Moreover, for improved numerical accuracy for all cases where `k == 0` (and a...
The PR adds a fast path for `loglikelihood(::Uniform, ::AbstractArray{ using Distributions, Zygote, BenchmarkTools julia> a = rand(); julia> b = a + rand(); julia> x = rand(Uniform(a, b), 10_000); julia>...
This PR - ~~simplifies the implementation of `convolve(::MvNormal, ::MvNormal)`,~~ - adds a Unicode alias `⊕`, - ~~adds documentation for `convolve` and `⊕` (Preview: https://juliastats.org/Distributions.jl/previews/PR1445/convolution/).~~ It seems `convolve` is difficult to...
Since there is no consensus yet on the Unicode alias for `convolve` yet in https://github.com/JuliaStats/Distributions.jl/pull/1445, I put together this PR with an alternative suggestion. This PR contains the same changes...
**Scope of the PR** The PR tries to address *some* of the problems with `rand` and `eltype` - but it does not intend to solve *all* inconsistencies and problems. I...
@oxinabox reported on Slack that https://github.com/JuliaStats/Distributions.jl/pull/1437 caused method ambiguity errors in an internal package. This PR reduces the amount of default fallback implementations and hence should make method ambiguity errors...
I am aware of https://github.com/JuliaStats/Distributions.jl/pull/823 but I think maybe it is more promising to start with a slightly simpler approach and add keyword argument constructors step by step. This PR...