Support truncated distribution.

[turinglang]

Support truncated distribution.

[xukai92]

I've looked into Distributions.jl. This error is due to the fact that helper functions for Truncated still are not defined with generic types (e.g. they use pdf(d::Truncated, x::Float64) = ... for pdf (https://github.com/JuliaStats/Distributions.jl/blob/master/src/truncate.jl). I will overload them in Turing and submit a PR soon.

[xukai92]

Submitted a PR for Distributions.jl JuliaStats/Distributions.jl#586

This example below (same as above but changed β to the right support range) is passed on my local Distribution.jl fork (https://github.com/xukai92/Distributions.jl/tree/generic-support-to-truncated-dist).

using Distributions, Turing

negbindata = [0, 1, 4, 0, 2, 2, 5, 0, 1]

@model negbinmodel(y) begin
  local α, β
  α ~ Truncated(Cauchy(0,10), 0, +Inf)
  β ~ Truncated(Cauchy(0,10), 0, 1.0)
  for i = 1:length(y)
    y[i] ~ NegativeBinomial(α, β)  # α > 0, 0 < β < 1
  end
  return(α, β)
end

@sample(negbinmodel(negbindata), HMC(1000, 0.02, 1))

[yebai]

@xukai92
Is there anything that we can do to speed up the merging by Distributions package developers?

[xukai92]

Let me chase them on the PR

[xukai92]

Closed because PR is merged.