trappmartin / Turkie.jl

Turing + Makie = Turkie

Turing + Makie -> Turkie!

WIP for an inference visualization package.

To plot during sampling :

• Trace of the chains
• Statistics (mean and var)
• Marginals (KDE/Histograms)
• Autocorrelation plots

Additional features :

• Selecting which variables are plotted
• Selecting what plots to show
• Giving a recording option
• Additional fine tuning features like
  • Thinning
  • Creating a buffer to limit the viewing

Extra Features

• Using a color mapping given some statistics
• Allow to apply transformation before plotting

Usage:

Small example:

using Turing
using Turkie
@model function demo(x)
    v ~ InverseGamma(3, 2)
    s ~ InverseGamma(2, 3)
    m ~ Normal(0, √s)
    for i in eachindex(x)
        x[i] ~ Normal(m, √s)
    end
end

xs = randn(100) .+ 1;
m = demo(xs);
ps = TurkieParams(m; nbins = 50, window = 200) # default behavior : will plot the marginals and trace of all variables
cb = TurkieCallback(ps) # Create a callback function to be given to sample
chain = sample(m, NUTS(0.65), 300; callback = cb)

If you want to show only some variables you can give a Dict to TurkieParams :

ps = TurkieParams(Dict(:v => [:trace, :mean],
                        :s => [:autocov, :var]))

If you want to record the video do

record(cb.scene, joinpath(@__DIR__, "video.webm")) do io
    addIO!(cb, io)
    sample(m,  NUTS(0.65), 300; callback = cb)
end