Markov Chain Monte Carlo methods are like cooking milk you better watch what's happening! In order to do this we propose two complementary approaches: Turkie.jl based on Makie.jl and TuringCallbacks.jl based on Tensorboard.
Both package allow you to finely pick which variables and which statistics to watch for and display the results in real time while sampling.
Théo Galy-Fajou
Hi! I am a PhD student at TU Berlin and Julia adept since 0.4! I develop packages for probabilistic inference and more especially Gaussian Processes. For more info check my website theogf.github.io