The Design of Turing.jl

📆 Download Calendar Event: Learn Bayes - The Design of Turing.jl

Type: SeminarSpeaker: Penelope YongDate & time: Thu May 15 2025, 13:00-14:00 (CEST)Place: Zoom

This is an online event. Zoom link will be provided here and through the email list.

In this seminar, Penelope will talk about the design Turing.jl and compare it to other probabilistic programming frameworks like Stan and PyMC.

Penelope Yong is a software engineer working on Turing.jl at the Alan Turing Institute.

In her talk, Penelope plans to cover the following points:

Modeling the classic ‘eight schools’ problem using Turing.jl, Stan, and PyMC
What Turing.jl brings to the table in Bayesian modeling, with a particular focus on composability with the broader Julia ecosystem
A discussion of some factors which might lead a user to gravitate towards one probabilistic programming framework
Reflections as a software developer on the challenges that Turing.jl (and open source software in general) face

For epidemiologists and biostatisticians, adopting Turing.jl offers several key advantages:

Speed and scalability: Julia’s performance rivals that of C, making Turing.jl well-suited for large-scale simulation and inference tasks.
Expressiveness: Define complex, custom models — including those with non-standard likelihoods or intricate dependencies — without sacrificing clarity or flexibility.
Interoperability: Leverage the rich Julia ecosystem, including data manipulation, plotting, and differential equation solvers.
Reproducibility and openness: As a fully open-source toolchain, Turing.jl encourages transparent, shareable, and extensible science.

For an in-depth look at Turing.jl, its design and applications, check out the recent publication by Fjelde et al. (2025).

Coming from R or Python? Julia has great interfaces to these languages with RCall.jl and PythonCall.jl, allowing you to seemlessly integrate your existing workflows with Julia!