ONNX Probabilistic Programming Working Group – Collaboration with the R-INLA Community

[brian....@gmail.com]

Hello everyone,

I wanted to share that the ONNX Probabilistic Programming Working Group has recently been formed and invite participation from the R-INLA community.

The goal of this working group is to bring probabilistic modeling and Bayesian inference into the ONNX ecosystem as first-class capabilities, enabling probabilistic models to be represented in a portable intermediate format and executed across frameworks and hardware environments.

ONNX has traditionally been used to represent neural network models, but the scope of the ecosystem is expanding to support uncertainty-aware models and probabilistic inference workflows.

Areas we are exploring

The working group is currently exploring several areas relevant to probabilistic computation:

• Probability distributions and log-density operators

• Bijectors and constrained parameter transformations

• Reproducible stateless RNG semantics for parallel inference

• Special mathematical functions used in Bayesian computation

• Inference algorithms such as Laplace, Pathfinder, INLA, HMC, NUTS, and SMC

• Export pathways for probabilistic programming frameworks

Frameworks we are looking to support

The goal is to support a range of probabilistic modeling ecosystems, including:

• Stan

• PyMC

• Pyro

• NumPyro

• TensorFlow Probability

• JAX-based probabilistic systems

• BayesFlow

• Julia probabilistic programming frameworks (Turing.jl, RxInfer.jl)

• INLA-based workflows

Why input from the R-INLA community matters

R-INLA represents one of the most important approaches to approximate Bayesian inference, particularly for latent Gaussian models and spatial statistics.

As the working group explores support for INLA-style inference operators and model representations, feedback from the R-INLA community will be extremely valuable, especially around:

• Representation of latent Gaussian models

• Graph structures and precision matrices

• Numerical stability of Laplace approximations

• Practical deployment of INLA-based models

Getting involved

If you're interested in participating, contributing ideas, or providing feedback from the R-INLA perspective, please feel free to reach out to:

• Andreas Fehlner https://www.linkedin.com/in/andreas-fehlner-60499971/ 

• Adam Pocock https://www.linkedin.com/in/craigacp/ 

• Brian Parbhu https://www.linkedin.com/in/brian-parbhu-99891133/ 

You are also welcome to attend the working group meetings:

🗓 Fridays @ 12 PM EST, every two weeks

Working group repository:

https://github.com/onnx/working-groups/tree/main/probabilistic-programming

We would very much welcome insights from the R-INLA community as we explore how probabilistic inference methods can be represented in portable computational formats.