New PyMCon Talk: Enabling Uncertainty Quantification by Anne Reinarz & Linus Seelinger

Welcome to the 13th and 1st PyMCon Web Series events of 2024!


  1. Anne Reinarz, assistant professor of Computer Science at Durham University in the Scientific Computing Group. She is the incoming Director of the Durham MSc in Scientific Computing and Data Analysis.
  2. Linus Seelinger, a postdoctoral researcher at Karlsruhe Institute of Technology, holds a fellowship with KIT’s Young Investigator Group Preparation Program.

Event type: Recorded Talk with Live Q&A
Async Talk:
Code: Tutorial — UM-Bridge documentation
Q&A -1: 2024-01-29T15:00:00Z (RSVP Meetup event)
Q&A -2: 2024-01-30T06:30:00Z (RSVP Meetup event)
Website: PyMCon Events · PyMCon Web Series


1. Intro to Introduction to Uncertainty Quantification (UQ) and UM-Bridge

2. Hands-on Tutorial:

:memo: We highly recommend following along with the tutorial on UM-Bridge docs:

:point_right: Presentation Slides: tutorial.pdf - Google Drive

Treating uncertainties is essential in the design of safe aircraft, medical decision making, and many other fields. UM-Bridge enables straightforward uncertainty quantification (UQ) on advanced models by removing technical barriers.

Complex numerical models often consist of large code bases that are difficult to integrate with UQ packages such as PyMC, holding back many interesting applications. UM-Bridge is a universal interface for linking UQ and models, greatly accelerating development from prototype to high-performance computing.

This hands-on tutorial teaches participants how to build UQ applications using PyMC and UM-Bridge. We cover a range of practical exercises ranging from basic toy examples all the way to controlling parallelized models on a live cloud cluster. Beyond that, we encourage participants to bring their own methods and problems.

Why UM-Bridge is needed:
The main idea is to make UQ more widely accessible and accelerate the development of UQ methods by establishing:

  • A new software architecture for UQ applications based on a universal interface between UQ and models.
  • A library of ready-to-run benchmark problems based on that interface.

Our approach closes the gap between advanced UQ methods and advanced models by removing the technical hurdle of integrating complex software stacks:

  • Applying advanced UQ methods on complex models becomes trivial from a technical perspective.
  • Collaboration between UQ and model experts is accelerated through the separation of concerns.
  • Application communities (geophysics, engineering, etc.) gain easy access to state-of-the-art UQ methods.
  • Reproducibility becomes considerably easier and more widespread through containerized models.
  • Scaling to HPC clusters for solving computationally challenging problems becomes much simpler.

If you are interested, you can read more on the interface and the benchmarks here:


The UM-Bridge Tutorial has finally here! :rocket:

:point_right: Part - 1: Introduction to Uncertainty Quantification (UQ) and UM-Bridge:
:point_right: Part - 2: Hands-on Tutorial:

:memo: We highly recommend following along with the tutorial on UM-Bridge docs:

:point_right: Presentation Slides: tutorial.pdf - Google Drive