Uncertainty Quantification for High-Dimensional Problems

In the second half of 2024 CWI's Scientific Computing group will organise a semester programme on uncertainty quantificaton. Co-organisers are: TU Eindhoven, Radboud University, TU Delft, and University of Amsterdam/University College London.

Overview

Uncertainty quantification (UQ) aims at establishing the reliability of predictions made by computational models. Such models are often fully deterministic, yet are used to make predictions of an uncertain reality. The discipline of UQ consists of numerical and statistical methods for quantifying uncertainty in real problems due to both epistemic uncertainty (lack-of-knowledge) and aleatoric uncertainty (true randomness) in physical and model parameters, boundary and initial conditions, model form, or even uncertainty caused by finite-precision floating point calculations. UQ also offers a mathematical framework for exploiting experimental data to inform models (data-assimilation) and solving inverse problems. However, the applicability of UQ solution concepts is challenged by the cost of a single model evaluation, the high dimensionality of the (parametric-and data-) spaces in which UQ problems are often posed (curse of dimensionality), and the resulting high-dimensional, dependent probability distributions that must be represented.

Aim

In this context, the aim of the proposed program is to introduce the topic of UQ to a wide range of researchers, and to bring together both national and international UQ experts. We intend to invite a mixture of colleagues from industry, and academia, and we also want to have a combination of young, and more experienced researchers.

Topics

  • (High-dimensional) forward uncertainty propagation
  • Inverse UQ and data assimilation
  • UQ software and HPC scalability

Events

7-11 October - AUTUMN SCHOOL

Link eventpage to see the tentative programme

11-15 November - WORKSHOP

Throughout the program - Hybrid seminars with international UQ experts

  • The dates for the events will be announced in the course of 2024.

For more information, please contact one of the organizers: Olga Mula, Laura Scarabosio or Wouter Edeling.

Programme WORKSHOP 11-15 November

Program workshop RSP 11-15 November 2024

Here you can download the abstract booklet of the workshop.