Scientific Machine Learning Meets Uncertainty Quantification

Overview

The rapid rise of scientific machine learning (SciML) is transforming computational science by enabling efficient surrogate models and data-driven approaches for complex physical systems. At the same time, uncertainty quantification (UQ) remains essential for assessing the reliability, robustness, and risk associated with predictions and decisions derived from these models. As SciML methods are increasingly deployed in high-consequence scientific and engineering applications, there is a growing need to understand both how machine learning can accelerate UQ workflows and how uncertainty can be rigorously characterized within SciML itself.

This workshop will bring together researchers from applied mathematics, statistics, scientific computing, engineering, and machine learning to discuss recent advances at the intersection of SciML and UQ. Topics will include machine-learned surrogate models, neural operators, inverse problems, probabilistic methods, Bayesian inference, uncertainty-aware optimization, digital twins, and data assimilation. Particular emphasis will be placed on developing reliable and interpretable methodologies for scientific decision-making under uncertainty.