Uncertainty Quantification via Gaussian Processes and Beyond

Overview

Uncertainty Quantification (UQ) plays a central role in modern science and engineering by providing a principled framework for inference, prediction, and decision-making in complex systems. As these problems grow in scale and complexity, they require integrated advances across mathematics, statistics, machine learning, optimization, and high-performance computing. Gaussian processes (GPs) have become a key tool in this landscape due to their flexibility as nonparametric models and their ability to quantify uncertainty, with wide applications in various scenarios, such as computer experiments, spatiotemporal analysis, active learning, multi-task learning, model calibration, and Bayesian optimization.

This workshop will bring together researchers across statistics, mathematics, machine learning, and related fields to explore recent advances in GP-based UQ, with an emphasis on both theoretical foundations and practical methodologies for modeling complex systems and extracting reliable insights from data. The program will highlight four interconnected themes at the forefront of current research: (i) calibration of computer models using GPs; (ii) GP-driven approaches to optimization and experimental design; (iii) scalable GP computation; and (iv) sensitivity analysis and broader UQ techniques. By integrating these directions, the workshop aims to advance both the theory and practice of uncertainty quantification, while fostering collaboration and knowledge exchange across disciplines and application domains.