You are here

Uncertainty Quantification for Learning and Data-Driven Predictive Modeling of Complex Systems

Danial Faghihi, University at Buffalo
Alireza Tabarraei, The University of North Carolina at Charlotte
Kathryn Maupin*, Sandia National Laboratories
Prashant K. Jha, The University of Texas at Austin
Peng Chen, Georgia Institute of Technology
Recent advances in computational science have resulted in the ability to perform large-scale simulations and process massive amounts of data obtained from measurements, images, or high-fidelity simulations of complex physical systems. Harnessing such large and heterogeneous observational data and integrating those with physics-based and scientific machine learning models have enabled advancing computational models' prediction capabilities.
This mini-symposium highlights novel efforts to develop predictive computational models and model-based decision-making. It provides a forum for advancing scientific knowledge of data-driven complex system modeling and discussing recent uncertainty quantification (UQ) developments in physics-informed scientific machine learning and data interpretation algorithms. Potential topics may include but are not limited to efforts on:
  • Bayesian validation and selection of computational models
  • UQ analyses of high-fidelity discrete (molecular dynamics, agent-based) models
  • Physics-informed machine/deep learning
  • Data-driven discovery of physical laws
  • The interface of UQ and scientific machine learning
  • Design, control, and decision-making under uncertainty
  • Integrated multi-scale modeling and image analyses
  • Computational imaging
  • Operator inference for model reduction and surrogate modeling
  • Learning from high-dimensional and uncertain data
  • Multi-level, multi-fidelity, and dimension reduction methods
  • Learning the structure of the high-fidelity physics-based model from data
  • UQ methods for stochastic models with high-dimensional parameter space
  • Scalable, adaptive, and efficient UQ algorithms
  • Extensible software framework for large-scale inference and UQ
*Sandia National Laboratories is a multimission laboratory managed and operated by National Technology & Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525.