Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore the intersection of machine learning, physics-based modeling, and uncertainty quantification in this comprehensive seminar presented by Professor Michael Shields from Johns Hopkins University. Discover how uncertainty quantification (UQ) and machine learning (ML) are transforming physics-based computational modeling, particularly with the emergence of scientific machine learning (SciML) and physics-informed ML approaches. Learn about the critical role of UQ in addressing epistemic and aleatory uncertainties present in physics-based models, including parameters, inputs, excitations, and model forms themselves. Understand the distinction between "UQ for ML" and "ML for UQ" and their respective contributions to modern computational modeling paradigms as the field advances toward scientific foundation models with broad generalizability. Examine recent advances in UQ/ML including neural operators for multi-resolution and multi-fidelity stochastic problems, scalable Bayesian neural networks, and neural operators designed for uncertainty quantification in foundation models. Investigate alternative architectures to neural networks such as polynomial chaos expansions and Gaussian processes that prove beneficial for UQ applications. Review diverse applications spanning multi-scale materials modeling to high energy-density physics, demonstrating the wide-ranging impact of these methodologies across scientific domains.
