Karen Willcox (MIT)
Abstract. Model reduction has become a powerful approach for extracting low-order, cheap surrogate models from high-fidelity simulation models, such as those arising from discretization of systems of partial differential equations. Recent developments in combining traditional projection-based model reduction methods with machine learning methods have led to new approaches that tune and adapt reduced models in the face of changing system properties and dynamic data. This provides new opportunities to discover and learn models, informed by physics-based first principles but guided by data. This talk will provide a brief overview of the basics of model reduction and discuss some of our recent work in reduced model localization and adaptation. I will also offer some future directions in multi-information-source approaches that aim to create a principled strategy for managing high-fidelity models, reduced models, and data in a design setting.