Efficient Model Order Reduction for Multiscale Kinetic Equations
Jingwei Hu
University of Washington
Applied Mathematics
Many model order reduction techniques, such as projection-based reduced-order methods and dynamical low-rank methods, rely on Galerkin projection to map high-dimensional objects onto low-dimensional manifolds. This process is often computationally expensive, especially when nonlinearities and multiple scales are involved. Using a simple kinetic equation from plasma physics as an example, we demonstrate that, by leveraging the special structure of the problem, one can design efficient model order reduction methods for both fast on-the-fly simulations and parametric equation solvers.