Hosted by Lisa Kusch

Abstract

Scaling up computational algorithms efficiently remains a key challenge in scientific computing. While modern approaches, such as machine learning, offer new possibilities, they often struggle with high computational costs and efficiency limitations. The first part of the talk explores how classical algorithmic strategies can help overcome these barriers and bring complex models to scale.

In the second part, we focus on Maxwell’s equations, which describe electromagnetic phenomena and are essential in fields like imaging and communications. Solving these equations efficiently requires techniques that balance robustness and scalability. By leveraging established techniques, such as domain decomposition and preconditioning, we can develop solutions that are both computationally efficient and robust w.r.t geometric features and physical parameters.