Graduate Student / Postdoc Seminar

---

Inverse Problems Governed by Large-Scale PDE Models

Speaker: Georg Stadler

Location: Warren Weaver Hall 1302

Date: Friday, November 7, 2014, 1 p.m.

Synopsis:

Inverse problems are a framework to convert measurements into information about a system, which cannot be observed directly. They are at the base of medical imaging, play an important role in weather prediction and in the detection of airplanes and submarines, and are the reason why we know a lot about how our planet looks inside. If the underlying mechanism is mathematically described by a partial differential equation, these problems can become computationally very challenging. As an example, I will show results for a time-domain seismic inverse problem. Due to the size of this problem, a parallel wave simulation code is used, and I will discuss some of the scalability, parallelism, and high-performance computing issues underlying this implementation.