Dynamical low-rank approximation (Emil Kieri - University of Bonn)

16.05.2017 14:15

In low-rank approximation, separation of variables is used to reduce the amount of data in computations with high-dimensional functions. Such techniques have proved their value, e.g., in quantum mechanics and recommendation algorithms. In this talk, we consider low-rank techniques for time-dependent problems. We study properties of the low-rank approximation space, and what consequences they have for time integration schemes. We investigate theoretically and experimentally the convergence of time-stepping schemes of two types: We consider explicit and implicit projection methods, and the so-called projection-splitting integrator which is developed explicitly for this type of problems.

Lieu

salle 624, Séminaire d'analyse numérique

entrée libre