Previous talks at the SCCS Colloquium

Michael Grad, Thomas Raith: Efficient Numerical Solvers for the Manifold Learning Framework datafold

SCCS Colloquium |


As part of an interdisciplinary project in cooperation with the faculty of Mathematics, our task was to implement an efficient Eigenproblem solving path for sparse matrices into the datafold framework.
Therefor we utilized the PETSc and SLEPc frameworks to achieve some run time improvements. In order to increase PETSc's and SLEPc's ease of use in our use case, we set up a Docker environment which allows deployment regardless of host platform and takes over the tedious tasks of collecting, installing and the setup of its dependencies.
Finally, we present the results of our Benchmarks that were performed both, on consumer grade machines and the LRZ's Linux Cluster.

Keywords: SLEPc, PETSc, Eigenvalue-Problems, Datafold

IDP submission talk. Michael & Thomas are advised by Dr. Tobias Neckel and Dr. Felix Dietrich.