ETNA 25
Title slide
Abstract
In this presentation we revisit the approximate rational Krylov method. We present two alternative but mathematically equivalent formulations of the same algorithm. The first reformulation uses a pole swapping technique and is an implicit method, just like the original algorithm. The second reformulation explicitly solves shifted linear systems using the Arnoldi Hessenberg matrix. This reformulation leads us to a connection between the approximate rational Krylov method and the full orthogonalization method (FOM). Finally, we show how the approximate rational Krylov method can be modified to obtain a similar connection with GMRES.
Daan Camps
Researcher in Advanced Technologies Group
My research interests include quantum algorithms, numerical linear algebra, tensor factorization methods and machine learning. I’m particularly interested in studying the interface between HPC and quantum computing.