Distributed Eigenvalue Computation

Betreuer: Maik Röper
Art der Arbeit: Projekt (MSc)
Ausgabe: 03/2020
Bearbeiter: Sheena Moin
Status: in Arbeit

In the context of digitial signal processing, the calculation of eigenvalues and their respective eigenvectors of some matrix A is of great interest. To compute them for large matrices, a parallel implementation of eigenvalue algorithms can reduce the computational time.

For distributed implementation, the processing units have no acces to a central memory. Therefore they have only local information about the matrix A available and need to exchange information between each other.

The main question for this work are

  • What kind of information have to be available at each processing unit to compute the eigenvalues and their respective eigenvectors in a fully distributed way?
  • Is it possible to compute only the largest/lowest eigenvalue of A with less information exchange?
  • Are distributed eigenvlaue algorithms suitable for distributed precoding/beamforming?

Requirements: Basic Matab programming skills, knowledge in linear algebra


[1] F. Penna and S. Stanczak, "Decentralized Eigenvalue Algorithms for Distributed Signal Detection in Wireless Networks"
[2] J. Mohammadi, S. Limmer and S. Stanczak, "A Decentralized Eigenvalue Computation Method for Spectrum Sensing Based on Average Consensus"

Zuletzt aktualisiert am 03.03.2020 von M. Röper
AIT ieee GOC tzi ith Fachbereich 1
© Arbeitsbereich Nachrichtentechnik - Universität BremenImpressum / Kontakt