This is awesome because this is an honors thesis which means that this subject area is really not for specialists anymore, not even computer science folks (see previous tutorial) ! woohoo ! (implementations are available at the end of the document).
A Randomized Proper Orthogonal Decomposition Method for Reducing Large Linear Systems by Brad Marvin
The proper orthogonal decomposition (POD) method is a powerful tool for reducing large data systems which can quickly overwhelm modern computing tools. In this thesis we provide a link between randomized projections and statistical methods by introducing the randomized POD method. We also apply the POD method to a heat transfer finite element model and image compression. In doing so we demonstrate the practical use and quantify the error introduced by the POD method.
Liked this entry ? subscribe to Nuit Blanche's feed, there's more where that came from. You can also subscribe to Nuit Blanche by Email, explore the Big Picture in Compressive Sensing or the Matrix Factorization Jungle and join the conversations on compressive sensing, advanced matrix factorization and calibration issues on Linkedin.