A Randomized Proper Orthogonal Decomposition Method for Reducing Large Linear Systems

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.