Hankel-Norm Model Reduction

Abstract

In the previous chapters, we assumed that a given upper operator or matrix T has a computational model of a sufficiently low order to warrant the (possibly expensive) step of deriving its state realization. Once a state model is known, we showed how multiplication by T or its inverse can be done efficiently, using the model rather than the entries of T. We also derived some useful factorizations, such as the external and inner-outer (~ QR) factorization. A spectral factorization/Cholesky factorization result is given in chapter 13.