No label defined (Q4435199)

Since 1950, generalized inverses theory (involving Moore-Penrose inverses and least squares problems) has been widely studied. This book deepens research in matrix analysis and numerical linear algebra. Concretely, it presents a generalization to weighted least squares problems, that is: \[ \min_x\{(b-Ax)^HW(b-Ax)\}, \quad W= \text{diag} (w_1,\dots,w_m), w_i > 0 \] and constrained weighted least squares problems, that is: \[ \min_{x \in S}\{(Kx-g)^HW_2(Kx-g)\}, \quad\text{subject to}\quad Lx=h, \] which applies to mathematical programming and optimization, statistical linear modelling, fluid systems, electronic circuits, signal processing, etc. This book contains three parts: analysis of supremum of weighted pseudoinverses, study the stability of weighted pseudoinverses, weighted least squares problems and constrained weighted least squares problems, and stable numerical methods to evaluate minimum norm solutions weighted least squares problems and constrained weighted least squares problems. The main focus of this book is what happens when \(W\) or \(W_2\) is severely ill-conditioned. The book is self-contained. It is clearly written and it is easy to read and it presents a collection of new results recently published.