Low phase-rank approximation (Q2074973): Difference between revisions

The authors study low phase-rank approximation problems, which serve as a counterpart to the well-known low-rank approximation problem and the Schmidt-Mirsky theorem (see [\textit{G. W. Stewart} and \textit{J.-g. Sun}, Matrix perturbation theory. Boston etc.: Academic Press, Inc. (1990; Zbl 0706.65013), Chap. IV; \textit{I. Markovsky}, Low rank approximation. Algorithms, implementation, applications. London: Springer (2012; Zbl 1245.93005), Chap. 2]). In this paper, the (canonical) phases of a matrix are defined for a special family of square matrices, called sectorial matrices, whose numerical ranges do not contain the origin. Based on the definition of canonical phases, the authors propose the concept of phase-rank (henceforth abbreviated as prank) defined as the number of the nonzero phases of a sectorial matrix. The authors derive a majorization inequality between the phases of the geometric mean and the arithmetic mean of the phases, similarly to the Ky-Fan inequality for eigenvalues of Hermitian matrices. A low phase-rank approximation problem is formulated based on the geometric mean of matrices and a theorem characterizing its solutions is obtained. It is also shown that low phase-rank approximation problems are connected to low-rank approximation problems, but that in general one cannot be reformulated as the other. Moreover, the authors provide an alternative formulation of the low phase-rank approximation problem using geodesic distances between sectorial matrices and characterized the set of optimal solutions.