Simultaneous singular value decomposition (Q2431173)

Given a set of \(m\times n\) real or complex matrices \(A_1,\dots,A_N\), the authors find an \(m\times m\) orthogonal or unitary matrix \(P\), and an \(n\times n\) orthogonal or unitary matrix \(Q\) such that \(P^*A_1Q,\dots,P^*A_NQ\) are in a common block-diagonal form with possibly rectangular diagonal blocks. This problem is called the simultaneous singular value decomposition (simultaneous SVD). The authors use the theory of *-algebras and bimodules to prove that a finest simultaneous SVD is uniquely determined. In addition, they present an algorithm for finding the finest simultaneous SVD on the basis of recent algorithms of Murota-Kanno-Kojima-Kojima [\textit{K. Murota}, et al., Japan J. Ind. Appl. Math. 27, No. 1, 125--160 (2010; Zbl 1204.65068)] and of \textit{T. Maehara} and \textit{K. Murota} [ibid. 27, No. 2, 263--293 (2010; Zbl 1204.65035)] for simultaneous block-diagonalization of square matrices under orthogonal or unitary similarity.