The Moore-Penrose inverse of a partitioned matrix M=pmatrix A&0\\ B&Cpmatrix.

From MaRDI portal
Publication:4101964





For a rectangular matrix \(M\) which can be partitioned as \(M=\begin{pmatrix} A&0\\ B&C\end{pmatrix}\) the Moore-Penrose inverse is given by \(^+ = \begin{pmatrix} A^+&B^*L^+ \\ 0&C^*L^+\end{pmatrix}\) where \(L =BB^*+ CC^*\) if and only if \(A^*B=0\). Similarly, \(M^+ = \begin{pmatrix} K^+A^* &K^+B^* \\ 0 &C^+\end{pmatrix}\) where \(K =A^*A+B^*B\) if and only if \(B^*C =0\). The author then presents a general (rather complicated) formula for \(M^+\). A theorem of C. Meyer giving necessary and sufficient conditions that \(M^+\) be triangular block form when \(M\) is, follows as a corollary.











This page was built for publication: The Moore-Penrose inverse of a partitioned matrix \(M=\begin{pmatrix} A&0\\ B&C\end{pmatrix}\).

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4101964)