About the generalized \(LM\)-inverse and the weighted Moore-Penrose inverse (Q961562)

The weighted Moore-Penrose inverse \(X\) of the matrix \(A\in{\mathbb C}^{m\times n}\) is defined by the equations \(AXA=A\), \(XAX=X\), \((MAX)^*=MAX\), \((NXA)^*=NXA\) where \(M\) and \(N\) are positive definite weighting matrices and the superscript * denotes the conjugate transpose. If the weighting matrices are unit matrices, this gives the ordinary Moore-Penrose inverse. An algorithm to compute this weighted Moore-Penrose inverse was given by \textit{G. Wang} and \textit{Y. Chen} [J. Comput. Math. 4, 74--85 (1986; Zbl 0617.65033)]. The generalized \(LM\)-inverse of \(A\) is defined as the matrix \(A^+\) such that \(x=A^+b\) minimizes \(\|L^{1/2}(Ax-b)\|^2=\|Ax-b\|_L\) and \(\|M^{1/2}x\|=\|x\|_M\). Again \(L\) and \(M\) are positive definite weighting matrices. An algorithm to compute the this generalized inverse was given by \textit{F. E. Udwadia} and \textit{P. Phohomsiri} [Appl. Math. Comput. 190, No.~2, 999--1006 (2007; Zbl 1125.15007)]. Both algorithms are explicitly given and then it is proved that both algorithms are equivalent and compute the same result. The appendix of this paper gives the \texttt{Mathematica} code of the algorithms.