The \(\{P,Q,k+1\}\)-reflexive solution of matrix equation \(AXB=C\) (Q2511364): Difference between revisions

A complex matrix \(P\) is called Hermitian \(\left\{ k+1\right\} \)-potent if \(P^{k+1}=P=P^{\ast}\). Since the eigenvalues of a Hermitian matrix are real, the eigenvalues of \(P\) must lie in \(\{0,1,-1\}\), so really only the cases where \(k=1\) or \(2\) are of interest. In what follows we assume that \(P\in \mathbb{C}^{m\times m}\) and \(Q\in \mathbb{C}^{n\times n}\) are Hermitian \(\left\{ k+1\right\} \)-potent. A matrix \(X\in \mathbb{C}^{m\times n}\) is called \(\left\{ P,Q,k+1\right\} \)-reflexive if \(PXQ=X\) and antireflexive if \(PXQ=-X\). Let \(\mathcal{S}_{1}\) (respectively, \(\mathcal{S}_{2}\)) denote the set of \(\left\{ P,Q,k+1\right\} \)-reflexive (respectively, antireflexive) solutions to \(AXB=C\), where \(A\), \(B\) and \(C\) are given matrices of appropriate sizes. The authors describe the two sets \(\mathcal{S}_{1}\) and \(\mathcal{S}_{2}\) in the cases where \(k=1\) or \(2\), and use this to find the maximum and minimum ranks of the solutions. The final section of the paper applies these results to solve the following optimization problem: Given \(\tilde{X}\in \mathbb{C}^{m\times n}\) find \(\hat{X}\in\mathcal{S}_{i}\) such that \(\left\| \tilde {X}-\hat{X}\right\| \) is minimum.