Computing the least-square solutions for centrohermitian matrix problems (Q2490228)

Algorithms for solving the following three problems are proposed: I. Given \(X,B\in \mathbb{C}^{n\times n}\), find a centrohermitian matrix \(A\in \mathbb{C}^{m\times n}\) such that \(\|AX-B\|=\min\); II. Given \(X,B\in \mathbb{C}^{n\times m}\), find a centrohermitian matrix \(A\in \mathbb{C}^{n\times n}\) such that \(AX=B\); III. Given \(\widetilde A\in \mathbb{C}^{n\times n}\), find \(A^*\in{\mathcal S}\) such that \(\|\widetilde A-A^*\|=\inf_{A\in{\mathcal S}} \|\widetilde A-A\|\), where: \(n>m\), \({\mathcal S}\) is the solution set of problem I or problem II, and \(\|\cdot\|\) is the Frobenius norm. The main tool to develop the algorithms is the reducibility of centrohermitian matrices. Finally, it is proved that the new algorithms introduced ensure significant savings in computational costs.