The linear targeting problem (Q6973053)

Let \(X\) and \(Y\) be \(m\times n\) real or complex matrices with \(m\geq n\) and \( X\neq 0\). The linear targeting problem is to determine whether, for a specified class of matrices, there exists an \(m\times m\) matrix \(A\) such that \(AX=Y\). The following are examples of the results obtained in the paper.\N\NWithout further restrictions, \(A\) can be chosen so that \(AX=Y\) if and only if \(\mathrm{null}(X)\subseteq \mathrm{null}(Y)\); and all solutions are of the form \(A=YX^{\dag }+Z(I-XX^{\dag })\), where \(Z\) an arbitrary \(m\times m\) matrix (\(\dag \) denotes the Moore-Penrose pseudoinverse.) Furthermore: \N\N(i) \(A \) can be chosen to be invertible if and only if the null space \(\mathrm{null} (X)=\mathrm{null}(Y)\); \N\N(ii) \(A\) can be chosen to be Hermitian if and only if \(X^{\ast }Y\) is Hermitian; \N\N(iii) \(A\) can be chosen to be positive definite if and only if \(X^{\ast }Y\geq 0\), \(\mathrm{null}(Y)=\mathrm{null}(X^{\ast }Y)\) and \textrm{rank}(\(X)=\mathrm{rank}(Y)\); \N\N(iv) \(A\) can be chosen to be unitary if and only if \(X^{\ast }X=Y^{\ast }Y\).