Frobenius norm minimization with additions

44 Views Asked by At

I am trying to find minimizer X for which

$$ \arg \min_{X} {\left\|| A X + X B - Y \right\||}_{F} $$

when all matrices are square. I know to use the pseudoinverse of A for

$$ \arg \min_{X} {\left\|| A X - Y \right\||}_{F} $$

but don't understand how to rewrite the argument of the first equation to a solvable form.

1

There are 1 best solutions below

0
On

Since $\left\|A X + X B - Y \right\|_F=\|(I\otimes A+B^T\otimes I)\operatorname{vec}(X)-\operatorname{vec}(Y)\|_2$ (see Wikipedia for the meaning of the $\operatorname{vec}$ operator), this is just an ordinary least-square problem and a global minimiser is given by $\operatorname{vec}(X)=(I\otimes A+B^T\otimes I)^+\operatorname{vec}(Y)$.