2
$\begingroup$

I have to find $x$ that minimizes: $$ \sum_{k}(x^H\textbf A_kx - b_k)^2$$ where $A_k$ are 4 x 4 positive definite matrices ($A_1, A_2,...A_k$), $x$ is 4 x 1 vector and $b_k$ are scalars ($b_1,b_2,...b_k)$. Can anyone give me some suggestion/hint inorder to solve this OP ?

$\endgroup$
4
  • $\begingroup$ You will probably have to use a numerical approach like Gauss-Newton or BFGS $\endgroup$
    – Doc Brown
    Dec 4, 2015 at 9:20
  • $\begingroup$ Thank you for your suggestion. Since this nonlinear OP will have multiple local minima (multiple $x$ that would solve the OP), is it possible to solve in a way that $x$ is a global minimum? $\endgroup$
    – zsha
    Dec 4, 2015 at 9:45
  • $\begingroup$ Since you did not mention any constraints: can one safely assume there are no constraints for $x$? $\endgroup$
    – Doc Brown
    Dec 4, 2015 at 10:57
  • $\begingroup$ Yes, there are no constraints for $x$ in general (the only condition is that $Rank(x^Hx)=1$, which is automatically satisfied based on definition of the OP so I haven't consider that as a constraint). I am looking for an approach that would yield an $x$ which is a global minimum. Any suggestions on this would be appreciated. $\endgroup$
    – zsha
    Dec 4, 2015 at 11:19

0