6
$\begingroup$

For an invertible matrix $A$, we have the identity \begin{align} \dfrac{\partial \det A}{\partial A} = \det A (A^{-1})^T \end{align} where the $T$ denotes the transpose operation.

How does this formula change when considering the Kronecker product of $A$ with some other (invertible) matrix $B$? For example, how does one compute the following:

\begin{align} \dfrac{\partial \det \left[A\otimes B\right]}{\partial A} \end{align}

$\endgroup$
3
$\begingroup$

I find this question very interesting. I could not find any reference to this problem in any book. So, here is my derivation for square matrices and I would like to hear other people to know if it is correct.

$A$ being ${n\times n}$ matrix and $B$ being ${p\times p}$ matrix. We know that $$\det[A\otimes B]=(\det{A})^p(\det{B})^n$$ (Notice how the power switches between $n$ and $p$)

Then the derivative is $$\frac{\partial(\det{A})^p(\det{B})^n}{\partial A}=(\det{B})^np(\det A)^{p-1}\det A(A^{-1})^T=p(\det B)^n(\det A)^p(A^{-1})^T$$

To put it "nicer":

$$\frac{\partial\det[A\otimes B]}{\partial A}=p\det[A\otimes B](A^{-1})^T$$

$\endgroup$
  • $\begingroup$ Thank you! I have confirmed that this is the correct answer. I had initially derived this result including the determinant of the Kronecker product, but I had made a mistake computing the determinant of my matrix leading me to think this answer was incorrect. $\endgroup$ – Matt Apr 24 '14 at 4:12
  • $\begingroup$ @Matt. I would be actually interested in some application where you use this result. Are you working on some "bigger" problem? $\endgroup$ – pisoir Apr 24 '14 at 6:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.