We have the following matrix

$\left[\begin{array}{c} A\otimes B_1\\ \vdots\\ A\otimes B_T \end{array}\right]$, where $A$ and the $B_i$ are matrices, and $\otimes$ is the Kronecker product.

Is it possible to rewrite it as a Kronecker product where we gather all the $B_i$ matrices as one big matrix?

For example something like $(\cdots) \otimes \left[\begin{array}{c} B_1\\ \vdots\\ B_T \end{array}\right]$


You could use the fact that $$\mathbf A \otimes \mathbf B = \mathbf K_{M,N} \cdot \left( \mathbf B \otimes \mathbf A\right) \cdot \mathbf K_{P,Q}^{\rm T},$$ for an $\mathbf A \in \mathbb{F}^{M \times P}$, $\mathbf B \in \mathbb{F}^{N \times Q}$ and $\mathbb{F}$ is the field you are considering.

Here, $\mathbf K_{M,N}$ represents the commutation matrix of size $MN \times MN$, which is the permutation matrix defined via $$\mathbf K_{M,N}\cdot {\rm vec}\{\mathbf X\} = {\rm vec}\{\mathbf X^{\rm T}\}$$ for any $\mathbf X \in \mathbb{F}^{M \times N}$. A lot is known on these and they have some fun properties. Magnus and Neudecker [1,2] would be a good source to study.

You could apply this to each row to "turn around" all your Kronecker products, which then allows to pull out $\mathbf A$ completely. This should give you something like a ${\rm diag}\{\mathbf K_1, \ldots, \mathbf K_T\}$ in front, which may be a bit unhandy. But at least you know that such permutation matrices exist.

Not sure if this helps anything!

[1] Magnus, Jan R.; Neudecker, H., The commutation matrix: Some properties and applications, Ann. Stat. 7, 381-394 (1979). ZBL0414.62040.

[2] Magnus, Jan R.; Neudecker, Heinz, Matrix differential calculus with applications in statistics and econometrics, Wiley Series in Probability and Mathematical Statistics. Applied Probability and Statistics. Chichester etc.: John Wiley & Sons. XVII, 393 p.; £ 24.50 (1988). ZBL0651.15001.


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.