Let $A_m$ and $B_{ij}$ be matrices of size $M\times N$ where $m\in\{1,2\}$ and $N \times N$ where $i,j\in\{1,2\}$, respectively. Note that $m,i,j$ can be any number in general, but here they are specified so that the question can be clearer and easy to see.

I am wondering whether there is an alternative way to write/decompose the following block matrix more succinctly. I have had a look at Khatri-Rao product or Tracy-Singh product, but neither of them fits into the block matrix I have. I'm looking for a kind of block-wise multiplication of block matrix although I am not sure whether this type of matrix multiplication exists.

The matrix what I have is as follows:

$$ \begin{bmatrix} A_1B_{11}A_1^T & A_1B_{12}A_2^T\\ A_2B_{21}A_1^T & A_2B_{22}A_2^T \end{bmatrix} $$

I would be so grateful if anyone can suggest how this matrix can be decomposed or written more succinctly.


One way is $$\begin{bmatrix} A_1B_{11}A_1^T & A_1B_{12}A_2^T\\ A_2B_{21}A_1^T & A_2B_{22}A_2^T\end{bmatrix} = \begin{bmatrix} A_1 & 0\\ 0 & A_2 \end{bmatrix} \begin{bmatrix} B_{11} & B_{12}\\ B_{21} & B_{22} \end{bmatrix} \begin{bmatrix} A_1 & 0\\ 0 & A_2 \end{bmatrix}^T\,.$$

  • $\begingroup$ Indeed, this is definitely one way to decompose the matrix I post!!!!! Great! $\endgroup$ – Creatlee Jun 4 at 20:44

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.