Say I have the following variable length vectors containing unknown values:

$$ A=\left (\begin{array}{c} a_1 \\ a_2 \\ \vdots\\ a_i\\ \end{array}\right) B=\left (\begin{array}{c} b_1 \\ b_2 \\ \vdots\\ b_j\\ \end{array}\right) $$

And the following variable length vectors and matrix (according to the size of the preceding vectors) containing known values:

$$ A_T=\left (\begin{array}{c} a'_1 \\ a'_2 \\ \vdots\\ a'_i\\ \end{array}\right) B_T=\left (\begin{array}{c} b'_1 \\ b'_2 \\ \vdots\\ b'_j\\ \end{array}\right) K=\left (\begin{array}{ccc} k_{11} & \cdots & k_{1j} \\ \vdots & \ddots & \vdots \\ k_{i1} & \cdots & k_{ij}\\ \end{array}\right) $$

These are all related by the following two equations: $$ A_T = A + diag(A)KB\\ B_T = B + diag(B)K^TA $$

Given all of the known values, I want to know if it is possible to solve for A and B. If so how should I go about figuring this out? I only have a basic education on matrix math.

  • $\begingroup$ My bad, I edited it. The values are different, just the same length. $\endgroup$ Mar 29, 2015 at 6:34
  • $\begingroup$ This looks pretty weird to me. In any case the equation are not linear in $A,B$, due to the diag factors. So standard linear algebra methods do not apply. $\endgroup$ Mar 29, 2015 at 9:05

1 Answer 1


As Marc writes, the problem is not a linear one. Yet, I did numerical tests and the results are surprising. Of course, we work with generic systems. When $i=j=3$ or $4$, the associated system of algebraic equations decompose in linear subsystems; that is, the solutions $(a_k),(b_k)$ are rational functions of the $(k_{*,*}),(a'_{*}),(b'_{*})$. Moreover, there is only a finite number of solutions in $(A,B)$; more precisely, $20$ solutions for $i=j=3$ and $70$ solutions for $i=j=4$.


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.