1
$\begingroup$

I have an equation of the following form:

$$BK=A$$

where $A\in\mathbb{R}^{n\times n}$, $B\in\mathbb{R}^{n\times l}$ and $K\in\mathbb{R}^{l\times n}$, with $l\leq n$. $B$ and $A$ are known, and I want to find some $K$ that satisfies the equation.

Due to the matrices $B$ and $K$ not being square (in general) the obvious problem is that $B$ cannot be inverted to solve the problem directly.

So, there are $n\times n$ equations and $n\times l$ unknowns. I'm trying to express this system in the usual, vector form:

$$\hat{B}k=a$$

with $\hat{B}$ being a square matrix, and $a$ and $k$ unidimensional vectors, to solve it as a common underdetermined linear system.

Is there a way to re-write the original equation in this way?

$\endgroup$
  • $\begingroup$ The equation has a solution if and only if each column of $A$ belongs to the column space of $B$. If you denote column $i$ of $A$ by $a_i$, this is equivalent to the solvability of each of the $\ell$ equations $B x = a_i$. $\endgroup$ – Umberto P. May 30 '18 at 16:48
  • $\begingroup$ @UmbertoP. What would $x$ mean? $\endgroup$ – Tendero May 31 '18 at 13:26
  • $\begingroup$ If $Bx = a_i$ has a solution for each $i$, then $x$ will be column $i$ of the matrix $K$. $\endgroup$ – Umberto P. May 31 '18 at 14:44
0
$\begingroup$

After some thought I managed to get what I wanted. I'll use an example to make it easier to understand. Suppose that:

$$ A= \left[ {\begin{array}{} a_1 & a_2 & a_3\\ a_4 & a_5 & a_6\\ a_7 & a_8 & a_9\\ \end{array} } \right] $$

$$ K= \left[ {\begin{array}{} k_1 & k_2 & k_3\\ k_4 & k_5 & k_6 \\ \end{array} } \right] $$

$$ B= \left[ {\begin{array}{} b_1 & b_2 \\ b_3 & b_4 \\ b_5 & b_6 \\ \end{array} } \right] $$

Then:

$$BK=A\implies \left[ {\begin{array}{} b_1 & b_2 \\ b_3 & b_4 \\ b_5 & b_6 \\ \end{array} } \right] \left[ {\begin{array}{} k_1 & k_2 & k_3\\ k_4 & k_5 & k_6 \\ \end{array} } \right] = \left[ {\begin{array}{} a_1 & a_2 & a_3\\ a_4 & a_5 & a_6\\ a_7 & a_8 & a_9\\ \end{array} } \right] $$

It can be seen that this set of equations can be also expressed as:

$$ \left[ {\begin{array}{} b_1 & b_2 & 0& 0& 0& 0\\ 0& 0&b_1 & b_2 & 0& 0\\ 0& 0 & 0& 0&b_1 & b_2\\ b_3 & b_4 & 0& 0& 0& 0\\ 0& 0&b_3 & b_4 & 0& 0\\ 0& 0 & 0& 0&b_3 & b_4\\ b_5 & b_6 & 0& 0& 0& 0\\ 0& 0&b_5 & b_6 & 0& 0\\ 0& 0 & 0& 0&b_5 & b_6\\ \end{array} } \right] \left[ {\begin{array}{} k_1 \\ k_4 \\ k_2\\ k_5 \\ k_3 \\ k_6 \\ \end{array} } \right] = \left[ {\begin{array}{} a_1 \\a_2\\ a_3\\ a_4 \\ a_5 \\ a_6\\ a_7 \\ a_8 \\ a_9\\ \end{array} } \right] $$

It's easy to take this to the general case where sizes are not $3$ and $2$ but $n$ and $l$.

| cite | improve this answer | |
$\endgroup$

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.