1
$\begingroup$

I'm starting to understand how to compute and think about the transformation matrix of a linear transformation $L:\mathbb{R}^n \to \mathbb{R}^m$. However, I'm trying to understand how this translates to a vector spaces of matrices. For example, consider the linear map $L:\mathbb{R}^{n \times m} \to \mathbb{R}^{m \times p}$, and the standard basis $\beta=\{E_{ij}|1\leq i \leq n, 1\leq j \leq m\}$ where each $E_{ij}$ contains zeros except in the $i$-th row and $j$-th column, which contains a ones. From this it follows that any matrix $Q\in \mathbb{R}^{n \times m}$ can be written as a linear combination of these basis matrices. Applying the transformation $L$ to this matrix $Q$ we obtain another matrix $L(Q)\in \mathbb{R}^{m \times p}$. In the case of vectors the transformation matrix $A$ of $L$ has columns containing the coordinates of each basis vector $L(v_i)$ w.r.t. to basis $\beta_W$. However, I'm having difficulty translating this idea to the context in which the vectors are $m\times n$ matrices.

Any help would be much appreciated. Thanks,

$\endgroup$
1
$\begingroup$

I think the key here is you are converting your matrices into vectors by writing them as linear combinations of the basis vectors.

For example given an ordered basis (order matters for transformation matrices, you always have to keep the basis vectors in the same order!) $\beta_{3\times 3}=\{ E_{11},E_{12},E_{13},E_{21},E_{22},E_{23},E_{31},E_{32},E_{33}\}$:

$\begin{bmatrix} a_{11} & a_{12} & a_{13}\\ a_{21} & a_{22} & a_{23}\\ a_{31} & a_{32} & a_{33} \end{bmatrix} = a_{11}\cdot E_{11}+a_{12}\cdot E_{12}+\cdot \cdot \cdot +a_{33}\cdot E_{33} \equiv(a_{11},a_{12},...,a_{33})\in \mathbb{R}^{3\cdot 3}$

The beauty about vector spaces is you can treat whatever you are working with, in this case matrices, as a vector, so you are essentially converting your $n\times m$ matrix into a vector in $\mathbb{R}^{n\cdot m}$, you find the matrix of the linear transformation as you would for any linear transformation because you are back to familiar territory, and then that matrix allows you to input a vector (a matrix written as a linear combination of the ordered basis vectors of the domain) and you get as output another vector (another matrix written as linear combination of the ordered basis vectors of the codomain).

Hope this helps!

$\endgroup$
2
  • $\begingroup$ Not sure I fully understand "converting your $n \times m$ matrix into a vector in $\mathbb{R}^{n \times m}$". My confusion comes from the fact that the linear combination of basis matrices has $nm$ terms as such the coordinate vector becomes an $[v]_{\beta_W} \in \mathbb{R}^{nm}$. But in your $3 \times 3$ example, you obtain a coordinate matrix $[v]_{\beta_W} \in \mathbb{R}^{3 \times 3}$. This would imply that mapping from $ V \to W$ is a matrix multiplication instead of matrix-vector multiplication as would be the case if the elements were vectors in $\mathbb{R}^{n}$. Can you please elaborate $\endgroup$
    – dpakasa
    Dec 14 '20 at 19:21
  • $\begingroup$ I understand now, that was just a typo what I meant by $\mathbb{R}^{3\times 3}$ was actually $\mathbb{R}^{3\cdot 3}$, sorry! $\endgroup$ Dec 14 '20 at 19:35

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.