Suppose i have a vector space $V=\mathbb{F}^{3\times2}$ of $3\times2$ matrices and a vector space $W=\mathbb{F}^{2\times2}$ of $2\times2$ matrices over the field $\mathbb{F}$.

Any Linear transformation $T: V\to W$ would take a $3\times2$ matrix and transform it into a $2\times2$.

I tried to compare this with the fact that multiplying a $2\times3$ matrix by a $3\times2$ matrix also results in a $2\times2$ matrix.

But obviously this $2\times3$ matrix is not the matrix representation of $T$.

To better understand, i defined a standard basis for $V$ and made some algebra resulting in this fact :

Where the elements of the $6\times1$ matrix are the vectors from the standard basis of $\mathbb{F}^{3\times2}$ after being transformed by $T$.

By all of this, i found that there are at least two ways to implement matrix multiplication to result in a $2\times2$ matrix.

My question :

Is the ( $1\times6$ times $6\times1$ ) matrix multiplication the only one that is equivalent to the ($2\times3$ times $3\times2$) matrix multiplication ?

I can't find detailed information about matrix multplication, even on linear algebra books ( even wikipedia http://en.wikipedia.org/wiki/Matrix_multiplication doesn't mention the equivalent matrix multiplications i found out empirically ).

Is there any topic name, site, resource, book that can teach much more than the algorithm approach (row dot column ) of matrix multiplication ?

  • $\begingroup$ I didn't understand your explicit question, but here's a comment about $T$ itself: $T$ is a linear transformation from a six-dimensional vector space $V$ to a four-dimensional vector space $W$. Therefore if you choose a basis for $V$ and a basis for $W$, the corresponding matrix representation of $T$ will be a $4\times6$ matrix. $\endgroup$ – Greg Martin Apr 23 '13 at 7:45

Not all linear maps between matrix spaces arise from one-sided matrix multiplication.

Indeed, the space of linear maps from $F^{3\times 2}$ to $F^{2\times 2}$ is $24$-dimensional ($6\cdot 4=24$). Some of them are realized as matrix multiplication $M\mapsto AM$ where $A\in F^{2\times 3}$. However, the matrix $A$ has only $6$ entires, so the linear spaces of all such transformations is $6$-dimensional: it is a proper subspace of $L(F^{3\times 2},F^{2\times 2})$.

However, every linear map from $F^{3\times 2}$ to $F^{2\times 2}$ can be written as a finite sum of two-sided multiplication maps $M\mapsto AMB$. Indeed, $24$ terms suffice: $$M\mapsto \sum_{j=1}^{24} A_j MB_j$$ where each matrix $A_j$ and $B_j$ has one nonzero entry. To see why, observe that by placing $1$ at one entry of $A$ and one entry of $B$ one can move any given entry of $M$ into any desired position in $AMB$.

| cite | improve this answer | |

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.