Matrix factorization refers to the process of decomposing a larger matrix $A$ into the product of two smaller matrices $X$ and $Y$. This decomposition reduces the number of elements required to represent the original matrix.
**Definition:** Consider a matrix $A$ with shape $(n \times m)$ and [[Matrix Rank]] $r$. This can be factorized as
$ A=XY$
where:
- $X$ is a $(n \times r)$ matrix
- $Y$ is a $(r \times m)$ matrix
![[matrix-factorization.png|center|300]]
While matrix $A$ has $(n*m)$ terms, the factorized version has only $(n*r)+(m*r)$ terms. When the rank $r$ is relatively small, this can lead to a significant decrease in matrix elements.
## Example
Consider the following matrix:
$ A= \begin{bmatrix} 1&2&3&5\\ 2&4&8&12\\ 3&6&7&13 \end{bmatrix} $
*Step1: Determine the Rank*
The matrix $A$ has rank $2$, as only two of its columns are [[Linearly Independent Vectors|linearly independent]]. The first and third columns $C_1$ and $C_3$ can serve as [[Basis Vectors]] for the column space of $A$.
*Step 2: Represent Columns as Linear Combinations*
Express each column of $A$ as a linear combination of the basis vectors $C_1$ and $C_3$.
$
\begin{align}
C_1 = 1*C_1+0*C_3 \\
C_2 = 2*C_1+0*C_3 \\
C_3 = 0*C_1+1*C_3 \\
C_4 = 2*C_1+1*C_3
\end{align}
$
*Step 3: Construct the Factorized Matrices*
The columns of $X$ are formed from the basis vectors $C_1$ and $C_3$.
$ X= \begin{bmatrix} 1&3\\ 2&8\\ 3&7 \end{bmatrix}$
The rows of $Y$ represent the coefficients of the linear combinations.
$ Y= \begin{bmatrix} 1&2&0&2\\ 0&0&1&1 \end{bmatrix}$
*Step 4: Verify the Factorization*
We see that the original matrix $A$ can be reconstructed by multiplying $X$ and $Y$.
$ A=XY=
\begin{bmatrix} 1&3\\ 2&8\\ 3&7 \end{bmatrix} \cdot
\begin{bmatrix} 1&2&0&2\\ 0&0&1&1 \end{bmatrix} =
\begin{bmatrix} 1&2&3&5\\ 2&4&8&12\\ 3&6&7&13 \end{bmatrix}
$