The magic factorization expresses a matrix as:
$ A = CW^{-1}R^* $
where:
- $C$ is an $m \times r$ matrix of linearly independent columns from $A$
- $R^*$ is an $r \times n$ matrix of linearly independent rows from $A$
- $W$ is the $r \times r$ invertible submatrix at the intersection
- $r = \text{rank}(A)$
Construction:
- Identify pivot positions in RREF$(A)$
- Extract pivot columns from $A$ to form $C$
- Extract pivot rows from $A$ to form $R^*$
- $W$ is the submatrix of $A$ at intersection of these rows/columns
Applications:
- Expresses $A$ in terms of its essential structure
- Generalizes LU factorization
- Useful in numerical linear algebra