Given a possibly rectangular m×n matrix A with rank k, the LUQ Decomposition takes the form:
A=^L^U^Q
where ^L is an invertible m×m matrix, ^Q is an invertible n×n matrix and ^U has the form:
^U=(^U11000),
where ^U11 is a k×k invertible upper triangular matrix.
To construct the LUQ decompostion, we start with a standard LU decomposition with pivoting determines matrices such that:
PAQ=LU,
where P is a m×m permutation matrix, Q is a n×n permutation matrix, L is an invertible m×m lower triangular matrix, and U is a m×n upper triangular matrix.
Lemma 1: There exists an m×m row permutation R and a n×n column permutation C of U from the decomposition of A above such that the matrix:
¯¯¯¯U=RUC
has the form
¯¯¯¯U=(¯¯¯¯U11¯¯¯¯U12¯¯¯¯U21¯¯¯¯U22)
where ¯¯¯¯U11 is a square r×r matrix has a non-zero diagonal and ¯¯¯¯U22 has a zero diagonal. Further, if A has rank k≤m≤n then r≤k.
Proof: (invoke some property of U maintained during LU decomposition) █
We can plug this into our decomposition:
PAQ=LU
PAQ=LRTRUCCT
PAQ=LRT¯¯¯¯UCT
PAQC=LRT¯¯¯¯U
Let’s let Q and L swallow these permutations: ¯¯¯¯Q=QC and ¯¯¯¯L=LRT, so that we have,
PA¯¯¯¯Q=¯¯¯¯L¯¯¯¯U
where ¯¯¯¯Q is an n×n permutation matrix and ¯¯¯¯L is an invertible (not necessarily lower triangular) m×n matrix.
Now we will transform ¯¯¯¯U into a block diagonal matrix by creating zeros below ¯¯¯¯U11 and to the right of ¯¯¯¯U11. Accomplish this by constructing an m×m matrix K and a n×n matrix S over the following form:
¯¯¯¯U=(I0¯¯¯¯U21¯¯¯¯U –111I)K(¯¯¯¯U1100¯¯¯¯U22−¯¯¯¯U21¯¯¯¯U –111¯¯¯¯U12)˜U(I¯¯¯¯U –111¯¯¯¯U120I)S,¯¯¯¯U=K˜US.