Block matrix pseudoinverse: Difference between revisions

Content deleted Content added
bmatrix
Excess q
Line 5:
Consider a column-wise partitioned matrix:
:<math>
\begin{bmatrix}\mathbf A & \mathbf B\end{bmatrix},\qquadquad
\mathbf A \in \reals^{m \times n},\qquadquad
\mathbf B \in \reals^{m \times p},\qquadquad
m \geq n + p.
</math>
Line 89:
=== Column-wise partitioning in over-determined least squares ===
 
Suppose a solution <math>
<math> \mathbf x = \begin{bmatrix}
\mathbf x_1 \\
\mathbf x_2 \\
Line 103:
\mathbf x_2 \\
\end{bmatrix} =
\mathbf d,\qquadquad
\mathbf d \in \reals^{m \times 1}.
</math>
Line 120:
Therefore, we have a decomposed solution:
:<math>
\mathbf x_1 = \left(\mathbf P_B^{\perp} \mathbf A\right)^{+}\,\mathbf d,\qquadquad
\mathbf x_2 = \left(\mathbf P_A^{\perp} \mathbf B\right)^{+}\,\mathbf d.
</math>
Line 126:
=== Row-wise partitioning in under-determined least squares ===
 
Suppose a solution <math> \mathbf x </math> solves an under-determined system:
:<math>
\begin{bmatrix}
Line 135:
\mathbf e \\
\mathbf f
\end{bmatrix},\qquadquad
\mathbf e \in \reals^{n \times 1},\qquadquad
\mathbf f \in \reals^{p \times 1}.
</math>