My question is very simple, but I it seems to be impossible for me to figure it out. Here we go, find $\mathbf Z$ such that $$ (\mathbf Z^t\mathbf Z)^{-1}\mathbf Z^t\mathbf W=X_1Y_1+X_2Y_2 $$ Where $\mathbf Z$, $\mathbf W$, $X_1$, $X_2$, $Y_1$ and $Y_2$ are all matrices.
Asked
Active
Viewed 48 times
0
-
Is this related to a least squares problem? – Sean Roberson May 31 '17 at 18:26
-
Yes it is related but in a somewhat complicated way – Chamberlain Mbah May 31 '17 at 22:01
1 Answers
0
If we assume that such a $Z$ exists with linearly independent columns, then $( Z^t Z)^{-1} Z^t = Z^+$ is the Moore-Penrose pseudo inverse of $Z$. So you could try to solve $ U W = X_1Y_1+X_2Y_2$ first by vectorization and afterwards recover $Z = (Z^+)^+ = U^+$.
Hyperplane
- 11,659
-
Thanks a lot the Moore-Penrose pseudo-inverse does the trick. I am grateful – Chamberlain Mbah May 31 '17 at 22:03