Is it true that:
If $b\ \bot\ im(A)$ then $b\ \bot\ im(A^T)$ ?
I think vaguely remembering LA class I would say it is not true. But this seems like this is what is being implied by scribe notes for linear regression
Is it true that:
If $b\ \bot\ im(A)$ then $b\ \bot\ im(A^T)$ ?
I think vaguely remembering LA class I would say it is not true. But this seems like this is what is being implied by scribe notes for linear regression
The answer is no. Take for example the matrix $$ A = \pmatrix{1 & 2 \\ -1 & -2}.$$ The image of $A$ is $\Im(A)= \text{span} \pmatrix{1 \\ -1}$
Take $$b = \pmatrix{1 \\ 1}.$$ Vector $b$ belongs to $\Im(A)^{\perp}$. However, since $\Im(A^{T}) = \text{span} \pmatrix{1 \\ 2}$, $b\notin \Im(A^T)^{\perp}$.