1
$\begingroup$

Which of the four subspaces contains $x$, for the given equation:

$$A^T Ax=A^Tb$$ where $b$ is a column vector NOT in the column space of $A_{m\times n}$. (This equation is to find the best solution for $Ax=b$). Also, $A$ is assumed to have independent columns.

(Question is from one of the MIT OCW problem sets.)

My attempt: Null space of $A$ is just the zero vector. So $\mathbb R^n$ is entirely the row space. Since $x$ has to be a $n\times 1$ vector, it belongs in the row space.

Is there any other more intuitive way to visualize this? Can I generalize this to any $Ax=b$ also, if rank $=n$ for $A$?

$\endgroup$
0

2 Answers 2

1
+50
$\begingroup$

Well, to me your attempt is as intuitive as it gets. Just a remark: you assumed that $m\neq n$ to conclude that $x$ can't be in the column space or left null space, which a priori is not given (at least, it wasn't stated in the question). But this is ok, since, if $m=n$ and $A$ has independent columns, $A$ would be an invertible square matrix, and $A^TAx=A^Tb\implies Ax=b$, but, since $b$ is assumed not to be in the column space of $A$, this case doesn't happen.

Maybe what you're asking for is a more explicit explanation? If so, Let $A=\begin{bmatrix}L_1 \\ \vdots\\ L_m\end{bmatrix}$, where $L_i$ is the $i$-th line, a $1\times n$ vector. If $b=\begin{bmatrix} b_1 \\ \vdots\\ b_m\end{bmatrix}$, we have that

\begin{align*} A^TAx=A^Tb&\implies \begin{bmatrix} L_1 & \cdots & L_m\end{bmatrix}\begin{bmatrix}L_1 \\ \vdots\\ L_m\end{bmatrix}\textbf{x}=\begin{bmatrix} L_1 & \cdots & L_m\end{bmatrix}\begin{bmatrix} b_1 \\ \vdots\\ b_m\end{bmatrix}\\ &\implies(L_1\cdot L_1+\dots+L_n\cdot L_n)\textbf{x}=b_1L_1+\dots+b_mL_m\\ &\implies\textbf{x}=\dfrac{1}{L_1\cdot L_1+\dots+L_n\cdot L_n}(b_1L_1+\dots+b_mL_m) \end{align*}

So $\textbf{x}$ is indeed in the row space of $A$ (notice that the denominator of the fraction is clearly non-zero).

$\endgroup$
0
$\begingroup$

Your solution is already perfectly simple. By the four subspaces theorem, $x$ can be written as $x = x_1 + x_2$, where $x_1 \in R(A^T)$ and $x_2 \in N(A)$. But the null space of $A$ is trivial, so $x_2 = 0$ and $x \in R(A^T)$.

$\endgroup$

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.