4.3-exam2-2001¶
Quiz 2, 2001 · Section 4.3: Least Squares Approximations Source: exam2_1.pdf, pp. 56-57
Problem¶
1 (36 pts.) Let \(A\) be the square matrix
\[
A = \begin{bmatrix} 1 \\ 1 \\ 1 \end{bmatrix} q_1^T + \begin{bmatrix} -1 \\ a \\ -1 \end{bmatrix} q_2^T,
\]
where \(q_1\) and \(q_2\) are orthonormal vectors in \(R^3\). (12p)
(a) Find \(x\) such that
\[
Ax = \begin{bmatrix} 1 \\ 1 \\ 1 \end{bmatrix}
\]
(b) Choose \(a\) such that the column space of \(A\) has dimension 1. (8p)
(c) If
\[
q_1 = \begin{bmatrix} 1 \\ 0 \\ 0 \end{bmatrix} \quad \text{and} \quad q_2 = \begin{bmatrix} 0 \\ 1 \\ 0 \end{bmatrix}
\]
and \(a = 0\), solve
\[
Ay = \begin{bmatrix} 1 \\ 2 \\ 3 \end{bmatrix}
\]
in the least squares sense. (16p)
Solution
1a. Take \(x = q_1\). Then \[ Aq_1 = \begin{bmatrix} 1 \\ 1 \\ 1 \end{bmatrix} (q_1^T q_1) + \begin{bmatrix} -1 \\ a \\ -1 \end{bmatrix} (q_2^T q_1) = \begin{bmatrix} 1 \\ 1 \\ 1 \end{bmatrix} \cdot 1 + \begin{bmatrix} -1 \\ a \\ -1 \end{bmatrix} \cdot 0 = \begin{bmatrix} 1 \\ 1 \\ 1 \end{bmatrix}. \] 1b. The column space of \(A\) is spanned by \((1, 1, 1)\) and \((-1, a, -1)\). Its dimension is 1 exactly when these two vectors are dependent, i.e. when \((-1, a, -1)\) is a multiple of \((1, 1, 1)\). Since \(-1 = -1 \cdot 1\), the multiple must be \(-1\), so \(a = -1\). 1c. With \(q_1 = e_1\), \(q_2 = e_2\) and \(a = 0\), \[ A = \begin{bmatrix} 1 \\ 1 \\ 1 \end{bmatrix} e_1^T + \begin{bmatrix} -1 \\ 0 \\ -1 \end{bmatrix} e_2^T = \begin{bmatrix} 1 & -1 & 0 \\ 1 & 0 & 0 \\ 1 & -1 & 0 \end{bmatrix}. \] Solve the normal equations \(A^T A y = A^T b\) with \(b = (1, 2, 3)\): \[ A^T A = \begin{bmatrix} 3 & -2 & 0 \\ -2 & 2 & 0 \\ 0 & 0 & 0 \end{bmatrix}, \qquad A^T b = \begin{bmatrix} 6 \\ -4 \\ 0 \end{bmatrix}. \] From \(\begin{bmatrix} 3 & -2 \\ -2 & 2 \end{bmatrix} \begin{bmatrix} y_1 \\ y_2 \end{bmatrix} = \begin{bmatrix} 6 \\ -4 \end{bmatrix}\) we get \(y_1 = 2, y_2 = 0\). The third component is free (column 3 is zero), so the least squares solutions are \[ y = \begin{bmatrix} 2 \\ 0 \\ y_3 \end{bmatrix} \] for any \(y_3\), and the projection is \(p = Ay = (2, 2, 2)\).Keywords: column space, rank one, least squares, normal equations