Skip to content

4.4-final-2001-b

Final Exam, 2001 · Section 4.4: Orthonormal Bases and Gram-Schmidt Source: finalexam_1.pdf, p. 57

Problem

5 (32 pts.) Let

\[ A = QR = \begin{bmatrix} 1/5 & -2/5 & -4/5 \\ 2/5 & 1/5 & 2/5 \\ 2/5 & -4/5 & 2/5 \\ 4/5 & 2/5 & -1/5 \end{bmatrix} \begin{bmatrix} 5 & -2 & 1 \\ 0 & 4 & -1 \\ 0 & 0 & a \end{bmatrix} \]

(\(Q\) has orthonormal columns).

(a) Give an orthonormal basis of \(C(A)\).

(b) For which values of \(a\) is the rank of \(A\) equal to 2?

(c) Solve \(A^Tx = \begin{bmatrix} 0 \\ 0 \\ a \end{bmatrix}\).

(d) Let \(a = 2\). Solve \(Ay = b\) in the least squares sense for

\[ b = \begin{bmatrix} -1 \\ 1 \\ 1 \\ -2 \end{bmatrix}. \]
Solution (a) \(A = QR\) with \(Q\) orthonormal, so an orthonormal basis of \(C(A)\) is given by the columns of \(Q\): \[ q_1 = \begin{bmatrix} 1/5 \\ 2/5 \\ 2/5 \\ 4/5 \end{bmatrix}, \quad q_2 = \begin{bmatrix} -2/5 \\ 1/5 \\ -4/5 \\ 2/5 \end{bmatrix}, \quad q_3 = \begin{bmatrix} -4/5 \\ 2/5 \\ 2/5 \\ -1/5 \end{bmatrix}. \] (b) \(Q\) has orthonormal columns, so rank \(A = \text{rank } R\). \(R\) is upper triangular with diagonal \(5, 4, a\), so rank \(A = 2\) exactly when \[ a = 0. \] (c) \(A^Tx = R^TQ^Tx\). Set \(y = Q^Tx\), so \(R^Ty = (0, 0, a)\): \[ \begin{bmatrix} 5 & 0 & 0 \\ -2 & 4 & 0 \\ 1 & -1 & a \end{bmatrix} \begin{bmatrix} y_1 \\ y_2 \\ y_3 \end{bmatrix} = \begin{bmatrix} 0 \\ 0 \\ a \end{bmatrix} \] gives \(y_1 = 0\), \(y_2 = 0\), and \(ay_3 = a\). For \(a \neq 0\), \(y = (0, 0, 1)\) and \(x = Qy = q_3 = (-4/5, 2/5, 2/5, -1/5)\). For \(a = 0\), \(y_3\) is free and \(x = t\,q_3\), \(t \in R\). (d) The normal equations reduce to \(R\hat{y} = Q^Tb\): compute \[ Q^Tb = \begin{bmatrix} q_1 \cdot b \\ q_2 \cdot b \\ q_3 \cdot b \end{bmatrix} = \begin{bmatrix} -1 \\ -1 \\ 2 \end{bmatrix}. \] Solve \(R\hat{y} = (-1, -1, 2)\): \(2\hat{y}_3 = 2\) so \(\hat{y}_3 = 1\); \(4\hat{y}_2 - \hat{y}_3 = -1\) so \(\hat{y}_2 = 0\); \(5\hat{y}_1 + \hat{y}_3 = -1\) so \(\hat{y}_1 = -2/5\). The least squares solution is \[ \hat{y} = \begin{bmatrix} -2/5 \\ 0 \\ 1 \end{bmatrix}. \]

Keywords: QR factorization, least squares, rank, normal equations