Skip to content

7.1-final-2013

Final Exam, 2013 · Section 7.1: Singular Values and Singular Vectors Source: finalexam_2.pdf, pp. 91-92

Problem

3 (22 pts.)

(a) (5 pts.) What are the two eigenvalues of \(A = \begin{bmatrix} 3 & 2 \\ 2 & 0 \end{bmatrix}\)?

(b) (5 pts.) What is \(\kappa\), the ratio of the maximum and minimum values of \(\|Ax\|\) over the unit circle \(\|x\| = 1\)?

(c) (6 pts.) What are the maximum and minimum value of \(v^T x\), for \(x\) over the unit circle \(\|x\| = 1\), when \(v = \begin{bmatrix} 2 \\ 0 \end{bmatrix}\)? Same question when \(v = \begin{bmatrix} 3 \\ 2 \end{bmatrix}\)? (Hint: the minimum is negative.)

(d) (6 pts.) When every point on the unit circle is multiplied by \(A\), the result is an ellipse. Find the dimensions of the tightest rectangle with sides parallel to the coordinate axes that encloses the ellipse. (Hint: the previous maximum and minimum question is meant to be a warm-up. Another hint: singular values are not so useful here.)

Solution (a) The eigenvalues are the roots of \[ \det(A - \lambda I) = -\lambda(3 - \lambda) - 4 = \lambda^2 - 3\lambda - 4 = (\lambda + 1)(\lambda - 4), \] so \[ \lambda = -1 \quad \text{or} \quad \lambda = 4. \] (b) \(A\) is symmetric, so the singular values are the absolute values of the eigenvalues: \(\sigma_1 = 4\), \(\sigma_2 = 1\). Over the unit circle the maximum and minimum of \(\|Ax\|\) are the largest and smallest singular values: \[ \kappa = \frac{\sigma_1}{\sigma_2} = \frac{4}{1} = 4. \] (c) By Cauchy-Schwarz, \(v^T x\) is maximized when \(x\) points along \(v\) and minimized when \(x\) points opposite to \(v\). For \(v = (2, 0)\): \[ \max = 2 \quad (x = (1, 0)), \qquad \min = -2 \quad (x = (-1, 0)). \] For \(v = (3, 2)\): \[ \max = \sqrt{13} \quad \left(x = \frac{1}{\sqrt{13}} \begin{bmatrix} 3 \\ 2 \end{bmatrix}\right), \qquad \min = -\sqrt{13} \quad \left(x = -\frac{1}{\sqrt{13}} \begin{bmatrix} 3 \\ 2 \end{bmatrix}\right). \] (d) A point \(Ax\) of the ellipse has coordinates \(((3, 2) \cdot x, \ (2, 0) \cdot x)\). From part (c) with \(v = (3, 2)\) (the first row of \(A\)), the \(x\)-coordinate ranges over \([-\sqrt{13}, \sqrt{13}]\); with \(v = (2, 0)\) (the second row of \(A\)), the \(y\)-coordinate ranges over \([-2, 2]\). So the tightest enclosing rectangle has corners \[ (\pm \sqrt{13}, \pm 2), \] with dimensions \[ 2\sqrt{13} \times 4. \]

Keywords: eigenvalues, singular values, condition number, ellipse