Skip to content

7.1-exam3-2015

Exam III, 2015 · Section 7.1: Singular Values and Singular Vectors Source: exam3_2.pdf, pp. 104, 106

Problem

3. (34 points)

(a) If \(C\) is any symmetric matrix, show that \(e^C\) is a positive definite matrix. We can see that \(e^C\) is symmetric -- which test will you use to show that \(e^C\) is positive definite?

(b) \(A\) is a 3 by 3 matrix. Suppose \(v_1, v_2, v_3\) are orthonormal eigenvectors (with eigenvalues 1, 2, 3) of the symmetric matrix \(A^TA\). Show that \(Av_1, Av_2, Av_3\) are orthogonal by rewriting and simplifying \((Av_i)^T(Av_j)\).

(c) For the 3 by 3 matrix \(A\) in part (b), find three matrices \(U, \Sigma, V\) that go into the Singular Value Decomposition \(A = U\Sigma V^T\).

(d) TRUE or FALSE: If \(A\) is any symmetric 4 by 4 matrix and \(M\) is any invertible 4 by 4 matrix, then \(B = M^{-1}AM\) is also symmetric. Give a reason for true or false.

Solution (a) We can write \(C = Q\Lambda Q^T\) for some orthogonal matrix \(Q\) and some diagonal matrix \(\Lambda\). Then \(e^C = Qe^\Lambda Q^T\), which immediately shows that \(e^C\) is symmetric. If \(\Lambda = diag(\lambda_1, \dots, \lambda_n)\), then \(e^\Lambda = diag(e^{\lambda_1}, \dots, e^{\lambda_n})\), so that each eigenvalue of \(e^C\) is of the form \(e^\lambda\). In particular, it is positive, so that \(e^C\) is positive definite. (b) \[ (Av_i)^T(Av_j) = v_i^T(A^TA)v_j = \lambda_j v_i^Tv_j = \lambda_j\, v_i \cdot v_j. \] Since \(v_1, v_2\), and \(v_3\) are orthogonal, we see that \((Av_i)^T(Av_j) = 0\) when \(i \neq j\), i.e. \(Av_1, Av_2\), and \(Av_3\) are orthogonal. (c) \[ V = (v_1 \mid v_2 \mid v_3), \qquad \Sigma = diag(1, \sqrt{2}, \sqrt{3}), \qquad U = \left(Av_1 \mid \frac{Av_2}{\sqrt{2}} \mid \frac{Av_3}{\sqrt{3}}\right). \] (d) False. Take \[ A = \begin{bmatrix} 0 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 \\ 0 & 0 & 1 & 0 \\ 0 & 0 & 0 & 1 \end{bmatrix}, \qquad M = \begin{bmatrix} 1 & 1 & 0 & 0 \\ 0 & 1 & 0 & 0 \\ 0 & 0 & 1 & 0 \\ 0 & 0 & 0 & 1 \end{bmatrix}. \] Then \[ M^{-1}AM = \begin{bmatrix} 0 & -1 & 0 & 0 \\ 0 & 1 & 0 & 0 \\ 0 & 0 & 1 & 0 \\ 0 & 0 & 0 & 1 \end{bmatrix}, \] which is not symmetric. In fact, any diagonal \(A\) with distinct eigenvalues, together with any \(M\) with nonorthogonal columns, will provide a counterexample.

Keywords: matrix exponential, positive definite, singular value decomposition, orthonormal eigenvectors, similar matrices