7.1-exam3-2001¶
Quiz 3, 2001 · Section 7.1: Singular Values and Singular Vectors Source: exam3_1.pdf, p. 55
Problem¶
2 (32 pts.) (a) Choose \(x\) and \(y\) so that
\[
M = \begin{bmatrix} 1/2 & x \\ y & 1/4 \end{bmatrix}
\]
is a Markov matrix. (4p)
Compute the steady state eigenvector \(x_1\) of unit length. (That is, \(\|x_1\| = 1\)). (8p)
(b) Is
\[
A = \begin{bmatrix} -1 & 2 \\ 2 & -1 \end{bmatrix}
\]
positive definite? (4p)
Find the singular value decomposition of \(A\). (16p)
Solution
(a) A Markov matrix has column sums equal to 1, so \(1/2 + y = 1\) and \(x + 1/4 = 1\): \(x = 3/4\), \(y = 1/2\). The steady state is the eigenvector for \(\lambda = 1\). From \((M - I)x_1 = 0\), \[ \begin{bmatrix} -1/2 & 3/4 \\ 1/2 & -3/4 \end{bmatrix} x_1 = 0 \quad \Rightarrow \quad x_1 = \begin{bmatrix} 3 \\ 2 \end{bmatrix} \] up to scaling. Normalized to unit length: \[ x_1 = \frac{1}{\sqrt{13}} \begin{bmatrix} 3 \\ 2 \end{bmatrix}. \] (b) \(A\) is not positive definite: the leading \(1 \times 1\) determinant is \(-1 < 0\) (equivalently, the eigenvalues are \(-3\) and \(1\), one of which is negative). For the SVD, compute \[ A^TA = \begin{bmatrix} -1 & 2 \\ 2 & -1 \end{bmatrix} \begin{bmatrix} -1 & 2 \\ 2 & -1 \end{bmatrix} = \begin{bmatrix} 5 & -4 \\ -4 & 5 \end{bmatrix}, \] whose eigenvalues are \(9\) and \(1\). The singular values are \(3\) and \(1\), with orthonormal eigenvectors \((1, -1)/\sqrt{2}\) (for \(9\)) and \((1, 1)/\sqrt{2}\) (for \(1\)). Hence \[ U = \frac{1}{\sqrt{2}} \begin{bmatrix} -1 & 1 \\ 1 & 1 \end{bmatrix}, \qquad \Sigma = \begin{bmatrix} 3 & 0 \\ 0 & 1 \end{bmatrix}, \qquad V = \frac{1}{\sqrt{2}} \begin{bmatrix} 1 & 1 \\ -1 & 1 \end{bmatrix}, \] and one checks that \[ U\Sigma V^T = \frac{1}{2} \begin{bmatrix} -1 & 1 \\ 1 & 1 \end{bmatrix} \begin{bmatrix} 3 & 0 \\ 0 & 1 \end{bmatrix} \begin{bmatrix} 1 & -1 \\ 1 & 1 \end{bmatrix} = \begin{bmatrix} -1 & 2 \\ 2 & -1 \end{bmatrix} = A. \]Keywords: Markov matrix, steady state, singular value decomposition, positive definite