Skip to content

6.2-exam3-2000-b

Quiz 3, 2000 · Section 6.2: Diagonalizing a Matrix Source: exam3_1.pdf, pp. 42, 45

Problem

2 (40 pts.) (a) Suppose the diagonalization \(A = S\Lambda S^{-1}\) is exactly the same as the singular value decomposition \(A = U\Sigma V^T\) (so \(S = U = V\) and \(\Lambda = \Sigma\)). What information does this give about \(A\)? Can it be singular?

(b) What are the eigenvalues of a 3 by 3 Markov projection matrix that has trace 2? Create one matrix that has these properties.

(c) Here is a matrix with orthogonal columns. Find its SVD \(A = U\Sigma V^T\).

\[ A = \begin{bmatrix} 3 & 0 \\ 4 & 0 \\ 0 & 7 \end{bmatrix} \]

(d) Suppose \(A\) is similar to a 3 by 3 matrix \(B\) that has eigenvalues 1, 1, 2. What can you say about

  1. the eigenvalues of \(A\)
  2. diagonalizability of \(A\)
  3. symmetry of \(A\)
  4. positive definiteness of \(A\)

In each of (2) (3) and (4) decide if \(A\) can't have or might have or must have this property.

Solution (a) \(S^{-1} = S^T\), so \(A = S\Lambda S^T\) is symmetric. Singular values are always nonnegative, so from \(\Lambda = \Sigma\) the eigenvalues of \(A\) are nonnegative, so \(A\) is symmetric positive semidefinite. It can be singular (the all zeros matrix is an example). (b) The eigenvalues of a projection matrix are either 0 or 1, and their sum is 2, so they must be 1, 1, 0. For example \[ \begin{bmatrix} 1 & 0 & 0 \\ 0 & 1/2 & 1/2 \\ 0 & 1/2 & 1/2 \end{bmatrix}. \] (c) \(A^TA = \begin{bmatrix} 25 & 0 \\ 0 & 49 \end{bmatrix}\), so the singular values are 7 and 5. Thus \[ V = \begin{bmatrix} 0 & 1 \\ 1 & 0 \end{bmatrix} \] and \[ A = \begin{bmatrix} 0 & 3/5 & -4/5 \\ 0 & 4/5 & 3/5 \\ 1 & 0 & 0 \end{bmatrix} \begin{bmatrix} 7 & 0 \\ 0 & 5 \\ 0 & 0 \end{bmatrix} \begin{bmatrix} 0 & 1 \\ 1 & 0 \end{bmatrix}. \] (d) 1. The eigenvalues of \(A\) are 1, 1, 2 — the same as the eigenvalues of \(B\). 2. \(A\) might or might not be diagonalizable. 3. \(A\) might or might not be symmetric. 4. \(A\) definitely has positive eigenvalues. However it might not be symmetric, so \(A\) might or might not be positive definite.

Keywords: singular value decomposition, Markov matrix, projection matrix, similar matrices