6.3-exam3-2014¶
Exam 3, 2014 · Section 6.3: Symmetric Positive Definite Matrices Source: exam3_2.pdf, pp. 81, 87
Problem¶
Problem 3. (a-c) Consider the matrix \(A = \begin{bmatrix} 2 & t & 0 \\ t & 2 & t \\ 0 & t & 2 \end{bmatrix}\) that depends on a parameter \(t\).
(a) Find all values of \(t\), for which the matrix \(A\) has 3 nonzero eigenvalues.
(b) Find all values of \(t\), for which the matrix \(A\) has 3 positive eigenvalues.
(c) Find all values of \(t\), for which the matrix \(A\) has 3 negative eigenvalues.
(d) Find a singular value decomposition \(B = U\Sigma V^T\) for the matrix \(B = \begin{bmatrix} 1 & 0 & 1 \\ 1 & 0 & 1 \end{bmatrix}\).
(e) Find orthonormal bases of the 4 fundamental subspaces of the matrix \(B\) from part (d).
Solution
(a) \(A\) has 3 nonzero eigenvalues whenever \(\det A = 8 - 4t^2\) is nonzero, i.e. whenever \(t \neq \pm\sqrt{2}\). (b) Since \(2 > 0\), \(A\) has 3 positive eigenvalues whenever \(\det A = 8 - 4t^2\) and \(\det \begin{bmatrix} 2 & t \\ t & 2 \end{bmatrix} = 4 - t^2\) are positive, i.e. whenever \(-\sqrt{2} < t < \sqrt{2}\). (c) Since \(\operatorname{tr}(A) = 6 > 0\), there are no values of \(t\) for which \(A\) has 3 negative eigenvalues. (d) We use the standard SVD algorithm. (a) \(BB^T = \begin{bmatrix} 2 & 2 \\ 2 & 2 \end{bmatrix}\) has orthonormal eigenvectors \(u_1 = (1, 1)^T/\sqrt{2}\) and \(u_2 = (1, -1)^T/\sqrt{2}\) with eigenvalues \(\lambda_1 = 4\) and \(\lambda_2 = 0\). (b) \(\sigma_1 = \sqrt{\lambda_1} = 2\) and \(v_1 = B^Tu_1/2 = (1, 0, 1)^T/\sqrt{2}\). (c) We can choose \(v_2 = (1, 0, -1)^T/\sqrt{2}\) and \(v_3 = (0, 1, 0)^T\), since they give an orthonormal basis for \(N(B^TB) = N(B)\). (d) This gives \(B = U\Sigma V^T\), where \[ U = \begin{bmatrix} 1 & 1 \\ 1 & -1 \end{bmatrix}/\sqrt{2}, \qquad \Sigma = \begin{bmatrix} 2 & 0 & 0 \\ 0 & 0 & 0 \end{bmatrix}, \qquad V = \begin{bmatrix} 1 & 1 & 0 \\ 0 & 0 & \sqrt{2} \\ 1 & -1 & 0 \end{bmatrix}/\sqrt{2}. \] (e) The fundamental subspaces are \[ C(B) = \left\langle (1, 1)^T/\sqrt{2} \right\rangle, \qquad N(B) = \left\langle (1, 0, -1)^T/\sqrt{2}, (0, 1, 0)^T \right\rangle, \] \[ C(B^T) = \left\langle (1, 0, 1)^T/\sqrt{2} \right\rangle, \qquad N(B^T) = \left\langle (1, -1)^T/\sqrt{2} \right\rangle. \]Keywords: eigenvalues, positive definite, determinant, singular value decomposition, fundamental subspaces