6.3-final-2015¶
Final Exam, 2015 · Section 6.3: Symmetric Positive Definite Matrices Source: finalexam_2.pdf, p. 160
Problem¶
8 (5+5+5 pts.)
(a) Factor this symmetric matrix into \(A = U^TU\) where \(U\) is upper triangular:
\[
A = \begin{bmatrix} 1 & 1 & 1 \\ 1 & 2 & 2 \\ 1 & 2 & 3 \end{bmatrix}.
\]
(b) Show by two different tests that \(A\) is symmetric positive definite.
(c) Find and explain an upper bound on the eigenvalues of \(A\). Find and explain a (positive) lower bound on those eigenvalues if you know that
\[
A^{-1} = \begin{bmatrix} 2 & -1 & 0 \\ -1 & 2 & -1 \\ 0 & -1 & 1 \end{bmatrix}.
\]
Solution
(a) Elimination gives the factorization \(A = LU\) with \[ A = \begin{bmatrix} 1 & 0 & 0 \\ 1 & 1 & 0 \\ 1 & 1 & 1 \end{bmatrix} \begin{bmatrix} 1 & 1 & 1 \\ 0 & 1 & 1 \\ 0 & 0 & 1 \end{bmatrix}. \] Here \(L = U^T\), so \[ A = U^TU \quad \text{with} \quad U = \begin{bmatrix} 1 & 1 & 1 \\ 0 & 1 & 1 \\ 0 & 0 & 1 \end{bmatrix}. \] (b) Two tests (any two of these work): - \(A = U^TU\) with \(U\) invertible (each diagonal entry of \(U\) is 1); - the energy test: \(x^TAx = x^TU^TUx = \|Ux\|^2 > 0\) for \(x \neq 0\), since \(U\) is invertible; - the pivots of \(A\) are the pivots of \(U\), namely \(1, 1, 1\), all positive; - the upper left determinants are \(1, 1, 1\), all positive. (c) The eigenvalues \(\lambda_1, \lambda_2, \lambda_3\) are positive and sum to the trace of \(A\): \[ \lambda_1 + \lambda_2 + \lambda_3 = \mathrm{trace}(A) = 6, \] so each satisfies \[ \lambda_i < 6 \] (an upper bound; the largest eigenvalue is at most 6). The eigenvalues of \(A^{-1}\) are \(1/\lambda_1, 1/\lambda_2, 1/\lambda_3\), and their sum is the trace of \(A^{-1}\): \[ \frac{1}{\lambda_1} + \frac{1}{\lambda_2} + \frac{1}{\lambda_3} = \mathrm{trace}(A^{-1}) = 2 + 2 + 1 = 5. \] Since each \(1/\lambda_i > 0\), each is at most 5, so each eigenvalue satisfies \[ \lambda_i \ge \frac{1}{5}. \] Thus the eigenvalues of \(A\) lie in the interval \([1/5, 6)\).Keywords: Cholesky factorization, positive definite, eigenvalue bounds, trace