7.1-exam3-1999¶
Quiz 3, 1999 · Section 7.1: Singular Values and Singular Vectors Source: exam3_1.pdf, pp. 32, 39
Problem¶
3 (24 pts.) Suppose that \(A\) is an invertible 3 by 3 matrix.
(a) Show me how to prove that \(x^TA^TAx\) is always positive if \(x\) is not the zero vector. Why will this fail if \(A\) is not invertible?
(b) Show me how to prove that \(A^TA\) is similar to \(AA^T\). Does it follow that these matrices have the same eigenvalues and eigenvectors?
(c) If the SVD is written in the usual form \(A = U\Sigma V^T\), what is the matrix \(A^TA\) (reduced to the simplest form)?
Solution
(a) \[ x^TA^TAx = (Ax)^T(Ax) = \|Ax\|^2 \] is positive unless \(Ax = 0\). If \(A\) is invertible, then \(Ax = 0\) forces \(x = 0\), so the expression is positive for every \(x \neq 0\). This fails when \(A\) is not invertible: then there is a nonzero \(x\) in the nullspace of \(A\), for which \(Ax = 0\) and hence \(x^TA^TAx = 0\). (b) \[ A(A^TA)A^{-1} = AA^T, \] so the matrices in parentheses are similar (the conjugating matrix is \(A\)). It follows that they have the same eigenvalues, but not necessarily the same eigenvectors. (c) \[ A^TA = (V\Sigma^TU^T)(U\Sigma V^T) = V\Sigma^T\Sigma V^T, \] the eigenvalue decomposition of \(A^TA\).Keywords: positive definite, similar matrices, singular value decomposition