A second order perturbation expansion for small singular values (Q788792): Difference between revisions

Let X be an \(n\times p\) matrix and let \(\tilde X=X+E\) be a perturbation of X. In this paper the author presents a formula for the square of the smallest singular value of \(\tilde X\) with an error of order \(0(\| E\|^ 3)\). Unfortunately, it is not easy to decode the formula. I believe that the reference [2] on page 232 should be [1], but even so quite a lot of deskwork is needed before one can understand the notations used in the formula. Moreover it is not easy to see how the lemma, by which the formula follows, should be proved. The author refers to Theorem 4.1 and Theorem 4.2 of [1] (the latter does not exist and shall probably be Theorem 4.3), but it is not trivial to see how these theorems imply the lemma in the paper.