6
$\begingroup$

How smooth are the singular values of a matrix $F$ in terms of entries of $F$? I am hoping for Lipschitz continuity, but was not able to find it.

$\endgroup$
1
  • $\begingroup$ I think the extreme (i.e smallest and largest) singular-values are $1$-Lipschitz functions of the input matrix (w.r.t operator norm). This can be seen via the variational characterization of these. $\endgroup$
    – dohmatob
    Commented Feb 20, 2021 at 15:21

3 Answers 3

8
$\begingroup$

The singular values of $F$ are the (square roots of ) eigenvalues of $F F^t,$ and the regularity of the latter have been studied half-to-death. See either T. Kato (perturbation theory of linear operators, ch. 1) or Golub-van Loan (Matrix Computations -- they almost certainly talk about singular values directly, without going through eigenvalues, but at worst talk about eigenvalues).

$\endgroup$
1
  • 2
    $\begingroup$ Thanks for the answer. Corollary 8.6.2. in Golub-van Loan gives exactly Lipschitz continuity. $\endgroup$ Commented Feb 10, 2012 at 20:48
1
$\begingroup$

Check these references

$\endgroup$
0
$\begingroup$

Since this got bumped to the front page...

Let $\sigma_i$ be the singular values of $A$ and $\tilde{\sigma}_i$ be those of $A+E$. Then, $|\sigma_i - \tilde{\sigma_i}| \leq \|E\|$, by Weyl's inequalities. That gives you 1-Lipschitz continuity.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .