Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Let $A \in \mathbb{R}^{n \times n}$ be a positive definite matrix and let $a > 0$ be some constant.

I am interested in the following quantity $$ \hat x = {\rm arg} \max_{x \in R^n}\ x' \left( \frac{1}{a + \|x\|^2_2} A \right) x $$ Is it possible to relate $\|\hat x\|_2^2$ to the eigenvalue and $\hat x$ to the eigenvector of $A$.

Furthermore, suppose that $E \in \mathbb{R}^{p \times p}$ is such that $A + E$ is still positive definite and $\|E\|$ is small in some norm (say, spectral). Consider $$ \hat y = {\rm arg} \max_{y \in R^n}\ y' \left( \frac{1}{a + \|y\|^2_2} (A+E) \right) y $$ Are there known results that can be used to quantify $| \|\hat x\|_2^2 - \|\hat y\|_2^2 |$? I suppose the norms should be close if the perturbation $E$ is small, but don't know how to quantify this notion.

share|improve this question
    
The minimum is always attained for x = 0. Are you missing something or is it me? –  Michael Ulm Nov 9 '10 at 11:42
    
Indeed, the minimum is always attained. I meant to ask about the maximum. Thanks for pointing this out. –  mkolar Nov 9 '10 at 11:59

1 Answer 1

up vote 3 down vote accepted

Let $\lambda$ be the biggest eigenvalue of A and $x_{\lambda}$ a corresponding eigenvector. Since $x' A x \leq \lambda \left\| x \right\|^2$, we obtain

$$ x' \left( \frac{1}{a + \left\| x \right\|_2^2} A \right) x \leq \frac{\lambda \left\| x \right\|^2}{a + \left\| x \right\|^2} < \lambda $$

But for $x = t x_{\lambda}$ and $t \to \infty$, the expression goes to $\lambda$.

This means the maximum is never attained.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.