I read this in my textbook have tried working through it - I keep getting max 2-norm(Ax), which is just the magnitude of Ax.

How should I do this proof? (note, this is not for homework, I'm just trying to understand why as no proof is provided).


From your question, I assume you mean the operator norm with respect to the 2-norm. Let $A = {\rm diag}(\lambda_1, \ldots, \lambda_n)$ a diagonal matrix. We have \begin{align*} \|A\| &= \max_{\|x\|_2 = 1} \|Ax\|_2\\ &= \max_{\|x\|_2 = 1} \left(\sum_{i=1}^n \lambda_i^2x_i^2\right)^{1/2}\\ &\le \max_{\|x\|_2 = 1} \max_i|\lambda_i| \left(\sum_{i=1}^n x_i^2\right)^{1/2}\tag 1\\ &= \max_i|\lambda_i| \cdot \max_{\|x\|_2 = 1} \|x\|_2\\ &= \max_i|\lambda_i| \end{align*} For (1), we can argue as follows: For each $i$: we have $x_i^2 \ge 0$, hence multiplying the inequality $|\lambda_i|^2 \le \max_i |\lambda_i|^2$ by $x_i^2$, we get $x_i^2|\lambda_i|^2 \le (\max_i|\lambda_i|)^2 x_i^2$. Now sum and take the square root of both sides.

On the other hand, let $x$ be an eigenvector, corresponding to the largest eigenvalue. Then $$ \|Ax\|_2 = \max_i|\lambda_i| \cdot \|x\|_2 $$ This gives $\|A\| = \max_i |\lambda_i|$.

  • $\begingroup$ so that line where you introduce the inequality - what inequality are you using to do that? $\endgroup$ – praks5432 Jun 6 '13 at 22:32
  • $\begingroup$ oh or is it just because you take the max lambda? $\endgroup$ – praks5432 Jun 6 '13 at 22:34
  • $\begingroup$ @praks5432 Added something, hope this helps. $\endgroup$ – martini Jun 6 '13 at 22:36
  • $\begingroup$ @praks5432 Yes, exactly $\endgroup$ – martini Jun 6 '13 at 22:36
  • $\begingroup$ How can we be sure that $\lambda_i$ is an eigenvalue $\endgroup$ – Tyler Hilton Dec 3 '14 at 22:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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