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.

I was assigned to make a program that finds the largest, the N-largest, the smallest and the N-smallest eigenvalues of a symmetric matrix, using the Power Method. So far, I've been able to succesfully calculate the largest eigenvalue using the traditional Power Method, the N-largest using the Power Method with Deflation, and the smallest using the Inverse Iteration (the Inverse Iteration as described here in section 3-2: Iterative Methods).

But, right now I have no idea how to determine the N-smallest. I tried using the Inverse Iteration with Deflation to calculate the N-smallest but it is not working. When I calculate the second smallest (and so on...) I don't get the expected results, as if it's not possible to simply apply the Inverse Iteration with deflation. What am I missing? What should be the right way to calculate de N-smallest?

Your help is deeply appreciated.

share|improve this question
add comment

2 Answers

I assume by "largest" and "smallest" you mean in absolute value. The $n$'th smallest eigenvalue of $A$ is $\pm \sqrt{t - \mu}$ where $\mu$ is the $n$'th largest eigenvalue of $t I - A^2$ if $t$ is large enough. See how this moves if you replace $A$ by $A-\epsilon I$ for small $\epsilon$ and you can tell whether it's $+$ or $-$.

share|improve this answer
    
Yes, I mean in absolute value, thanks for your reply. I was totally unaware of that way of calculating the N-smallest eigenvalues of a matrix, and definitely will keep it in mind, next time I need to do it. Now, since my homework requires me to do it using the Power Method, I wonder if you know how to determine the N-smallest eigenvalues using the Power Method, I feel like it's easy and I'm missing something obvious but I can't see it. –  Robst Sep 17 '12 at 1:39
add comment

Is your matrix positive-definite ? (all eigenvalues positives). If yes, let $M$ be the largest eigenvalue. Find the N-largest eigenvalues of $A-MId$ by power method w/ deflation to find the N smallest eigenvalues of $A$.

share|improve this answer
add comment

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.