0

1

So, I know that $||AB|| \leq ||A||\cdot||B||$ (2-norm)

I'm doing a work on matrix algorithms and i seem to get as a result that $||A^TA|| = ||A||^2$

Does this always apply, or when and why does it happen?

flag
1 
 Which matrix norm are you using? – Igor Rivin Dec 13 2011 at 14:05
2-norm, sorry for not mentioning that – Koeneuze Dec 13 2011 at 14:15
I think this would be much better asked on math.stackexchange.com as it's not really research level. – Matthew Daws Dec 13 2011 at 14:53
2 
 ...and, if you do ask at stackexchange, it would be better to give, explicitly, what your definition of the 2-norm is. – Matthew Daws Dec 13 2011 at 14:59

closed as too localized by Martin Brandenburg, S. Sra, Matthew Daws, Igor Rivin, Denis Serre Dec 13 2011 at 15:21

1 Answer

2

If I'm not missing something, this is very obvious: $|A|^2_2$ is the maximal absolute value of the eigenvalues of $B = A^T A$. Thus $|A^TA|_2 =|B|_2$ is the square root of the maximal absolute value of the eigenvalues of $B^TB = B^2$. Since $B$ is diagonalizable, the eigenvalues of $B^2$ are just the squares of those of $B$.

link|flag
sorry, shouldn't this be a comment then? – S. Sra Dec 13 2011 at 14:50
3 
 looks like an answer to me... – Igor Rivin Dec 13 2011 at 14:55
Yes, but I thought implicit MO etiquette suggested not answering blatantly non research questions, where comments would suffice. anyhow, not a big deal ;-) – S. Sra Dec 13 2011 at 17:08

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