0
$\begingroup$

In my linear algebra self-study I can't understand why the vector p-norm for square matrices is submultiplicative for p=1,2 in a sum norm.

I found this Stack Overflow post that discusses the case $1 \leqslant p \leqslant 2$ Vector $p$-norm for square matrices is submultiplicative for $1 \le p \le 2$ but it says "For $p=1$ and $p=2$ the result follows easily from the Cauchy-Schwarz inequality", and I don't understand how it follows. I'm particularly unable to see how Cauchy-Schwarz Inequality works for $p=1$.

$\endgroup$
1
$\begingroup$

For $p=1$ you don't need the Cauchy-Schwarz inequality. This is solely a consequence of the triangle inequality (and positivity) of the norm: Denote $A = (a_{ij})_{i,j=1,\dotsc,n}$ and $B=(b_{kl})_{k,l=1,\dotsc,n}$. Then \begin{align*} \lVert AB\rVert_p &= \sum_{i,l=1}^n \left\lvert \sum_{j=1}^n a_{ij}b_{jl} \right\rvert \le \sum_{i,l=1}^n\sum_{j=1}^n\lvert a_{ij}\rvert\cdot \lvert b_{jl}\rvert\\ &\le \sum_{i,l=1}^n\sum_{j=1}^n\left(\lvert a_{ij}\rvert\cdot \lvert b_{jl}\rvert + \sum_{\substack{k=1\\k\neq j}}^n \lvert a_{ij}\rvert \cdot \lvert b_{kl}\rvert\right)\\ &= \sum_{i,j=1}^n\lvert a_{ij}\rvert \cdot \sum_{k,l=1}^n\lvert b_{kl}\rvert = \lVert A\rVert_p\cdot \lVert B\rVert_p. \end{align*}

$\endgroup$
  • $\begingroup$ Thank you! Can you pls explain how you went from the second inequality to the next expression? That part isn't clear. $\endgroup$ – Cogicero Jan 21 '16 at 7:24
  • 1
    $\begingroup$ I added the sum (on the right in the parentheses), which is positive, since every summand is positive. Hence the whole expression becomes larger. $\endgroup$ – Claudius Jan 21 '16 at 7:39
  • 1
    $\begingroup$ Ah sorry, I misread what you meant. The term in the big parentheses equals $\sum_{k=1}^n \lvert a_{ij}\rvert\cdot \lvert b_{kl}\rvert$. Rearranging the sum yields the next expression. $\endgroup$ – Claudius Jan 21 '16 at 7:44
  • $\begingroup$ Thanks! The line with the big parentheses, is that second sigma a typo? I mean, I thought the sum from j=1 to n is what became the big parenthesis? Sorry if I am just being slow... $\endgroup$ – Cogicero Jan 21 '16 at 7:49
  • 1
    $\begingroup$ No, the sum is not a typo. The third sum is exactly, what is need to get from $\lvert a_{ij}\rvert \cdot \lvert b_{jl}\rvert$ to $\sum_{k=1}^n\lvert a_{j}\rvert \cdot \lvert b_{kl}\rvert$. And the second Sigma is still the one from the first line. $\endgroup$ – Claudius Jan 21 '16 at 7:51

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.