suppose we have a matrix operator $A=(a_{ij})_{n\times n}:(R^n,\|\cdot\|_{\infty})\to(R^n,\|\cdot\|_{1})$, how to calculate $\|A\|$? or if there really is an exactly closed form for $\|A\|$?

if $x\in (R^n,\|\cdot\|_{\infty})$, then $\|x\|_{\infty}=\max \{|x_n|\}$, and by the definition $$\|A\|=\sup_{{x\in R^n}\atop{\|x\|_{\infty}=1}}\frac{\|Ax\|_1}{\|x\|_{\infty}} =\sup_{{x\in R^n}\atop{\|x\|_{\infty}=1}}\frac{\sum_{i=1}^n\left|\sum_{j=1}^na_{ij}x_j\right|}{\max\{|x_n|\}}.$$

before then, I have thought about the orthogonal transform, such that $P^{-1}AP=J$, $J$ is the Jordan form of $A$. but unfortunately, the norm of $J$ is not the same with $A$, because there are examples with the same Jordan form that have different norm, i.e. $A_1=\left(\begin{array}{cc}1 & n \\ & 2\end{array}\right)$ or $A_2=\left(\begin{array}{ccc}1 & x & y \\ & 2 & z\\ & & 3\end{array}\right)$, the calculation gives $\|A_1\|=n+3$ and $\|A_2\|=6+x+y+z$ for positive $x,y,z$.

So I have to go back to the definition of the norm. But the supremum $$\sup_{{x\in R^n}\atop{\|x\|_{\infty}=1}}\frac{\sum_{i=1}^n\left|\sum_{j=1}^na_{ij}x_j\right|}{\max\{|x_n|\}}$$ is not easy to calculate as in $(R^n,\|\cdot\|_{1})\to(R^n,\|\cdot\|_{1})$: $$\sup_{{x\in R^n}\atop{\|x\|_{\infty}=1}}\frac{\sum_{i=1}^n\left|\sum_{j=1}^na_{ij}x_j\right|}{\sum_{i=1}^n|x_i|} \le \sup_{{x\in R^n}\atop{\|x\|_{\infty}=1}}\frac{\sum_{j=1}^n\sup_j\sum_{i=1}^n|a_{ij}|\cdot |x_j|}{\sum_{i=1}^n|x_i|} = \sup_j\sum_{i=1}^n|a_{ij}|.$$

finally, I've tried some other examples such as $A=\left(\begin{array}{cc}1 & 2 \\ -3 & 4\end{array}\right)$, $x=(x_1,x_2)^T$ with $|x_i|\le 1$. then $\|Ax\|_1=|x_1+2x_2|+|3x_1-4x_2|\le8$, the equality achieved at $x_1=-1$ and $x_2=1$; for $A=\left(\begin{array}{cc}1 & 2 \\ -3 & -4\end{array}\right)$, $\|A\|=10$.


If the columns of the matrix $A$ are such that all elements of the column have the same sign then in that case the norm is simple to calculate because

$\frac{||Ax||_1}{||x||_\infty}\leq \sum_{i,j}^n |a_{ij}|$

and if you consider the vector

$x_0=(sign(a_{11}),\dots , sign(a_{1,n}))$

then you have that

$\frac{||Ax_0||_1}{||x_0||_\infty}=\sum_{i=1}^n(|\sum_{j=1}^n a_{ij}sign(a_{1j})|=$

$=\sum_{i=1}^n(|\sum_{j=1}^n a_{ij}sign(a_{ij})|=$

$= \sum_{i,j}^n |a_{ij}|$

so you have that

$||A||= \sum_{i,j}^n |a_{ij}|$

for example if you consider

$\left[\begin{matrix}\frac{1}{9} & -\frac{1}{3} & \frac{1}{2} \\ \frac{1}{2} & -3 & \frac{1}{2} \\ \frac{1}{2} &- \frac{1}{2} & \frac{1}{2} \end{matrix}\right]$

then its norm will be


  • 1
    $\begingroup$ thanks for your answer, yes, this is a simple situation for my question, so the result for the general one is hard to imagine. $\endgroup$ – Larry Eppes Jun 15 at 8:51
  • $\begingroup$ Exactly, infact it is difficult think to a closed formula in which if the column of the matrix have all the same sign then you have that that formula becomes my simple formula. I don’t know how to generalize better. $\endgroup$ – Federico Fallucca Jun 15 at 8:53

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.