Let's define the following subordinate norm of a $(NM \times NM)$ matrix A norm as follows

\begin{eqnarray*} ||A||_{2,b} = \mathrm{max}_{x \in \mathbb{C}^{NM}} \left \{ \frac{||A x||_b}{||x||_2} \right \} \end{eqnarray*}

where $||.||_2$ denotes the Euclidean norm of the $(NM \times 1)$ vector x, and where $||x||_b$ denotes the block maximum norm of the $(NM \times 1)$ vector x, i.e.

\begin{eqnarray*} ||x||_{2,b} = \mathrm{max}_{1 \leq k \leq N} \left \{ ||x_k||_2 \right \} \end{eqnarray*}

with $x = [x_1^T \, \, x_2^T \ldots \, x_N^T]$ and $x_k$ equal to a $(M \times 1)$ sub-vector of x for all $k$. It is well-known that

$||A||_2 \geq \rho(A)$

and that

$||A||_b \geq \rho(A)$

where $\rho(A)$ denotes the spectral radius of A (i.e. $\mathrm{max} |\lambda_i|$ with $\{\lambda_i\}$ denoting the eigenvalues of the matrix $A$), and where \begin{eqnarray*} ||A||_{2} = \mathrm{max}_{x \in \mathbb{C}^{NM}} \left \{ \frac{||A x||_2}{||x||_2} \right \} \end{eqnarray*} and \begin{eqnarray*} ||A||_{b} = \mathrm{max}_{x \in \mathbb{C}^{NM}} \left \{ \frac{||A x||_b}{||x||_b} \right \}. \end{eqnarray*} However, I would like to know if $||A||_{2,b} \geq \rho(A)$. By using the material that there is in the literature, I am only able to prove that $||A||_{2,b} \geq \rho(A)c$ with $c \leq 1$.


1 Answer 1


This ratio $\|A\|_{2,b}/\rho(A)$ can be arbitrarily small. Consider the $N\times N$ matrix $A$ all of whose entries are equal to $1$, and interpret this as a block matrix with blocks of size $M=1$. Then $\rho(A)=N$, $$ \|Ax\|_b = \max_j |x_1+\ldots + x_N| \le \|x\|_1 , $$ so $\|A\|_{2,b}\le \sup \|x\|_1/\|x\|_2 = \sqrt{N}$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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