matrix product, norm

125 Views Asked by At

Let $M_1, M_2, M_3$ are $n\times n$ matrices with real entries, and modulus of the eigenvalues are strictly less than one. For $x,y$ are any $n\times 1$ vector, Could anyone tell me given a suitable norm can I say this norm-is less than something? $$\|M_1M_3M_2x- M_2M_1M_3y\|< ?$$

I know that the Matrix norm has some relation with its spectral-radius, from there we know that each of the matrix norms is $<1$. Thanks for any help.

The norm one can use Hilbert metric(norm) too if that works.

1

There are 1 best solutions below

1
On

I'm not sure if this is the kind of result you are looking for, but if $M_2$ is invertible you can reason as follows.

Set $A:=M_1 M_3 M_2$ and $z:=M_2^{-1}y$. Then

\begin{align} M_1 M_3 M_2 x- M_2 M_1 M_3 y &= Ax - M_2 A z \\ &= (1-M_2) Ax + M_2 Ax - M_2 Az \\ &= (1-M_2) Ax + M_2 A(x-z) \end{align}

At this point you can use triangle inequality and your properties to infer

\begin{align} \Vert M_1 M_3 M_2 x- M_2 M_1 M_3 y \Vert \le & \,\Vert (1-M_2) \Vert \Vert x \Vert + \\ & + \Vert x-M_2^{-1}y \Vert \end{align}

NB I assumed the supremum norm, but any (submultiplicative) norm satisfying $\Vert AB x\Vert \le \Vert A\Vert \Vert B \Vert \Vert x\Vert$ would do