Norm estimation

Go to: Back / Home Page

The matrix-vector product by a QT matrix $A$ can be computed cheaply, and therefore the norm of $QT$ matrices is amenable to be computed by a power iteration.

The estimate norm is the spectral norm, and should be used to get rough estimates. Note that in general norm is relatively cheap for QT matrices, and can be a much more accurate alternative.

Contents

Syntax

Example

We use normest, and compare the result with the "exact" norm compute by the norm command.

A = cqt([3 2 1], [3 1]);
nrmA = norm(A, 2)
nrmEst = normest(A)

nrmA =

    6.8890


nrmEst =

    6.9094