The von Mises iteration applies powers of a matrix to
a start vector .
The resulting normalized sequence
will generally approximate a dominant eigenvector.
Sufficient for convergence is that
is diagonalizable and has an eigenvalue
with largest absolute value.
Then, for any vector
with a nontrivial component in the eigenspace of
,
with
and
an eigenvalue of with second largest absolute value.