next up previous contents index
Next: Inverse Iteration Up: Single- and Multiple-Vector Iterations Previous: Single- and Multiple-Vector Iterations   Contents   Index


Power Method

The power method, described in Algorithm 4.1, can be used to solve the NHEP without any apparent change.


\begin{algorithm}{The Power Method for the Computation of $\lambda_{\max}(A)$} ... ...\\ (4) \> \> \> Else $z := y /\Vert y\Vert _2$. \end{tabbing}} \end{algorithm}
Under conditions similar to those in the Hermitian case, the power method for the non-Hermitian matrix $A$ converges to $\lambda_{\max}(A)$, the largest eigenvalue in magnitude. The convergence rate depends on the ratio $\vert \lambda_2 / \lambda_{\max}\vert$, where $\lambda_2$ is the second largest eigenvalue of $A$ in magnitude. For detailed discussions of the power method, see Wilkinson [457], Golub and Van Loan [198], and Demmel [114].


Susan Blackford 2000-11-20