next up previous contents index
Next: Selective Reorthogonalization. Up: Reorthogonalization Previous: Reorthogonalization   Contents   Index

Full Reorthogonalization.

We could play it safe and do a full reorthogonalization either by modified or classical Gram-Schmidt, in the latter case computing the orthogonalization coefficients in a vector

\begin{displaymath}h=V_j^{\ast} v_{j+1}\end{displaymath}

and then subtracting

\begin{displaymath}v'_{j+1}=v_{j+1}-V_jh\end{displaymath}

to get the improved vector $v'_{j+1}$. If the norm is decreased by a nontrivial amount, say $\Vert v'_{j+1}\Vert<\frac{1}{\sqrt{2}}\Vert v_{j+1}\Vert$, this will have to be repeated, but this will hardly ever happen when treating properly computed Lanczos vectors.



Susan Blackford 2000-11-20