Re: another sparse eigenvector question

rif wrote:
> 

> The matrix is 60K by 60K.  Assuming that I could store the
> elements of the matrix as floats and that I only needed the lower
> triangle, I would need over 7 GB to store a dense matrix of this size.

But a Cholesky factorization of a sparse matrix with only 8 nonzeros 
per row cannot be dense; it has at least 1000 sparse rows, no matter
how poorly you do the factorization. Of course for efficiency you 
need to reorder the matrix. 

If you order the rows and columns suitably
(AMD will gnenerate a feather-like structure pointing from
top left top to bottom right) you can probably find a quite 
sparse factorization. 

Using the sparse materix facilities of Matlab makes the task 
quite easy for you.

Arnold Neumaier