how can i find k-eigenvalues faster than eig for hermitian dense matrix

Hi! eigs is not faster than eig for large hermitian dense matrix. I need many eigenvalues. For a matrix n > 2000 elements, i need to find the 250 largest eigenvalues. I think that the divide and conquer method or bissection method.

are there routines( or package) for these methods?

thanks

[Matt J]

Any special structure to the matrix, besides that it is Hermitian? It would be great if it were circulant, obviously.

Answer 1

[Andrew Knyazev]

try http://www.mathworks.com/matlabcentral/fileexchange/48-lobpcg-m