svd of a VERY LARGE sparse matrix

Question

I have a 75000 x 75000 sparse matrix, and I'm interested in computing the full SVD. Whenever I use:

[U,D,V] = svds(A,k)

I get an out of memory error for k larger than 200. Is there a tracable way for computing the matrices U, V and D in this case?

Answer 1

The approximate matrix decomposition into a product of three matrices is called CUR. However I'm not sure the matlab contains its implementation.