Is there a function in torch parallel to “scipy.sparse.linalg.svds”?
or Matlab’s “svds”?
If not, is there a way to calculate only the k largest SVD values in a different way?
Is there a function in torch parallel to “scipy.sparse.linalg.svds”?
or Matlab’s “svds”?
If not, is there a way to calculate only the k largest SVD values in a different way?
For a sparse matrix? Unfortunately no such function is available right now. We are always happy if you want to contribute one implementation.