Possible replacement for scipy svds
Scipy sparse.linalg.svds is a computation of the singular value decomposition of A, by a call to ARPACK on the matrix A'A or AA' depending on the size of the matrices. However, this computation can be unstable. An alternative is to compute the eigenvalues of the matrix [ 0 A; A' 0] which has eigendecomposition