Reference: https://www.cs.emory.edu/~benzi/Web_papers/copper.pdf
I am interested in building a RIF preconditioner for least square problem based on the C-orthogonalization in the above paper: start with e1-en (cols of identity matrix), using conjugate Gram–Schmidt process with inner product <u,v>_c = u^T @ C @ v = u^T @ A^T @ A @ v, then apply some dropping strategy to get a sparse incomplete decomposition.
I have no experience to work directly on the sparse matrix using scipy. Anyone could provide some similar code sample?
Thanks.
Related algorithms: algorithm
