On Friday, June 10, 2016 5:22:03 PM CDT [email protected] wrote: > In the case of the PostiveFactorizations.jl package, it appears > that the package is aimed at dense rather than sparse matrices.
No, it's just that sparse matrices haven't been implemented yet. Sparse is definitely on the horizon; the advantage is that it would not suffer from the "discard the factorization" problem you noted. For sparse, in principle it seems like that one could use the analysis routines of CHOLMOD to determine a good permutation; computing the actual factorization would then be fairly trivial, once one wraps one's head around the communication between CHOLMOD and julia. This is not in my immediate TODO list, but contributions are of course welcome! Best, --Tim
