Closed
Description
I suppose this is more of a feature request than anything else. There are several implementations of PCA that can compute the decomposition on noncentered, sparse data, while the implementation here does not support sparse matrices at all.
A matlab implementation can be found here and a Python implementation here. So far, I've been using the Python implementation, but it's missing some things and will eventually be deprecated (facebookarchive/fbpca#9).
I haven't looked at the code or math too much, but as far as I'm aware, it's just a matter of adding a term to randomized_range_finder to account for centering.
Is this something that you guys are aware of, and is anyone working on this? This would an awesome feature to have.