Ilhan Polat
Ilhan Polat
@danielhanchen You are always welcome. The starting point can be reading the current code for SVD and see whether the actual call to the LAPACK subroutines can be replaced with...
[LAPACK did not include](http://icl.cs.utk.edu/lapack-forum/viewtopic.php?f=2&t=2646) the code that probably(!) matlab uses from LINPACK since nobody actually polished that code. There is Daniel Kressner's `QRUPDATE` package but I think the license is...
@ewmoore Care to chime in?
@xoviat matlab uses lapack (linpack actually) already for that.
Not only that but also even if it is added tomorrow to LAPACK, we won't be able to use it since we are barely trying to upgrade to LAPACK v3.4.1....
@ashimb9 LAPACK is like firmware, you change a single comma and people will start complaining about checksums not matching :) It's one of "the" most central libraries so people often...
@ashimb9 No problem at all. Just tried to summarize the intertwined story. Took me a while to figure it out too.
Thanks @ewmoore I'll have a look at them to see what I can do. In the meantime, I've found [DOI:10.1007/BF01933218](https://dx.doi.org/10.1007/BF01933218) as the starting point. Julia implementation looks a bit naive...
@ashimb9 Thanks. The paper I've linked to is from the second author of Alexander et al. There is also a comparison within the paper about that. That's why I've picked...
OK here is a basic prototype with very little attention to optimization ``` from numpy import zeros_like, empty_like, sqrt, spacing from scipy.linalg import LinAlgError def algoB(R, z, lower=False): """ Cholesky...