Benchmarking mir.ndslice + lubeck against numpy and Julia

Dennis dkorpel at
Sun Jan 12 15:53:16 UTC 2020

On Sunday, 12 January 2020 at 14:41:11 UTC, bachmeier wrote:
> You left out an important detail in your description. gesdd is 
> more efficient, but at the expense of being less accurate, and 
> can easily fail on you.

Interesting, I didn't actually know that. I just quoted the scipy 
documentation. I have only used SVD once before today, and it was 
on a 3x3 matrix for point set registration. Performance and 
accuracy weren't a problem.

It was wondering why Matlab and Octave default to the slower 
method too, but that explains why.

More information about the Digitalmars-d mailing list