numpy, scipy If you feel a little slow, check if blas, lapack, atlas are included.

In my case, only atlas was missing, and when I put in atlas and then rebuild it, it became really fast.

Confirmation method

import numpy
import scipy

numpy.__version__ # 1.9.3
numpy.show_config() # blas, lapack,You can check if it is built with atlas

scipy.__version__ # 0.16.0
scipy.show_config() #Same as for numpy

How fast

The following GaP-NMF python benchmark (although a maniac) is 1.75 times faster. See comments GaP-NMF short benchmark scripts for Python and Julia.

Julia is slower and I have tears

ATLAS installation

For ubuntu

sudo apt-get install libatlas-base-dev

If so, k

Recommended Posts

numpy, scipy If you feel a little slow, check if blas, lapack, atlas are included.
Use multithreaded BLAS / LAPACK with numpy / scipy