zclfly / pymf Goto Github PK

scipy >= v0.10
pymf 0.1.1
python > 2.6

pymf barfs at:

>>> import pymf
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/python2.7/site-packages/pymf/__init__.py", line 41, in <module>
    from .sivm_sgreedy import *
  File "/usr/lib/python2.7/site-packages/pymf/sivm_sgreedy.py", line 25, in <module>
    from vol import *
  File "/usr/lib/python2.7/site-packages/pymf/vol.py", line 20, in <module>
    from scipy import factorial
ImportError: cannot import name factorial


This can be fixed by editing pymf/vol.py and changing:

from scipy import factorial

to 

from scipy.misc import factorial

In vol.py you're trying to do

    from scipy import factorial

while for newer SciPy versions the correct way is to do

    from scipy.misc.common import factorial

This is also the way you're doing it in sivm_search.py.

What steps will reproduce the problem?
1. can this program support python 
csc_matrix?http://docs.scipy.org/doc/scipy/reference/generated/scipy.sparse.csc_
matrix.html

What steps will reproduce the problem?
1. python scripts/testpymf.py


What is the expected output? What do you see instead?

'module' object has no attribute 'eigen_symmetric'


What version of the product are you using? On what operating system?

PyMF-0.1.1, scipy 0.9.0, Fedora linux 64-bit


Please provide any additional information below.

replaced 'eigen_symmetric' with 'eigsh' according to 
https://github.com/marinkaz/MF/issues/6 can made the testpymf.py run though, 
not sure whether this is a good solution

What steps will reproduce the problem?
1. mf = pymf.CHNMF(data)
2. mf.factorize()
3. mf.W

What is the expected output? What do you see instead?

The documentation makes me believe, that I can find the columns of mf.W in 
data. This, however, is not the case. The data points closest to the columns of 
mf.W are stored in mf.Wmapped and are thus probably not used for the 
factorization.

What version of the product are you using? On what operating system?

0.1.9 checked out of subversion 16.Nov. 2012

Please provide any additional information below.

Is this a bug, or intended behaviour? I mean, the way Wmapped is created 
(selecting the closest data point), it is very probable, that the columns of 
Wmapped are not on the convex hull of the data.