[pymvpa] Trying to use Scikit-Learn PCA yields error

Thomas Nickson thomas.nickson at gmail.com
Thu Dec 18 15:43:34 UTC 2014


Hello All,

I'm trying to use the scikit-learn implementation of PCA because the
pymvpa/mdp version seems to end in a meomory error.

All I'm doing is wrapping the scikit-learn pca in the SKLTransformer like
so:

SKPCA = SKLTransformer(PCA())

and then attempting:

newds = ds.get_mapped(SKPCA)

but I get the error:

---------------------------------------------------------------------------TypeError
                                Traceback (most recent call
last)<ipython-input-191-3fb69e6da8d4> in <module>()     60 zscore(ds,
chunks_attr=None)     61 ---> 62 newds = ds.get_mapped(SKPCA)
/home/orkney_01/tnickson/Programming/pyVirtualEnv/lib/python2.6/site-packages/mvpa2/datasets/base.pyc
in get_mapped(self, mapper)     52         #  forward would lead to
different behavior     53         #mds = mapper(self)---> 54
mds = mapper.forward(self)     55         mds._append_mapper(mapper)
  56         return mds
/home/orkney_01/tnickson/Programming/pyVirtualEnv/lib/python2.6/site-packages/mvpa2/mappers/base.pyc
in forward(self, data)    185                 debug('MAP',
"Forward-map %s-shaped dataset through '%s'."    186
      % (data.shape, self))--> 187             return
self._forward_dataset(data)    188         else:    189             if
hasattr(data, 'ndim') and data.ndim < 2:
/home/orkney_01/tnickson/Programming/pyVirtualEnv/lib/python2.6/site-packages/mvpa2/mappers/skl_adaptor.pyc
in _forward_dataset(self, ds)     89             self._set_trained()
  90         else:---> 91             out = tf.transform(ds.samples,
self._get_y(ds))     92         return out
TypeError: transform() takes exactly 2 arguments (3 given)

SKPCA.train(ds) works fine but forward does not and throws the same error.

Any thoughts?

Thanks,

TOm
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pkg-exppsy-pymvpa/attachments/20141218/ad7b5354/attachment.html>


More information about the Pkg-ExpPsy-PyMVPA mailing list