>From your code it doesn't seem like it, but are you using
multiprocessing (ie. n_jobs > 1)? It causes issues on certain
configurations.
Either way, try to pass `verbose=2` to the grid search constructor.
Yours,
Vlad
On Wed, Jul 3, 2013 at 9:36 PM, Josh Wasserstein <ribonucle...@gmail.com> wrote:
> This is odd. I can successfully run the example `grid_search_digits.py`.
> However, I am unable to do a grid search on my own data.
>
> I have the following setup:
> ===============
> import sklearn
> from sklearn.svm import SVC
> from sklearn.grid_search import GridSearchCV
> from sklearn.cross_validation import LeaveOneOut
> from sklearn.metrics import auc_score
>
> # ... Build X and y ....
>
> tuned_parameters = [{'kernel': ['rbf'], 'gamma': [1e-3, 1e-4],
> 'C': [1, 10, 100, 1000]},
> {'kernel': ['linear'], 'C': [1, 10, 100, 1000]}]
>
> loo = LeaveOneOut(len(y))
> clf = GridSearchCV(SVC(C=1), tuned_parameters, score_func=auc_score)
> clf.fit(X, y, cv=loo)
> ....
> print clf.best_estimator_
> ....
> ===============
> But I never get passed `clf.fit` (I left it run for ~1hr).
>
> I have tried also with
>
> clf.fit(X, y, cv=10)
>
> and with
>
> skf = StratifiedKFold(y,2)
> clf.fit(X, y, cv=skf)
>
> and had the same problem (it never finishes the clf.fit statement). My data
> is simple:
>
> > X.shape
> (27,26)
>
> > y.shape
> 5
>
> > y.dtype
> dtype('int64')
>
>
> >?y
> Type: ndarray
> String Form:[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1]
> Length: 27
> File:
> /home/jacob04/opt/python/numpy/numpy-1.7.1/lib/python2.7/site-
> packages/numpy/__init__.py
> Docstring: <no docstring>
> Class Docstring:
> ndarray(shape, dtype=float, buffer=None, offset=0,
> strides=None, order=None)
>
> > ?X
> Type: ndarray
> String Form:
> [[ -3.61238468e+03 -3.61253920e+03 -3.61290196e+03
> -3.61326679e+03
> 7.84590361e+02 0.0000 <...> 0000e+00 2.22389150e+00
> 2.53252959e+00
> 2.11606216e+00 -1.99613432e+05 -1.99564828e+05]]
> Length: 27
> File:
> /home/jacob04/opt/python/numpy/numpy-1.7.1/lib/python2.7/site-
> packages/numpy/__init__.py
> Docstring: <no docstring>
> Class Docstring:
> ndarray(shape, dtype=float, buffer=None, offset=0,
> strides=None, order=None)
>
> This is all with the latest version of scikit-learn (0.13.1) and:
>
> $ pip freeze
> Cython==0.19.1
> PIL==1.1.7
> PyXB==1.2.2
> PyYAML==3.10
> argparse==1.2.1
> distribute==0.6.34
> epc==0.0.5
> ipython==0.13.2
> jedi==0.6.0
> matplotlib==1.3.x
> nltk==2.0.4
> nose==1.3.0
> numexpr==2.1
> numpy==1.7.1
> pandas==0.11.0
> pyparsing==1.5.7
> python-dateutil==2.1
> pytz==2013b
> rpy2==2.3.1
> scikit-learn==0.13.1
> scipy==0.12.0
> sexpdata==0.0.3
> six==1.3.0
> stemming==1.0.1
> -e
> git+https://github.com/PyTables/PyTables.git@df7b20444b0737cf34686b5d88b4e674ec85575b#egg=tables-dev
> tornado==3.0.1
> wsgiref==0.1.2
>
> Thanks,
>
> Jacob
>
> PS: This thread is based on the following StackOverflow post:
> http://stackoverflow.com/questions/17455302/clf-fit-freezes-on-small-dataset-in-scikit-learn
>
>
> ------------------------------------------------------------------------------
> This SF.net email is sponsored by Windows:
>
> Build for Windows Store.
>
> http://p.sf.net/sfu/windows-dev2dev
> _______________________________________________
> Scikit-learn-general mailing list
> Scikit-learn-general@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>
------------------------------------------------------------------------------
This SF.net email is sponsored by Windows:
Build for Windows Store.
http://p.sf.net/sfu/windows-dev2dev
_______________________________________________
Scikit-learn-general mailing list
Scikit-learn-general@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
