LogisticRegressionCV not compatible with LeaveOneGroupOut #8950
Comments
|
yes. sigh. Workaround for now, as long as you don't need this in a nested
context, is that you should be able to pass a list of (train, test) index
pairs directly to cv... I hope.
This is another request for sample properties etc (#4696). It perhaps also
suggests, from the perspective of maintainability that we should move away
from specialised CV objects, in preference for something like *SearchCV
with use_warm_start (#8230).
…On 29 May 2017 11:10 am, "Shudong Hao" ***@***.***> wrote:
In LogisticRegressionCV, the documentation says the argument cv could be
a cross-validation object from model_selection. However, when I pass
LeaveOneGroupOut as an argument, it has errors in it. I read the source
code. I think the problem is that in LogisticRegressionCV, when calling
the split method, it only pass X and y, but LeaveOneGroupOut requires an
additional argument groups, and we have nowhere to pass this argument.
This also raises problems when I use LogisticRegressionCV as the base
estimator for something like OneVsRestClassifier.
Here's a small example:
from sklearn.model_selection import LeaveOneGroupOut
from sklearn.linear_model import LogisticRegressionCV
X = np.array([[1, 2], [3, 4], [5, 6], [7, 8]])
y = np.array([1, 2, 1, 2])
groups = np.array([1, 1, 2, 2])
logo = LeaveOneGroupOut()
clf = LogisticRegressionCV(cv = logo)
clf.fit(X,y)
Error message:
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/site-packages/sklearn/linear_model/logistic.py", line 1581, in fit
folds = list(cv.split(X, y))
File "/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/site-packages/sklearn/model_selection/_split.py", line 91, in split
for test_index in self._iter_test_masks(X, y, groups):
File "/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/site-packages/sklearn/model_selection/_split.py", line 778, in _iter_test_masks
raise ValueError("The groups parameter should not be None")
ValueError: The groups parameter should not be None
------------------------------
Darwin-15.6.0-x86_64-i386-64bit
Python 3.4.3 (v3.4.3:9b73f1c3e601, Feb 23 2015, 02:52:03)
[GCC 4.2.1 (Apple Inc. build 5666) (dot 3)]
NumPy 1.11.1
SciPy 0.17.1
Scikit-Learn 0.18.1
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#8950>, or mute the
thread
<https://github.com/notifications/unsubscribe-auth/AAEz60Ct1f1FYarNz_EfQ7WYgXIrQ11kks5r-hrugaJpZM4No0XD>
.
|
|
The same here for GroupKFold and Gridsearch. The error I get is
|
|
|
In LogisticRegressionCV, the documentation says the argument
cvcould be a cross-validation object frommodel_selection. However, when I passLeaveOneGroupOutas an argument, it has errors in it. I read the source code. I think the problem is that inLogisticRegressionCV, when calling thesplitmethod, it only passXandy, butLeaveOneGroupOutrequires an additional argumentgroups, and we have nowhere to pass this argument. This also raises problems when I useLogisticRegressionCVas the base estimator for something likeOneVsRestClassifier.Here's a small example:
Error message:
Darwin-15.6.0-x86_64-i386-64bit
Python 3.4.3 (v3.4.3:9b73f1c3e601, Feb 23 2015, 02:52:03)
[GCC 4.2.1 (Apple Inc. build 5666) (dot 3)]
NumPy 1.11.1
SciPy 0.17.1
Scikit-Learn 0.18.1
The text was updated successfully, but these errors were encountered: