How to pass parameter of only one part of a pipeline object to scikit learn?

Question

I need to pass the parameter sample_weightto mine RandomForestClassifieras follows:

X = np.array([[2.0, 2.0, 1.0, 0.0, 1.0, 3.0, 3.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0,
        1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 5.0, 3.0,
        2.0, '0'],
       [15.0, 2.0, 5.0, 5.0, 0.466666666667, 4.0, 3.0, 2.0, 0.0, 0.0, 0.0,
        0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 1.0, 0.0, 0.0,
        7.0, 14.0, 2.0, '0'],
       [3.0, 4.0, 3.0, 1.0, 1.33333333333, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0,
        0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0,
        9.0, 8.0, 2.0, '0'],
       [3.0, 2.0, 3.0, 0.0, 0.666666666667, 2.0, 2.0, 1.0, 0.0, 0.0, 0.0,
        0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0,
        5.0, 3.0, 1.0, '0']], dtype=object)

y = np.array([ 0.,  0.,  1.,  0.])

m = sklearn.ensemble.RandomForestClassifier(
        random_state=0, 
        oob_score=True, 
        n_estimators=100,
        min_samples_leaf=5, 
        max_depth=10)

m.fit(X, y, sample_weight=np.array([3,4,2,3]))

The above code works fine. Then I try to do this in a similar pipeline object using a pipeline object, not just a random forest:

m = sklearn.pipeline.Pipeline([
    ('feature_selection', sklearn.feature_selection.SelectKBest(
        score_func=sklearn.feature_selection.f_regression,
        k=25)),
    ('model', sklearn.ensemble.RandomForestClassifier(
        random_state=0, 
        oob_score=True, 
        n_estimators=500,
        min_samples_leaf=5, 
        max_depth=10))])

m.fit(X, y, sample_weight=np.array([3,4,2,3]))

Now this is interrupted in the method fitwith " ValueError: need more than 1 value to unpack".

ValueError                                Traceback (most recent call last)
<ipython-input-212-c4299f5b3008> in <module>()
     25         max_depth=10))])
     26 
---> 27 m.fit(X, y, sample_weights=np.array([3,4,2,3]))

/usr/local/lib/python2.7/dist-packages/sklearn/pipeline.pyc in fit(self, X, y, **fit_params)
    128         data, then fit the transformed data using the final estimator.
    129         """
--> 130         Xt, fit_params = self._pre_transform(X, y, **fit_params)
    131         self.steps[-1][-1].fit(Xt, y, **fit_params)
    132         return self

/usr/local/lib/python2.7/dist-packages/sklearn/pipeline.pyc in _pre_transform(self, X, y, **fit_params)
    113         fit_params_steps = dict((step, {}) for step, _ in self.steps)
    114         for pname, pval in six.iteritems(fit_params):
--> 115             step, param = pname.split('__', 1)
    116             fit_params_steps[step][param] = pval
    117         Xt = X

ValueError: need more than 1 value to unpack

I am using the sklearnversion 0.14.
I think the problem is that the step F selectionin the pipeline does not accept the argument for sample_weights. How to pass this parameter to only one step in the pipeline when I run " fit"? Thank.

Answer 1

:

, . , , , '__, .

, model__ kwargs, 'model':

m.fit(X, y, model__sample_weight=np.array([3,4,2,3]))

Answer 2

set_params .

m = sklearn.pipeline.Pipeline([
    ('feature_selection', sklearn.feature_selection.SelectKBest(
        score_func=sklearn.feature_selection.f_regression,
        k=25)),
    ('model', sklearn.ensemble.RandomForestClassifier(
        random_state=0, 
        oob_score=True, 
        n_estimators=500,
        min_samples_leaf=5, 
        max_depth=10))])

m.set_params(model__sample_weight=np.array([3,4,2,3]))

Answer 3

, @rovyko , stackoverflow, , .

:

Pipeline.set_params(model__sample_weight=np.array([3,4,2,3])

RandomForestClassifier.fit(). Pipeline.set_params() (), . RandomForestClassifier , sample_weight (. __init__() ). sample_weight RandomForestClassifier fit() RandomForestClassifier , : @ali_m,

m.fit(X, y, model__sample_weight=np.array([3,4,2,3])).