如何将参数仅传递给 scikit 学习中管道对象的一部分?

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

我需要向我的 RandomForestClassifier 传递一个参数 sample_weight,如下所示:

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]))

上面的代码工作得很好。然后,我尝试像这样在管道对象中执行此操作,使用管道对象而不是仅使用随机森林:

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]))

现在在 fit 方法中使用“ValueError: need more than 1 value to unpack”中断。

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

我正在使用 sklearn 版本 0.14.
我认为问题在于管道中的 F selection 步骤没有接受 sample_weights 的参数。如何使用 I 运行“fit”将此参数仅传递给管道中的一个步骤?谢谢。

From the documentation:

The purpose of the pipeline is to assemble several steps that can be cross-validated together while setting different parameters. For this, it enables setting parameters of the various steps using their names and the parameter name separated by a ‘__’, as in the example below.

因此,您只需在要传递给 'model' 步骤的任何拟合参数 kwargs 前面插入 model__

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

您也可以使用方法 set_params 并在步骤名称前添加。

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]))

希望我可以在上面的@rovyko post 上发表评论,而不是单独回答,但我还没有足够的 Whosebug 声誉来发表评论,所以在这里。

您不能使用:

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

RandomForestClassifier.fit()方法设置参数。 Pipeline.set_params() 如代码中所示 (here) is only for initialization parameters for individual steps in the Pipeline. RandomForestClassifier has no initialization parameter called sample_weight (see its __init__() method here)。 sample_weight 实际上是 RandomForestClassifierfit() 方法的输入参数,因此只能通过正确标记答案中提供的方法设置 @ali_m,即,

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