python - sklearn 包装估计器会导致 get_params 丢失 self 错误

Question

我试图继承 BaseEstimator 和 MetaEstimatorMixin 来为 base_estimator 创建包装，但我遇到了问题。我试图遵循存储库中的 base_ensemble 代码，但没有帮助。在运行下面调用 check_estimator(Wrapper) 的测试时，我收到 TypeError: get_params() Missing 1 requiredpositional argument: 'self' 。根据文档，如果我继承自 BaseEstimator，则不必实现 get_params。看起来有些东西是一个类而不是一个实例，但我无法确定它。

from sklearn.base import BaseEstimator, ClassifierMixin, RegressorMixin, MetaEstimatorMixin, clone
from functools import lru_cache
import numpy as np
from sklearn.linear_model import LogisticRegression

'''
this is a module containing classes which wraps a classifier or a regressor sklearn estimator
'''


class Wrapper(BaseEstimator, MetaEstimatorMixin):
    def __init__(self, base_estimator=LogisticRegression, estimator_params=None):
        super().__init__()
        self.base_estimator = base_estimator
        self.estimator_params = estimator_params

    def fit(self, x, y):
        self.model = self._make_estimator().fit(x,y)

    def _make_estimator(self):
        """Make and configure a copy of the `base_estimator_` attribute.
        Warning: This method should be used to properly instantiate new
        sub-estimators. taken from sklearn github
        """
        estimator = self.base_estimator()
        estimator.set_params(**dict((p, getattr(self, p))
                                    for p in self.estimator_params))

        return estimator

    def predict(self, x):
        self.model.predict(x)


import unittest
from sklearn.utils.estimator_checks import check_estimator
class Test(unittest.TestCase):
    def test_check_estimator(self):
        check_estimator(Wrapper)

Answer 1

base_estimator 字段必须使用对象初始化，而不是类。

....
def __init__(self, base_estimator=LogisticRegression(), ...
....

发生错误是因为某些测试中使用了clone(safe=False)。

safe: boolean, optional
    If safe is false, clone will fall back to a deepcopy on objects
    that are not estimators.