LDA 在 Python 中作为 AdaBoost 的基础学习器

Question

我正在使用 AdaBoost 进行多 class class 化，将基础学习器作为判别式（线性或二次）。我在 scikit-learn 或任何其他库中找不到任何功能来实现这个，我该怎么做？

Answer 1

尽管 scikit-learn 的 AdaBoostClassifier 允许您选择一个基本估算器（参见 documentation), it requires the estimator to support sample_weight. Take a look at the source:

if not has_fit_parameter(self.base_estimator_, "sample_weight"):
    raise ValueError("%s doesn't support sample_weight."
                     % self.base_estimator_.__class__.__name__)

不幸的是，LinearDiscriminantAnalysis 和 QuadraticDiscriminantAnalysis 都不属于这一类。这是一个玩具示例：

from sklearn.ensemble import AdaBoostClassifier
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis as LDA
from sklearn.discriminant_analysis import QuadraticDiscriminantAnalysis as QDA
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split

iris = load_iris()
X_train, X_test, y_train, y_test = train_test_split(iris.data, iris.target)

clf = AdaBoostClassifier(base_estimator=LDA())
clf.fit(X_train, y_train)

您会看到如下回溯：

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "//anaconda/lib/python2.7/site-packages/sklearn/ensemble/weight_boosting.py", line 411, in fit
    return super(AdaBoostClassifier, self).fit(X, y, sample_weight)
  File "//anaconda/lib/python2.7/site-packages/sklearn/ensemble/weight_boosting.py", line 128, in fit
    self._validate_estimator()
  File "//anaconda/lib/python2.7/site-packages/sklearn/ensemble/weight_boosting.py", line 429, in _validate_estimator
    % self.base_estimator_.__class__.__name__)
ValueError: LinearDiscriminantAnalysis doesn't support sample_weight.

这是一项您不会在 scikit-learn 中绕过的要求。文档清楚地表明这是一项硬性要求：

"...Support for sample weighting is required, as well as proper classes_ and n_classes_ attributes."

但是，如果您只想使用集成，则始终可以使用 bagging 而不是 boosting：

from sklearn.ensemble import BaggingClassifier
clf = BaggingClassifier(base_estimator=LDA())
clf.fit(X_train, y_train)

LDA 在 Python 中作为 AdaBoost 的基础学习器

LDA as base learner for AdaBoost in Python

python

classification

adaboost

scikit-learn