[MRG] ENH: Make StackingRegressor support Multioutput #27704
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@OmarManzoor would you maybe have time to have a look here? |
OmarManzoor
reviewed
Apr 18, 2024
There was a problem hiding this comment.
Thank you for the PR @hmasdev. A few comments. I think we will also need a changelog entry for this.
OmarManzoor
reviewed
Apr 25, 2024
Comment on lines
888
to
892
| # NOTE: In this case the estimator can predict almost exactly the target | ||
| assert_allclose( | ||
| y_pred, | ||
| # NOTE: when the target is 2D but with a single output, | ||
| # the predictions are 1D because of column_or_1d |
There was a problem hiding this comment.
Suggested change
| # NOTE: In this case the estimator can predict almost exactly the target | |
| assert_allclose( | |
| y_pred, | |
| # NOTE: when the target is 2D but with a single output, | |
| # the predictions are 1D because of column_or_1d | |
| # NOTE: In this case the estimator can predict almost exactly the target. | |
| # When the target is 2D but with a single output the predictions are 1D | |
| # because of column_or_1d | |
| assert_allclose( | |
| y_pred, |
| rtol=acceptable_relative_tolerance, | ||
| atol=acceptable_aboslute_tolerance, | ||
| ) | ||
| # transform |
There was a problem hiding this comment.
Suggested change
| # transform |
| ) | ||
|
|
||
| reg.fit(X_train, y_train) | ||
| # predict |
There was a problem hiding this comment.
Suggested change
| # predict |
|
|
||
|
|
||
| def test_stacking_regressor_multioutput_with_passthrough(): | ||
| """Check that a stacking regressor with multioutput works""" |
There was a problem hiding this comment.
Suggested change
| """Check that a stacking regressor with multioutput works""" | |
| """Check that a stacking regressor with passthrough works with multioutput""" |
| rtol=acceptable_relative_tolerance, | ||
| atol=acceptable_aboslute_tolerance, | ||
| ) | ||
| # transform |
There was a problem hiding this comment.
Suggested change
| # transform |
|
|
||
|
|
||
| def test_stacking_regressor_multioutput(): | ||
| """Check that a stacking regressor with multioutput works""" |
There was a problem hiding this comment.
Suggested change
| """Check that a stacking regressor with multioutput works""" | |
| """Check that a stacking regressor works with multioutput""" |
|
@OmarManzoor
Actually, I don't have a good idea yet on how to handle an estimator that does not support multiple outputs when it is used in a multi-output problem. Do you know such an API? Python 3.10.13 (main, Feb 22 2024, 10:50:12) [GCC 11.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from sklearn.ensemble import StackingRegressor
>>> from sklearn.svm import SVR
>>> from sklearn.linear_model import LinearRegression
>>> lr = LinearRegression()
>>> svr = SVR()
>>> model = StackingRegressor(estimators=[('lr', lr), ('svr', svr)])
>>> import numpy as np
>>> X = np.random.randn(10, 2)
>>> Y = X ** 2
>>> model.fit(X, Y)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/root/workspace/scikit-learn/sklearn/ensemble/_stacking.py", line 973, in fit
return super().fit(X, y, sample_weight)
File "/root/workspace/scikit-learn/sklearn/base.py", line 1473, in wrapper
return fit_method(estimator, *args, **kwargs)
File "/root/workspace/scikit-learn/sklearn/ensemble/_stacking.py", line 224, in fit
self.estimators_ = Parallel(n_jobs=self.n_jobs)(
File "/root/workspace/scikit-learn/sklearn/utils/parallel.py", line 67, in __call__
return super().__call__(iterable_with_config)
File "/root/workspace/scikit-learn/sklearn-env/lib/python3.10/site-packages/joblib/parallel.py", line 1918, in __call__
return output if self.return_generator else list(output)
File "/root/workspace/scikit-learn/sklearn-env/lib/python3.10/site-packages/joblib/parallel.py", line 1847, in _get_sequential_output
res = func(*args, **kwargs)
File "/root/workspace/scikit-learn/sklearn/utils/parallel.py", line 129, in __call__
return self.function(*args, **kwargs)
File "/root/workspace/scikit-learn/sklearn/ensemble/_base.py", line 40, in _fit_single_estimator
estimator.fit(X, y, **fit_params)
File "/root/workspace/scikit-learn/sklearn/base.py", line 1473, in wrapper
return fit_method(estimator, *args, **kwargs)
File "/root/workspace/scikit-learn/sklearn/svm/_base.py", line 190, in fit
X, y = self._validate_data(
File "/root/workspace/scikit-learn/sklearn/base.py", line 650, in _validate_data
X, y = check_X_y(X, y, **check_params)
File "/root/workspace/scikit-learn/sklearn/utils/validation.py", line 1282, in check_X_y
y = _check_y(y, multi_output=multi_output, y_numeric=y_numeric, estimator=estimator)
File "/root/workspace/scikit-learn/sklearn/utils/validation.py", line 1303, in _check_y
y = column_or_1d(y, warn=True)
File "/root/workspace/scikit-learn/sklearn/utils/validation.py", line 1370, in column_or_1d
raise ValueError(
ValueError: y should be a 1d array, got an array of shape (10, 2) instead. |
|
Note that >>> import numpy as np
>>> from sklearn.linear_model import LogisticRegression
>>> from sklearn.ensemble import StackingClassifier
>>> from sklearn.multioutput import MultiOutputClassifier
>>> X = np.random.randn(10, 2)
>>> Y = X > 0 # multilabel classification
>>> model = StackingClassifier(estimators=[('lr', MultiOutputClassifier(LogisticRegression(C=1e3))), ('lr2', MultiOutputClassifier(LogisticRegression(C=1e3)))], final_estimator=MultiOutputClassifier(LogisticRegression(C=1e3)))
>>> model.fit(X, Y)
StackingClassifier(estimators=[('lr',
MultiOutputClassifier(estimator=LogisticRegression(C=1000.0))),
('lr2',
MultiOutputClassifier(estimator=LogisticRegression(C=1000.0)))],
final_estimator=MultiOutputClassifier(estimator=LogisticRegression(C=1000.0)))
>>> model.predict(X)[:3]
array([[ True, True],
[False, True],
[False, False]])
>>> model.predict_proba(X)[:3]
array([[1.55247883e-03, 7.05602027e-04],
[9.99983741e-01, 2.31839536e-03],
[9.99873471e-01, 9.99473360e-01]])
>>> Z = np.random.choice(range(3), size=X.shape) # multiclass-multioutput classification
>>> model.fit(X, Z)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/root/workspace/scikit-learn/sklearn/ensemble/_stacking.py", line 669, in fit
self._label_encoder = LabelEncoder().fit(y)
File "/root/workspace/scikit-learn/sklearn/preprocessing/_label.py", line 97, in fit
y = column_or_1d(y, warn=True)
File "/root/workspace/scikit-learn/sklearn/utils/validation.py", line 1370, in column_or_1d
raise ValueError(
ValueError: y should be a 1d array, got an array of shape (10, 2) instead.Ref. https://scikit-learn.org/stable/modules/multiclass.html |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Reference Issues/PRs
Related to #25597
Similar to #8547
Similar to #19223
What does this implement/fix? Explain your changes.
Any other comments?
I am concerned the followings: