BUG: Fixes for old sklearn (#13308)

Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>

Author: larsoner

environment.yml

@@ -42,14 +42,14 @@ dependencies:
   - pyarrow
   - pybv
   - pymatreader
-  - PySide6 !=6.8.0,!=6.8.0.1
+  - PySide6 !=6.9.1
   - python-neo
   - python-picard
   - pyvista >=0.32,!=0.35.2,!=0.38.0,!=0.38.1,!=0.38.2,!=0.38.3,!=0.38.4,!=0.38.5,!=0.38.6,!=0.42.0
   - pyvistaqt >=0.4
   - qdarkstyle !=3.2.2
   - qtpy
-  - scikit-learn
+  - scikit-learn >=1.3.0
   - scipy >=1.11
   - sip
   - snirf
@@ -60,5 +60,5 @@ dependencies:
   - trame
   - trame-vtk
   - trame-vuetify
-  - vtk =9.3.1=qt_*
+  - vtk >=9.2
   - xlrd

mne/decoding/_fixes.py

@@ -0,0 +1,132 @@
+# Authors: The MNE-Python contributors.
+# License: BSD-3-Clause
+# Copyright the MNE-Python contributors.
+
+try:
+    from sklearn.utils.validation import validate_data
+except ImportError:
+    from sklearn.utils.validation import check_array, check_X_y
+
+    # Use a limited version pulled from sklearn 1.7
+    def validate_data(
+        _estimator,
+        /,
+        X="no_validation",
+        y="no_validation",
+        reset=True,
+        validate_separately=False,
+        skip_check_array=False,
+        **check_params,
+    ):
+        """Validate input data and set or check feature names and counts of the input.
+
+        This helper function should be used in an estimator that requires input
+        validation. This mutates the estimator and sets the `n_features_in_` and
+        `feature_names_in_` attributes if `reset=True`.
+
+        .. versionadded:: 1.6
+
+        Parameters
+        ----------
+        _estimator : estimator instance
+            The estimator to validate the input for.
+
+        X : {array-like, sparse matrix, dataframe} of shape \
+                (n_samples, n_features), default='no validation'
+            The input samples.
+            If `'no_validation'`, no validation is performed on `X`. This is
+            useful for meta-estimator which can delegate input validation to
+            their underlying estimator(s). In that case `y` must be passed and
+            the only accepted `check_params` are `multi_output` and
+            `y_numeric`.
+
+        y : array-like of shape (n_samples,), default='no_validation'
+            The targets.
+
+            - If `None`, :func:`~sklearn.utils.check_array` is called on `X`. If
+            the estimator's `requires_y` tag is True, then an error will be raised.
+            - If `'no_validation'`, :func:`~sklearn.utils.check_array` is called
+            on `X` and the estimator's `requires_y` tag is ignored. This is a default
+            placeholder and is never meant to be explicitly set. In that case `X` must
+            be passed.
+            - Otherwise, only `y` with `_check_y` or both `X` and `y` are checked with
+            either :func:`~sklearn.utils.check_array` or
+            :func:`~sklearn.utils.check_X_y` depending on `validate_separately`.
+
+        reset : bool, default=True
+            Whether to reset the `n_features_in_` attribute.
+            If False, the input will be checked for consistency with data
+            provided when reset was last True.
+
+            .. note::
+
+            It is recommended to call `reset=True` in `fit` and in the first
+            call to `partial_fit`. All other methods that validate `X`
+            should set `reset=False`.
+
+        validate_separately : False or tuple of dicts, default=False
+            Only used if `y` is not `None`.
+            If `False`, call :func:`~sklearn.utils.check_X_y`. Else, it must be a tuple
+            of kwargs to be used for calling :func:`~sklearn.utils.check_array` on `X`
+            and `y` respectively.
+
+            `estimator=self` is automatically added to these dicts to generate
+            more informative error message in case of invalid input data.
+
+        skip_check_array : bool, default=False
+            If `True`, `X` and `y` are unchanged and only `feature_names_in_` and
+            `n_features_in_` are checked. Otherwise, :func:`~sklearn.utils.check_array`
+            is called on `X` and `y`.
+
+        **check_params : kwargs
+            Parameters passed to :func:`~sklearn.utils.check_array` or
+            :func:`~sklearn.utils.check_X_y`. Ignored if validate_separately
+            is not False.
+
+            `estimator=self` is automatically added to these params to generate
+            more informative error message in case of invalid input data.
+
+        Returns
+        -------
+        out : {ndarray, sparse matrix} or tuple of these
+            The validated input. A tuple is returned if both `X` and `y` are
+            validated.
+        """
+        no_val_X = isinstance(X, str) and X == "no_validation"
+        no_val_y = y is None or (isinstance(y, str) and y == "no_validation")
+
+        if no_val_X and no_val_y:
+            raise ValueError("Validation should be done on X, y or both.")
+
+        default_check_params = {"estimator": _estimator}
+        check_params = {**default_check_params, **check_params}
+
+        if skip_check_array:
+            if not no_val_X and no_val_y:
+                out = X
+            elif no_val_X and not no_val_y:
+                out = y
+            else:
+                out = X, y
+        elif not no_val_X and no_val_y:
+            out = check_array(X, input_name="X", **check_params)
+        elif no_val_X and not no_val_y:
+            out = check_array(y, input_name="y", **check_params)
+        else:
+            if validate_separately:
+                # We need this because some estimators validate X and y
+                # separately, and in general, separately calling check_array()
+                # on X and y isn't equivalent to just calling check_X_y()
+                # :(
+                check_X_params, check_y_params = validate_separately
+                if "estimator" not in check_X_params:
+                    check_X_params = {**default_check_params, **check_X_params}
+                X = check_array(X, input_name="X", **check_X_params)
+                if "estimator" not in check_y_params:
+                    check_y_params = {**default_check_params, **check_y_params}
+                y = check_array(y, input_name="y", **check_y_params)
+            else:
+                X, y = check_X_y(X, y, **check_params)
+            out = X, y
+
+        return out

mne/decoding/tests/test_base.py

@@ -49,6 +49,7 @@
     get_coef,
 )
 from mne.decoding.search_light import SlidingEstimator
+from mne.utils import check_version
 
 
 def _make_data(n_samples=1000, n_features=5, n_targets=3):
@@ -97,7 +98,8 @@ def test_get_coef():
     """Test getting linear coefficients (filters/patterns) from estimators."""
     lm_classification = LinearModel()
     assert hasattr(lm_classification, "__sklearn_tags__")
-    print(lm_classification.__sklearn_tags__())
+    if check_version("sklearn", "1.4"):
+        print(lm_classification.__sklearn_tags__())
     assert is_classifier(lm_classification.model)
     assert is_classifier(lm_classification)
     assert not is_regressor(lm_classification.model)

mne/decoding/tests/test_csp.py

@@ -495,4 +495,5 @@ def test_csp_component_ordering():
 @parametrize_with_checks([CSP(), SPoC()])
 def test_sklearn_compliance(estimator, check):
     """Test compliance with sklearn."""
+    pytest.importorskip("sklearn", minversion="1.4")  # TODO VERSION remove on 1.4+
     check(estimator)

mne/decoding/tests/test_ems.py

@@ -97,4 +97,5 @@ def test_ems():
 @parametrize_with_checks([EMS()])
 def test_sklearn_compliance(estimator, check):
     """Test compliance with sklearn."""
+    pytest.importorskip("sklearn", minversion="1.4")  # TODO VERSION remove on 1.4+
     check(estimator)

mne/decoding/tests/test_ssd.py

@@ -499,6 +499,7 @@ def test_non_full_rank_data():
 )
 def test_sklearn_compliance(estimator, check):
     """Test LinearModel compliance with sklearn."""
+    pytest.importorskip("sklearn", minversion="1.4")  # TODO VERSION remove on 1.4+
     ignores = (
         "check_methods_sample_order_invariance",
         # Shape stuff

mne/decoding/tests/test_time_frequency.py

@@ -57,4 +57,5 @@ def test_timefrequency_basic():
 @parametrize_with_checks([TimeFrequency([300, 400], 1000.0, n_cycles=0.25)])
 def test_sklearn_compliance(estimator, check):
     """Test LinearModel compliance with sklearn."""
+    pytest.importorskip("sklearn", minversion="1.4")  # TODO VERSION remove on 1.4+
     check(estimator)

mne/decoding/tests/test_transformer.py

@@ -339,6 +339,7 @@ def test_bad_triage():
 )
 def test_sklearn_compliance(estimator, check):
     """Test LinearModel compliance with sklearn."""
+    pytest.importorskip("sklearn", minversion="1.4")  # TODO VERSION remove on 1.4+
     ignores = []
     if estimator.__class__.__name__ == "FilterEstimator":
         ignores += [

mne/decoding/transformer.py

@@ -6,7 +6,7 @@
 from sklearn.base import BaseEstimator, TransformerMixin, check_array, clone
 from sklearn.preprocessing import RobustScaler, StandardScaler
 from sklearn.utils import check_X_y
-from sklearn.utils.validation import check_is_fitted, validate_data
+from sklearn.utils.validation import check_is_fitted
 
 from .._fiff.pick import (
     _pick_data_channels,
@@ -18,7 +18,8 @@
 from ..epochs import BaseEpochs
 from ..filter import filter_data
 from ..time_frequency import psd_array_multitaper
-from ..utils import _check_option, _validate_type, fill_doc
+from ..utils import _check_option, _validate_type, check_version, fill_doc
+from ._fixes import validate_data  # TODO VERSION remove with sklearn 1.4+
 
 
 class MNETransformerMixin(TransformerMixin):
@@ -41,7 +42,9 @@ def _check_data(
         # array to make everyone happy.
         if isinstance(epochs_data, BaseEpochs):
             epochs_data = epochs_data.get_data(copy=False)
-        kwargs = dict(dtype=np.float64, allow_nd=True, order="C", force_writeable=True)
+        kwargs = dict(dtype=np.float64, allow_nd=True, order="C")
+        if check_version("sklearn", "1.4"):  # TODO VERSION sklearn 1.4+
+            kwargs["force_writeable"] = True
         if hasattr(self, "n_features_in_") and check_n_features:
             if y is None:
                 epochs_data = validate_data(

mne/inverse_sparse/tests/test_mxne_inverse.py

@@ -589,8 +589,12 @@ def data_fun(times):
         forward, stc, info, noise_cov, nave=nave, use_cps=False, iir_filter=None
     )
     evoked = evoked.crop(tmin=0, tmax=10e-3)
-    stc_ = mixed_norm(evoked, forward, noise_cov, loose=0.9, n_mxne_iter=5, depth=0.9)
-    assert_array_equal(stc_.vertices, stc.vertices)
+    stc_ = mixed_norm(
+        evoked, forward, noise_cov, loose=0.9, n_mxne_iter=5, depth=0.9, random_state=0
+    )
+    assert len(stc_.vertices) == len(stc.vertices) == 2
+    for si in range(len(stc_.vertices)):
+        assert_array_equal(stc_.vertices[si], stc.vertices[si], err_msg=f"{si=}")
 
 
 @pytest.mark.slowtest  # slow on Azure
@@ -609,7 +613,13 @@ def test_mxne_inverse_empty():
     cov = read_cov(fname_cov)
     with pytest.warns(RuntimeWarning, match="too big"):
         stc, residual = mixed_norm(
-            evoked, forward, cov, n_mxne_iter=3, alpha=99, return_residual=True
+            evoked,
+            forward,
+            cov,
+            n_mxne_iter=3,
+            alpha=99,
+            return_residual=True,
+            random_state=0,
         )
         assert stc.data.size == 0
         assert stc.vertices[0].size == 0