patch: scikit-learn 1.6 compatibility (#726)

* WIP: low hanging fix

* add sklearn-compat dependency

* preprocessing module

* decomposition module

* mixture and feature_selection modules

* meta module

* top level modules

* WIP: do not use validate_data

* check_X_y with changed check_array

* use validate_data

Author: FBruzzesi

sklego/_sklearn_compat.py

@@ -5,7 +5,9 @@
 import numpy as np
 import pandas as pd
 from sklearn.base import BaseEstimator, TransformerMixin
-from sklearn.utils.validation import check_array, check_is_fitted, check_X_y
+from sklearn.utils.validation import check_is_fitted
+
+from sklego._sklearn_compat import validate_data
 
 
 class TrainOnlyTransformerMixin(TransformerMixin, BaseEstimator):
@@ -79,11 +81,11 @@ def fit(self, X, y=None):
             The fitted transformer.
         """
         if y is None:
-            check_array(X, estimator=self)
+            validate_data(self, X=X, reset=True)
         else:
-            check_X_y(X, y, estimator=self, multi_output=True)
+            validate_data(self, X=X, y=y, multi_output=True, reset=True)
+
         self.X_hash_ = self._hash(X)
-        self.n_features_in_ = X.shape[1]
         return self
 
     @staticmethod
@@ -145,10 +147,7 @@ def transform(self, X, y=None):
             If the input dimension does not match the training dimension.
         """
         check_is_fitted(self, ["X_hash_", "n_features_in_"])
-        check_array(X, estimator=self)
-
-        if X.shape[1] != self.n_features_in_:
-            raise ValueError(f"Unexpected input dimension {X.shape[1]}, expected {self.n_features_in_}")
+        validate_data(self, X=X, reset=False)
 
         if self._hash(X) == self.X_hash_:
             return self.transform_train(X)

sklego/common.py

@@ -1,7 +1,9 @@
 import numpy as np
 from sklearn.base import BaseEstimator, OutlierMixin
 from sklearn.decomposition import PCA
-from sklearn.utils.validation import FLOAT_DTYPES, check_array, check_is_fitted
+from sklearn.utils.validation import FLOAT_DTYPES, check_is_fitted
+
+from sklego._sklearn_compat import validate_data
 
 
 class PCAOutlierDetection(OutlierMixin, BaseEstimator):
@@ -94,7 +96,7 @@ def fit(self, X, y=None):
         ValueError
             If `threshold` is `None`.
         """
-        X = check_array(X, estimator=self, dtype=FLOAT_DTYPES)
+        X = validate_data(self, X=X, dtype=FLOAT_DTYPES, reset=True)
         if not self.threshold:
             raise ValueError("The `threshold` value cannot be `None`.")
 
@@ -108,8 +110,6 @@ def fit(self, X, y=None):
         )
         self.pca_.fit(X, y)
         self.offset_ = -self.threshold
-
-        self.n_features_in_ = X.shape[1]
         return self
 
     def difference(self, X):
@@ -126,6 +126,8 @@ def difference(self, X):
             The calculated difference.
         """
         check_is_fitted(self, ["pca_", "offset_"])
+        X = validate_data(self, X=X, dtype=FLOAT_DTYPES, reset=False)
+
         reduced = self.pca_.transform(X)
         diff = np.sum(np.abs(self.pca_.inverse_transform(reduced) - X), axis=1)
         if self.variant == "relative":
@@ -157,8 +159,8 @@ def predict(self, X):
         array-like of shape (n_samples,)
             The predicted data. 1 for inliers, -1 for outliers.
         """
-        X = check_array(X, estimator=self, dtype=FLOAT_DTYPES)
         check_is_fitted(self, ["pca_", "offset_"])
+        X = validate_data(self, X=X, dtype=FLOAT_DTYPES, reset=False)
         result = np.ones(X.shape[0])
         result[self.difference(X) > self.threshold] = -1
         return result.astype(int)

sklego/decomposition/pca_reconstruction.py

@@ -8,7 +8,9 @@
 
 import numpy as np
 from sklearn.base import BaseEstimator, OutlierMixin
-from sklearn.utils.validation import FLOAT_DTYPES, check_array, check_is_fitted
+from sklearn.utils.validation import FLOAT_DTYPES, check_is_fitted
+
+from sklego._sklearn_compat import validate_data
 
 
 class UMAPOutlierDetection(OutlierMixin, BaseEstimator):
@@ -100,9 +102,10 @@ def fit(self, X, y=None):
             - If `n_components` is less than 2.
             - If `threshold` is `None`.
         """
-        X = check_array(X, estimator=self, dtype=FLOAT_DTYPES)
         if y is not None:
-            y = check_array(y, estimator=self, ensure_2d=False)
+            X, y = validate_data(self, X=X, y=y, dtype=FLOAT_DTYPES, reset=True)
+        else:
+            X = validate_data(self, X=X, dtype=FLOAT_DTYPES, reset=True)
 
         if not self.threshold:
             raise ValueError("The `threshold` value cannot be `None`.")
@@ -116,7 +119,6 @@ def fit(self, X, y=None):
         )
         self.umap_.fit(X, y)
         self.offset_ = -self.threshold
-        self.n_features_in_ = X.shape[1]
         return self
 
     def difference(self, X):
@@ -133,6 +135,8 @@ def difference(self, X):
             The calculated difference.
         """
         check_is_fitted(self, ["umap_", "offset_"])
+        X = validate_data(self, X=X, dtype=FLOAT_DTYPES, reset=False)
+
         reduced = self.umap_.transform(X)
         diff = np.sum(np.abs(self.umap_.inverse_transform(reduced) - X), axis=1)
         if self.variant == "relative":
@@ -155,8 +159,8 @@ def predict(self, X):
         array-like of shape (n_samples,)
             The predicted data. 1 for inliers, -1 for outliers.
         """
-        X = check_array(X, estimator=self, dtype=FLOAT_DTYPES)
         check_is_fitted(self, ["umap_", "offset_"])
+        X = validate_data(self, X=X, dtype=FLOAT_DTYPES, reset=False)
         result = np.ones(X.shape[0])
         result[self.difference(X) > self.threshold] = -1
         return result.astype(int)
@@ -172,3 +176,8 @@ def score_samples(self, X):
 
     def _more_tags(self):
         return {"non_deterministic": True}
+
+    def __sklearn_tags__(self):
+        tags = super().__sklearn_tags__()
+        tags.non_deterministic = True
+        return tags

sklego/decomposition/umap_reconstruction.py

@@ -2,13 +2,9 @@
 
 import numpy as np
 from sklearn.base import BaseEstimator, RegressorMixin
-from sklearn.utils import check_X_y
-from sklearn.utils.validation import (
-    FLOAT_DTYPES,
-    check_array,
-    check_is_fitted,
-    check_random_state,
-)
+from sklearn.utils.validation import FLOAT_DTYPES, check_is_fitted, check_random_state
+
+from sklego._sklearn_compat import validate_data
 
 
 class RandomRegressor(RegressorMixin, BaseEstimator):
@@ -72,8 +68,7 @@ def fit(self, X: np.array, y: np.array) -> "RandomRegressor":
         """
         if self.strategy not in self._ALLOWED_STRATEGIES:
             raise ValueError(f"strategy {self.strategy} is not in {self._ALLOWED_STRATEGIES}")
-        X, y = check_X_y(X, y, estimator=self, dtype=FLOAT_DTYPES)
-        self.n_features_in_ = X.shape[1]
+        X, y = validate_data(self, X=X, y=y, dtype=FLOAT_DTYPES, reset=True)
 
         self.min_ = np.min(y)
         self.max_ = np.max(y)
@@ -99,9 +94,7 @@ def predict(self, X):
         rs = check_random_state(self.random_state)
         check_is_fitted(self, ["n_features_in_", "min_", "max_", "mu_", "sigma_"])
 
-        X = check_array(X, estimator=self, dtype=FLOAT_DTYPES)
-        if X.shape[1] != self.n_features_in_:
-            raise ValueError(f"Unexpected input dimension {X.shape[1]}, expected {self.dim_}")
+        X = validate_data(self, X=X, dtype=FLOAT_DTYPES, reset=False)
 
         if self.strategy == "normal":
             return rs.normal(self.mu_, self.sigma_, X.shape[0])
@@ -127,3 +120,9 @@ def allowed_strategies(self):
 
     def _more_tags(self):
         return {"poor_score": True, "non_deterministic": True}
+
+    def __sklearn_tags__(self):
+        tags = super().__sklearn_tags__()
+        tags.non_deterministic = True
+        tags.regressor_tags.poor_score = True
+        return tags

sklego/dummy.py

@@ -4,7 +4,9 @@
 from sklearn.base import BaseEstimator
 from sklearn.feature_selection import f_classif, f_regression
 from sklearn.feature_selection._base import SelectorMixin
-from sklearn.utils.validation import check_is_fitted, check_X_y
+from sklearn.utils.validation import check_is_fitted
+
+from sklego._sklearn_compat import validate_data
 
 
 def _redundancy_pearson(X, selected, left):
@@ -201,13 +203,12 @@ def fit(self, X, y):
 
                 k parameter is not integer type or is < n_features_in (X.shape[1]) or < 1
         """
-        X, y = check_X_y(X, y, dtype="numeric", y_numeric=True)
+        X, y = validate_data(self, X=X, y=y, dtype="numeric", y_numeric=True, reset=True)
         self._y_dtype = y.dtype
 
         relevance = self._get_relevance
         redundancy = self._get_redundancy
 
-        self.n_features_in_ = X.shape[1]
         left_features = list(range(self.n_features_in_))
         selected_features = []
         selected_scores = []

sklego/feature_selection/mrmr.py

@@ -21,11 +21,12 @@
 from sklearn.utils.validation import (
     FLOAT_DTYPES,
     _check_sample_weight,
-    check_array,
     check_is_fitted,
     column_or_1d,
 )
 
+from sklego._sklearn_compat import check_array, validate_data
+
 
 class LowessRegression(RegressorMixin, BaseEstimator):
     """`LowessRegression` estimator: LOWESS (Locally Weighted Scatterplot Smoothing) is a type of
@@ -96,7 +97,7 @@ def fit(self, X, y):
             - If `span` is not between 0 and 1.
             - If `sigma` is negative.
         """
-        X, y = check_X_y(X, y, estimator=self, dtype=FLOAT_DTYPES)
+        X, y = validate_data(self, X=X, y=y, dtype=FLOAT_DTYPES, reset=True)
         if self.span is not None:
             if not 0 <= self.span <= 1:
                 raise ValueError(f"Param `span` must be 0 <= span <= 1, got: {self.span}")
@@ -138,8 +139,8 @@ def predict(self, X):
         array-like of shape (n_samples,)
             The predicted values.
         """
-        X = check_array(X, estimator=self, dtype=FLOAT_DTYPES)
         check_is_fitted(self, ["X_", "y_"])
+        X = validate_data(self, X=X, dtype=FLOAT_DTYPES, reset=False)
 
         try:
             results = np.stack([np.average(self.y_, weights=self._calc_wts(x_i=x_i)) for x_i in X])
@@ -233,7 +234,7 @@ def fit(self, X, y):
         self : ProbWeightRegression
             The fitted estimator.
         """
-        X, y = check_X_y(X, y, estimator=self, dtype=FLOAT_DTYPES)
+        X, y = validate_data(self, X=X, y=y, dtype=FLOAT_DTYPES, reset=True)
 
         # Construct the problem.
         betas = cp.Variable(X.shape[1])
@@ -263,8 +264,8 @@ def predict(self, X):
         array-like of shape (n_samples,)
             The predicted data.
         """
-        X = check_array(X, estimator=self, dtype=FLOAT_DTYPES)
         check_is_fitted(self, ["coef_"])
+        X = validate_data(self, X=X, dtype=FLOAT_DTYPES, reset=False)
         return np.dot(X, self.coef_)
 
     @property
@@ -345,8 +346,6 @@ class DeadZoneRegressor(RegressorMixin, BaseEstimator):
 
     print(y_pred)
     ```
-
-
     """
 
     _ALLOWED_EFFECTS = ("linear", "quadratic", "constant")
@@ -381,7 +380,8 @@ def fit(self, X, y):
         ValueError
             If `effect` is not one of "linear", "quadratic" or "constant".
         """
-        X, y = check_X_y(X, y, estimator=self, dtype=FLOAT_DTYPES)
+        X, y = validate_data(self, X=X, y=y, dtype=FLOAT_DTYPES, reset=True)
+
         if self.effect not in self._ALLOWED_EFFECTS:
             raise ValueError(f"effect {self.effect} must be in {self._ALLOWED_EFFECTS}")
 
@@ -458,8 +458,9 @@ def predict(self, X):
         array-like of shape (n_samples,)
             The predicted data.
         """
-        X = check_array(X, estimator=self, dtype=FLOAT_DTYPES)
         check_is_fitted(self, ["coef_"])
+        X = validate_data(self, X=X, dtype=FLOAT_DTYPES, reset=False)
+
         return np.dot(X, self.coef_)
 
     @property
@@ -970,8 +971,6 @@ def __init__(
         self.fit_intercept = fit_intercept
         self.copy_X = copy_X
         self.positive = positive
-        if method not in ("SLSQP", "TNC", "L-BFGS-B"):
-            raise ValueError(f'method should be one of "SLSQP", "TNC", "L-BFGS-B", ' f"got {method} instead")
         self.method = method
 
     @abstractmethod
@@ -1021,6 +1020,10 @@ def fit(self, X, y, sample_weight=None):
         self : BaseScipyMinimizeRegressor
             Fitted linear model.
         """
+        if self.method not in {"SLSQP", "TNC", "L-BFGS-B"}:
+            msg = f"method should be one of 'SLSQP', 'TNC', 'L-BFGS-B', got {self.method} instead"
+            raise ValueError(msg)
+
         X_, grad_loss, loss = self._prepare_inputs(X, sample_weight, y)
 
         d = X_.shape[1] - self.n_features_in_  # This is either zero or one.
@@ -1051,7 +1054,8 @@ def _prepare_inputs(self, X, sample_weight, y):
         This method is called by `fit` to prepare the inputs for the optimization problem. It adds an intercept column
         to `X` if `fit_intercept=True`, and returns the loss function and its gradient.
         """
-        X, y = check_X_y(X, y, y_numeric=True)
+        X, y = validate_data(self, X=X, y=y, y_numeric=True, reset=True)
+
         sample_weight = _check_sample_weight(sample_weight, X)
         self.n_features_in_ = X.shape[1]
 
@@ -1081,7 +1085,7 @@ def predict(self, X):
             The predicted data.
         """
         check_is_fitted(self)
-        X = check_array(X)
+        X = validate_data(self, X=X, reset=False)
 
         return X @ self.coef_ + self.intercept_
 

sklego/linear_model.py

@@ -5,9 +5,10 @@
 import narwhals.stable.v1 as nw
 import pandas as pd
 from scipy.sparse import issparse
-from sklearn.utils import check_array
 from sklearn.utils.validation import _ensure_no_complex_data
 
+from sklego._sklearn_compat import check_array
+
 
 def parse_X_y(X, y, groups, check_X=True, **kwargs) -> nw.DataFrame:
     """Converts X, y to narwhals dataframe.

sklego/meta/_grouped_utils.py

@@ -2,8 +2,9 @@
 from sklearn.base import BaseEstimator, ClassifierMixin, MetaEstimatorMixin
 from sklearn.metrics import confusion_matrix
 from sklearn.utils.multiclass import unique_labels
-from sklearn.utils.validation import FLOAT_DTYPES, check_array, check_is_fitted, check_X_y
+from sklearn.utils.validation import FLOAT_DTYPES, check_is_fitted
 
+from sklego._sklearn_compat import validate_data
 from sklego.base import ProbabilisticClassifier
 
 
@@ -63,7 +64,8 @@ def fit(self, X, y):
             If the underlying estimator does not have a `predict_proba` method.
         """
 
-        X, y = check_X_y(X, y, estimator=self.estimator, dtype=FLOAT_DTYPES)
+        X, y = validate_data(self, X=X, y=y, dtype=FLOAT_DTYPES, reset=True)
+
         if not isinstance(self.estimator, ProbabilisticClassifier):
             raise ValueError(
                 "The ConfusionBalancer meta model only works on classification models with .predict_proba."
@@ -72,7 +74,6 @@ def fit(self, X, y):
         self.classes_ = unique_labels(y)
         cfm = confusion_matrix(y, self.estimator_.predict(X)).T + self.cfm_smooth
         self.cfm_ = cfm / cfm.sum(axis=1).reshape(-1, 1)
-        self.n_features_in_ = X.shape[1]
         return self
 
     def predict_proba(self, X):
@@ -90,7 +91,7 @@ def predict_proba(self, X):
             The predicted values.
         """
         check_is_fitted(self, ["cfm_", "classes_", "estimator_"])
-        X = check_array(X, dtype=FLOAT_DTYPES)
+        X = validate_data(self, X=X, dtype=FLOAT_DTYPES, reset=False)
         preds = self.estimator_.predict_proba(X)
         return (1 - self.alpha) * preds + self.alpha * preds @ self.cfm_
 
@@ -108,5 +109,5 @@ def predict(self, X):
             The predicted values.
         """
         check_is_fitted(self, ["cfm_", "classes_", "estimator_"])
-        X = check_array(X, dtype=FLOAT_DTYPES)
+        X = validate_data(self, X=X, dtype=FLOAT_DTYPES, reset=False)
         return self.classes_[self.predict_proba(X).argmax(axis=1)]