Merge pull request #632 from GAA-UAM/feature/numpy2

Allow to use NumPy 2.
GAA-UAM · Aug 31, 2024 · d0bc82f · d0bc82f
2 parents 2f1fb20 + 58796e1
commit d0bc82f
Show file tree

Hide file tree

Showing 23 changed files with 84 additions and 83 deletions.
diff --git a/pyproject.toml b/pyproject.toml
@@ -37,7 +37,7 @@ dependencies = [
 	"lazy_loader",
 	"matplotlib",
 	"multimethod>=1.5, !=1.11, != 1.11.1",
-	"numpy>=1.16, <2",
+	"numpy>=1.16",
 	"pandas>=1.0",
 	"rdata",
 	"scikit-datasets[cran]>=0.2.2",
@@ -61,6 +61,7 @@ docs = [
   "sphinxcontrib-bibtex",
 ]
 test = [
+  "numpy>=2", # Changes in array representation.
   "pytest",
   "pytest-env",
   "pytest-subtests",

diff --git a/skfda/_utils/_utils.py b/skfda/_utils/_utils.py
@@ -201,22 +201,22 @@ def _cartesian_product(  # noqa: WPS234
         >>> from skfda._utils import _cartesian_product
         >>> axes = [[0,1],[2,3]]
         >>> _cartesian_product(axes)
-        array([[0, 2],
-               [0, 3],
-               [1, 2],
-               [1, 3]])
+        array([[ 0, 2],
+               [ 0, 3],
+               [ 1, 2],
+               [ 1, 3]])
 
         >>> axes = [[0,1],[2,3],[4]]
         >>> _cartesian_product(axes)
-        array([[0, 2, 4],
-               [0, 3, 4],
-               [1, 2, 4],
-               [1, 3, 4]])
+        array([[ 0, 2, 4],
+               [ 0, 3, 4],
+               [ 1, 2, 4],
+               [ 1, 3, 4]])
 
         >>> axes = [[0,1]]
         >>> _cartesian_product(axes)
-        array([[0],
-               [1]])
+        array([[ 0],
+               [ 1]])
     """
     cartesian = np.stack(np.meshgrid(*axes, indexing='ij'), -1)
 

diff --git a/skfda/exploratory/outliers/_directional_outlyingness.py b/skfda/exploratory/outliers/_directional_outlyingness.py
@@ -316,7 +316,7 @@ class MSPlotOutlierDetector(  # noqa: WPS230
         >>> fd = skfda.FDataGrid(data_matrix, grid_points)
         >>> out_detector = MSPlotOutlierDetector()
         >>> out_detector.fit_predict(fd)
-        array([1, 1, 1, 1])
+        array([ 1, 1, 1, 1])
 
     References:
         Dai, Wenlin, and Genton, Marc G. "Multivariate functional data

diff --git a/skfda/misc/_math.py b/skfda/misc/_math.py
@@ -257,14 +257,14 @@ def inner_product(
         >>> array1 = np.array([1, 2, 3])
         >>> array2 = np.array([4, 5, 6])
         >>> inner_product(array1, array2)
-        32
+        np.int64(32)
 
         If the arrays contain more than one sample
 
         >>> array1 = np.array([[1, 2, 3], [2, 3, 4]])
         >>> array2 = np.array([[4, 5, 6], [1, 1, 1]])
         >>> inner_product(array1, array2)
-        array([32, 9])
+        array([ 32, 9])
 
         The inner product of the :math:`f(x) = x` and the constant
         :math:`y=1` defined over the interval :math:`[0,1]` is the area of
@@ -592,7 +592,7 @@ def cosine_similarity(
         >>> array1 = np.array([1, 2, 3])
         >>> array2 = np.array([4, 5, 6])
         >>> cosine_similarity(array1, array2)
-        0.9746318461970762
+        np.float64(0.9746318461970762)
 
         If the arrays contain more than one sample
 

diff --git a/skfda/misc/covariances.py b/skfda/misc/covariances.py
@@ -29,7 +29,7 @@ def _squared_norms(x: NDArrayFloat, y: NDArrayFloat) -> NDArrayFloat:
 
 def _transform_to_2d(t: ArrayLike) -> NDArrayFloat:
     """Transform 1d arrays in column vectors."""
-    t = np.asfarray(t)
+    t = np.asarray(t, dtype=np.float64)
 
     dim = t.ndim
     assert dim <= 2

diff --git a/skfda/misc/operators/_linear_differential_operator.py b/skfda/misc/operators/_linear_differential_operator.py
@@ -616,14 +616,14 @@ def _optimized_operator_evaluation_in_grid(
         [ 0.,  0.,  0.,  0.,  1., -2.,  1.,  4.],
         [ 0.,  0.,  0.,  0.,  0.,  1., -2., -5.],
         [ 0.,  0.,  0.,  0.,  0.,  0.,  1.,  2.]]),
-        array([[0, 1],
-                [0, 2],
-                [0, 3],
-                [0, 4],
-                [3, 7],
-                [4, 7],
-                [5, 7],
-                [6, 7]]))
+        array([[ 0, 1],
+               [ 0, 2],
+               [ 0, 3],
+               [ 0, 4],
+               [ 3, 7],
+               [ 4, 7],
+               [ 5, 7],
+               [ 6, 7]]))
 
     Explanation:
         Each row of the first array contains the values of the linear operator

diff --git a/skfda/ml/classification/_centroid_classifiers.py b/skfda/ml/classification/_centroid_classifiers.py
@@ -63,7 +63,7 @@ class and return a :class:`FData` object with only one sample
         We can predict the class of new samples
 
         >>> neigh.predict(fd[::2]) # Predict labels for even samples
-        array([0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1])
+        array([ 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1])
 
     See also:
         :class:`~skfda.ml.classification.DTMClassifier`
@@ -164,7 +164,7 @@ class DTMClassifier(NearestCentroid[Input, Target]):
         We can predict the class of new samples
 
         >>> clf.predict(X_test) # Predict labels for test samples
-        array([1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1,
+        array([ 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1,
                 1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 1, 1])
 
         Finally, we calculate the mean accuracy for the test data

diff --git a/skfda/ml/classification/_depth_classifiers.py b/skfda/ml/classification/_depth_classifiers.py
@@ -106,8 +106,8 @@ class DDClassifier(
         We can predict the class of new samples
 
         >>> clf.predict(X_test)
-        array([1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1,
-               1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 1, 1])
+        array([ 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1,
+                1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 1, 1])
 
         Finally, we calculate the mean accuracy for the test data
 
@@ -278,8 +278,8 @@ class DDGClassifier(
         We can predict the class of new samples
 
         >>> clf.predict(X_test)
-        array([1, 1, 1, 0, 1, 1, 1, 1, 0, 0, 0, 1,
-               1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 1, 1])
+        array([ 1, 1, 1, 0, 1, 1, 1, 1, 0, 0, 0, 1,
+                1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 1, 1])
 
         Finally, we calculate the mean accuracy for the test data
 
@@ -299,8 +299,8 @@ class DDGClassifier(
         >>> clf.fit(X_train, y_train)
         DDGClassifier(...)
         >>> clf.predict(X_test)
-        array([1, 1, 1, 0, 1, 1, 1, 1, 0, 0, 0, 1,
-               1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 1, 1])
+        array([ 1, 1, 1, 0, 1, 1, 1, 1, 0, 0, 0, 1,
+                1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 1, 1])
         >>> clf.score(X_test, y_test)
         0.875
 
@@ -475,7 +475,7 @@ class _ArgMaxClassifier(
         We can predict the class of new samples
 
         >>> clf.predict(X) # Predict labels for test samples
-        array([1, 0, 0])
+        array([ 1, 0, 0])
     """
 
     def fit(self, X: NDArrayFloat, y: Target) -> _ArgMaxClassifier[Target]:
@@ -543,8 +543,8 @@ class MaximumDepthClassifier(DDGClassifier[Input, Target]):
         We can predict the class of new samples
 
         >>> clf.predict(X_test) # Predict labels for test samples
-        array([1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1,
-               1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 1, 1])
+        array([ 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1,
+                1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 1, 1])
 
         Finally, we calculate the mean accuracy for the test data
 

diff --git a/skfda/ml/classification/_neighbors_classifiers.py b/skfda/ml/classification/_neighbors_classifiers.py
@@ -89,7 +89,7 @@ class KNeighborsClassifier(
         We can predict the class of new samples
 
         >>> neigh.predict(fd[::2]) # Predict labels for even samples
-        array([0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1])
+        array([ 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1])
 
         And the estimated probabilities.
 
@@ -254,7 +254,7 @@ class RadiusNeighborsClassifier(
         We can predict the class of new samples.
 
         >>> neigh.predict(fd[::2]) # Predict labels for even samples
-        array([0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1])
+        array([ 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1])
 
     See also:
         :class:`~skfda.ml.classification.KNeighborsClassifier`

diff --git a/skfda/ml/classification/_qda.py b/skfda/ml/classification/_qda.py
@@ -95,9 +95,9 @@ class QuadraticDiscriminantAnalysis(
 
         We can predict the class of new samples.
 
-        >>> list(qda.predict(X_test))
-        [0, 1, 0, 0, 1, 1, 0, 1, 0, 1, 0, 0, 1, 1, 1, 0, 1, 1, 1, 1,
-         1, 0, 1, 0, 1, 0, 1, 1]
+        >>> qda.predict(X_test)
+        array([ 0, 1, 0, 0, 1, 1, 0, 1, 0, 1, 0, 0, 1, 1, 1, 0, 1, 1, 1, 1,
+         1, 0, 1, 0, 1, 0, 1, 1], dtype=int8)
 
         Finally, we calculate the mean accuracy for the test data.
 

diff --git a/skfda/ml/clustering/_hierarchical.py b/skfda/ml/clustering/_hierarchical.py
@@ -144,7 +144,7 @@ class AgglomerativeClustering(  # noqa: WPS230
         >>> clustering.fit(X)
         AgglomerativeClustering(...)
         >>> clustering.labels_.astype(np.int_)
-        array([0, 0, 1, 0, 0, 1])
+        array([ 0, 0, 1, 0, 0, 1])
     """
 
     LinkageCriterion = LinkageCriterion

diff --git a/skfda/ml/regression/_historical_linear_model.py b/skfda/ml/regression/_historical_linear_model.py
@@ -264,19 +264,19 @@ class HistoricalLinearRegression(
         >>> import scipy.integrate
 
         >>> random_state = np.random.RandomState(0)
-        >>> data_matrix = random_state.choice(10, size=(8, 6))
+        >>> data_matrix = random_state.choice(10, size=(8, 6)).astype(float)
         >>> data_matrix
-        array([[5, 0, 3, 3, 7, 9],
-               [3, 5, 2, 4, 7, 6],
-               [8, 8, 1, 6, 7, 7],
-               [8, 1, 5, 9, 8, 9],
-               [4, 3, 0, 3, 5, 0],
-               [2, 3, 8, 1, 3, 3],
-               [3, 7, 0, 1, 9, 9],
-               [0, 4, 7, 3, 2, 7]])
-        >>> intercept = random_state.choice(10, size=(1, 6))
+        array([[ 5., 0., 3., 3., 7., 9.],
+               [ 3., 5., 2., 4., 7., 6.],
+               [ 8., 8., 1., 6., 7., 7.],
+               [ 8., 1., 5., 9., 8., 9.],
+               [ 4., 3., 0., 3., 5., 0.],
+               [ 2., 3., 8., 1., 3., 3.],
+               [ 3., 7., 0., 1., 9., 9.],
+               [ 0., 4., 7., 3., 2., 7.]])
+        >>> intercept = random_state.choice(10, size=(1, 6)).astype(float)
         >>> intercept
-        array([[2, 0, 0, 4, 5, 5]])
+        array([[ 2., 0., 0., 4., 5., 5.]])
         >>> y_data = scipy.integrate.cumulative_trapezoid(
         ...              data_matrix,
         ...              initial=0,

diff --git a/skfda/preprocessing/dim_reduction/variable_selection/recursive_maxima_hunting.py b/skfda/preprocessing/dim_reduction/variable_selection/recursive_maxima_hunting.py
@@ -44,7 +44,7 @@
 
 
 def _transform_to_2d(t: ArrayLike) -> NDArrayFloat:
-    t = np.asfarray(t)
+    t = np.asarray(t, dtype=np.float64)
 
     dim = t.ndim
     assert dim <= 2
@@ -907,7 +907,7 @@ def fit(  # type: ignore[override] # noqa: D102
         """Recursive maxima hunting algorithm."""
         self.features_shape_ = X.data_matrix.shape[1:]
 
-        y = np.asfarray(y)
+        y = np.asarray(y, dtype=np.float64)
 
         self.correction_ = (
             UniformCorrection()

diff --git a/skfda/preprocessing/feature_construction/_function_transformers.py b/skfda/preprocessing/feature_construction/_function_transformers.py
@@ -226,7 +226,7 @@ class NumberCrossingsTransformer(
         FDataGrid created.
         >>> tf = NumberCrossingsTransformer(levels=0, direction="up")
         >>> tf.fit_transform(fd_grid)
-        array([[2]])
+        array([[ 2]])
     """
 
     def __init__(

diff --git a/skfda/preprocessing/feature_construction/_functions.py b/skfda/preprocessing/feature_construction/_functions.py
@@ -349,7 +349,7 @@ def number_crossings(
         FDataGrid created.
 
         >>> number_crossings(fd_grid, levels=0, direction="up")
-        array([[2]])
+        array([[ 2]])
     """
     # This is only defined for univariate functions
     check_fdata_dimensions(fd, dim_domain=1, dim_codomain=1)

diff --git a/skfda/preprocessing/feature_construction/_per_class_transformer.py b/skfda/preprocessing/feature_construction/_per_class_transformer.py
@@ -107,7 +107,7 @@ class PerClassTransformer(TransformerMixin[Input, Output, NDArrayInt]):
         Finally we can predict and check the score:
 
         >>> neigh1.predict(X_test1)
-        array([0, 0, 1, 0, 1, 1, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1,
+        array([ 0, 0, 1, 0, 1, 1, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1,
             1, 1, 1], dtype=int8)
 
         >>> round(neigh1.score(X_test1, y_test1), 3)
@@ -145,7 +145,7 @@ class PerClassTransformer(TransformerMixin[Input, Output, NDArrayInt]):
         >>> neigh2 = KNeighborsClassifier()
         >>> neigh2 = neigh2.fit(X_train2, y_train2)
         >>> neigh2.predict(X_test2)
-        array([1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1,
+        array([ 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1,
                1, 1, 1], dtype=int8)
 
         >>> round(neigh2.score(X_test2, y_test2), 3)

diff --git a/skfda/preprocessing/registration/validation.py b/skfda/preprocessing/registration/validation.py
@@ -103,10 +103,10 @@ class AmplitudePhaseDecompositionStats():
     :func:`mse_r_squared`, returned when `return_stats` is `True`.
 
     Args:
-        r_square (float): Squared correlation index :math:`R^2`.
-        mse_amplitude (float): Mean square error of amplitude
+        r_square: Squared correlation index :math:`R^2`.
+        mse_amplitude: Mean square error of amplitude
             :math:`\text{MSE}_{amp}`.
-        mse_phase (float): Mean square error of phase :math:`\text{MSE}_{pha}`.
+        mse_phase: Mean square error of phase :math:`\text{MSE}_{pha}`.
         c_r (float): Constant :math:`C_R`.
 
     """
@@ -297,10 +297,10 @@ def stats(
         X_mean = X.mean()
         y_mean = y.mean()
 
-        c_r = np.sum(l2_norm(X)**2) / np.sum(l2_norm(y)**2)
+        c_r = float(np.sum(l2_norm(X)**2) / np.sum(l2_norm(y)**2))
 
-        mse_amplitude = c_r * np.mean(l2_distance(y, y.mean())**2)
-        mse_phase = (c_r * l2_norm(y_mean)**2 - l2_norm(X_mean)**2).item()
+        mse_amplitude = float(c_r * np.mean(l2_distance(y, y.mean())**2))
+        mse_phase = float(c_r * l2_norm(y_mean)**2 - l2_norm(X_mean)**2)
 
         # Should be equal to np.mean(l2_distance(X, X_mean)**2)
         mse_total = mse_amplitude + mse_phase

diff --git a/skfda/preprocessing/smoothing/validation.py b/skfda/preprocessing/smoothing/validation.py
@@ -210,7 +210,7 @@ class SmoothingParameterSearch(
         >>> np.array(grid.cv_results_['mean_test_score']).round(2)
         array([-11.67, -12.37])
         >>> round(grid.best_score_, 2)
-        -11.67
+        np.float64(-11.67)
         >>> grid.best_params_['kernel_estimator__n_neighbors']
         2
         >>> grid.best_estimator_.hat_matrix().round(2)

diff --git a/skfda/representation/basis/_fdatabasis.py b/skfda/representation/basis/_fdatabasis.py
@@ -435,7 +435,7 @@ def sum(  # noqa: WPS125
         if min_count > 0:
             valid = ~np.isnan(self.coefficients)
             n_valid = np.sum(valid, axis=0)
-            coefs[n_valid < min_count] = np.NaN
+            coefs[n_valid < min_count] = np.nan
 
         return self.copy(
             coefficients=coefs,
@@ -1026,7 +1026,7 @@ def construct_array_type(cls) -> Type[FDataBasis]:  # noqa: D102
     def _na_repr(self) -> FDataBasis:
         return FDataBasis(
             basis=self.basis,
-            coefficients=((np.NaN,) * self.basis.n_basis,),
+            coefficients=((np.nan,) * self.basis.n_basis,),
         )
 
     def __eq__(self, other: Any) -> bool:

diff --git a/skfda/representation/grid.py b/skfda/representation/grid.py
@@ -591,7 +591,7 @@ def sum(  # noqa: WPS125
         if min_count > 0:
             valid = ~np.isnan(self.data_matrix)
             n_valid = np.sum(valid, axis=0)
-            data[n_valid < min_count] = np.NaN
+            data[n_valid < min_count] = np.nan
 
         return self.copy(
             data_matrix=data,
@@ -1518,7 +1518,7 @@ def _na_repr(self) -> FDataGrid:
             + (self.dim_codomain,)
         )
 
-        data_matrix = np.full(shape=shape, fill_value=np.NaN)
+        data_matrix = np.full(shape=shape, fill_value=np.nan)
 
         return FDataGrid(
             grid_points=self.grid_points,