Typing inference module.

skfda/_utils/__init__.py

@@ -9,7 +9,6 @@
     ],
     submod_attrs={
         "_utils": [
-            "RandomStateLike",
             "_cartesian_product",
             "_check_array_key",
             "_check_estimator",
@@ -36,7 +35,6 @@
 if TYPE_CHECKING:
 
     from ._utils import (
-        RandomStateLike as RandomStateLike,
         _cartesian_product as _cartesian_product,
         _check_array_key as _check_array_key,
         _check_estimator as _check_estimator,

skfda/_utils/_utils.py

@@ -31,8 +31,6 @@
 from ..typing._numpy import NDArrayAny, NDArrayFloat, NDArrayInt, NDArrayStr
 from ._sklearn_adapter import BaseEstimator
 
-RandomStateLike = Optional[Union[int, np.random.RandomState]]
-
 ArrayDTypeT = TypeVar("ArrayDTypeT", bound="np.generic")
 
 if TYPE_CHECKING:

skfda/datasets/_samples_generators.py

@@ -3,19 +3,14 @@
 
 import numpy as np
 import scipy.integrate
-import sklearn.utils
 from scipy.stats import multivariate_normal
 
-from .._utils import (
-    RandomStateLike,
-    _cartesian_product,
-    _to_grid_points,
-    normalize_warping,
-)
+from .._utils import _cartesian_product, _to_grid_points, normalize_warping
 from ..misc import covariances
+from ..misc.validation import validate_random_state
 from ..representation import FDataGrid
 from ..representation.interpolation import SplineInterpolation
-from ..typing._base import DomainRangeLike, GridPointsLike
+from ..typing._base import DomainRangeLike, GridPointsLike, RandomStateLike
 from ..typing._numpy import NDArrayFloat
 
 MeanCallable = Callable[[np.ndarray], np.ndarray]
@@ -62,7 +57,7 @@ def make_gaussian(
         Gaussian processes.
 
     """
-    random_state = sklearn.utils.check_random_state(random_state)
+    random_state = validate_random_state(random_state)
 
     if cov is None:
         cov = covariances.Brownian()
@@ -193,7 +188,7 @@ def make_sinusoidal_process(
         :class:`FDataGrid` object comprising all the samples.
 
     """
-    random_state = sklearn.utils.check_random_state(random_state)
+    random_state = validate_random_state(random_state)
 
     t = np.linspace(start, stop, n_features)
 
@@ -256,7 +251,7 @@ def make_multimodal_landmarks(
         sample i.
 
     """
-    random_state = sklearn.utils.check_random_state(random_state)
+    random_state = validate_random_state(random_state)
 
     modes_location = np.linspace(start, stop, n_modes + 2)[1:-1]
     modes_location = np.repeat(
@@ -329,7 +324,7 @@ def make_multimodal_samples(
         :class:`FDataGrid` object comprising all the samples.
 
     """
-    random_state = sklearn.utils.check_random_state(random_state)
+    random_state = validate_random_state(random_state)
 
     if modes_location is None:
 
@@ -446,7 +441,7 @@ def make_random_warping(
     # Based on the original implementation of J. D. Tucker in the
     # package python_fdasrsf <https://github.com/jdtuck/fdasrsf_python>.
 
-    random_state = sklearn.utils.check_random_state(random_state)
+    random_state = validate_random_state(random_state)
 
     freq = shape_parameter + 1
 

skfda/exploratory/outliers/_directional_outlyingness.py

@@ -8,11 +8,11 @@
 import scipy.stats
 from numpy import linalg as la
 from sklearn.covariance import MinCovDet
-from sklearn.utils.validation import check_random_state
 
-from ..._utils import RandomStateLike
 from ..._utils._sklearn_adapter import BaseEstimator, OutlierMixin
+from ...misc.validation import validate_random_state
 from ...representation import FDataGrid
+from ...typing._base import RandomStateLike
 from ...typing._numpy import NDArrayFloat, NDArrayInt
 from ..depth.multivariate import Depth, ProjectionDepth
 from . import _directional_outlyingness_experiment_results as experiments
@@ -495,7 +495,7 @@ def fit_predict(  # noqa: D102
         y: object = None,
     ) -> NDArrayInt:
 
-        self.random_state_ = check_random_state(self.random_state)
+        self.random_state_ = validate_random_state(self.random_state)
         self.points_ = self._compute_points(X)
 
         # The square mahalanobis distances of the samples are

skfda/inference/__init__.py

@@ -1 +1,9 @@
-from . import anova, hotelling
+import lazy_loader as lazy
+
+__getattr__, __dir__, __all__ = lazy.attach(
+    __name__,
+    submodules=[
+        "anova",
+        "hotelling",
+    ],
+)

skfda/inference/anova/_anova_oneway.py

@@ -3,13 +3,13 @@
 from typing import Sequence, Tuple, TypeVar, overload
 
 import numpy as np
-from sklearn.utils import check_random_state
 from typing_extensions import Literal
 
-from ..._utils import RandomStateLike
 from ...datasets import make_gaussian
 from ...misc.metrics import lp_distance
+from ...misc.validation import validate_random_state
 from ...representation import FData, FDataGrid, concatenate
+from ...typing._base import RandomStateLike
 from ...typing._numpy import ArrayLike, NDArrayFloat
 
 
@@ -95,7 +95,7 @@ def _v_asymptotic_stat_with_reps(
     *fds: FData,
     weights: ArrayLike,
     p: int = 2,
-) -> float:
+) -> NDArrayFloat:
     """Vectorized version of v_asymptotic_stat for repetitions."""
     weights = np.asarray(weights)
     if len(weights) != len(fds):
@@ -112,7 +112,7 @@ def _v_asymptotic_stat_with_reps(
         right_fd = fds[pair[0]] * coef
         results[i] = lp_distance(left_fd, right_fd, p=p) ** p
 
-    return np.sum(results, axis=0)
+    return np.sum(results, axis=0)  # type: ignore[no-any-return]
 
 
 def v_asymptotic_stat(
@@ -206,7 +206,7 @@ def _anova_bootstrap(
     sizes = [fd.n_samples for fd in fd_grouped]
 
     # Instance a random state object in case random_state is an int
-    random_state = check_random_state(random_state)
+    random_state = validate_random_state(random_state)
 
     if equal_var:
         k_est = concatenate(fd_grouped).cov().data_matrix[0, ..., 0]
@@ -233,7 +233,6 @@ def _anova_bootstrap(
         for i in range(n_groups)
     ]
 
-    v_samples = np.empty(n_reps)
     return _v_asymptotic_stat_with_reps(*sim, weights=sizes, p=p)
 
 
@@ -385,7 +384,7 @@ def oneway_anova(
         equal_var=equal_var,
     )
 
-    p_value = np.sum(simulation > vn) / len(simulation)
+    p_value = float(np.sum(simulation > vn) / len(simulation))
 
     if return_dist:
         return vn, p_value, simulation

skfda/inference/hotelling/_hotelling.py

@@ -5,11 +5,11 @@
 
 import numpy as np
 import scipy.special
-from sklearn.utils import check_random_state
 from typing_extensions import Literal
 
-from ..._utils import RandomStateLike
+from ...misc.validation import validate_random_state
 from ...representation import FData, FDataBasis
+from ...typing._base import RandomStateLike
 from ...typing._numpy import NDArrayFloat
 
 
@@ -188,8 +188,8 @@ def hotelling_test_ind(
         TypeError: In case of bad arguments.
 
     Examples:
-        >>> from skfda.inference.hotelling import hotelling_t2
-        >>> from skfda.representation import FDataGrid, basis
+        >>> from skfda.inference.hotelling import hotelling_test_ind
+        >>> from skfda.representation import FDataGrid
         >>> from numpy import printoptions
 
         >>> fd1 = FDataGrid([[1, 1, 1], [3, 3, 3]])
@@ -223,7 +223,7 @@ def hotelling_test_ind(
     indices = np.arange(n)
 
     if n_reps is not None:  # Computing n_reps random permutations
-        random_state = check_random_state(random_state)
+        random_state = validate_random_state(random_state)
         dist = np.empty(n_reps)
         for i in range(n_reps):
             random_state.shuffle(indices)
@@ -238,7 +238,7 @@ def hotelling_test_ind(
             sample1, sample2 = sample[sample1_i], sample[sample2_i]
             dist[i] = hotelling_t2(sample1, sample2)
 
-    p_value = np.sum(dist > t2_0) / len(dist)
+    p_value = float(np.sum(dist > t2_0) / len(dist))
 
     if return_dist:
         return t2_0, p_value, dist

skfda/misc/validation.py

@@ -7,9 +7,16 @@
 from typing import Container, Sequence, Tuple, cast
 
 import numpy as np
+from sklearn.utils import check_random_state as _check_random_state
 
 from ..representation import FData, FDataBasis, FDataGrid
-from ..typing._base import DomainRange, DomainRangeLike, EvaluationPoints
+from ..typing._base import (
+    DomainRange,
+    DomainRangeLike,
+    EvaluationPoints,
+    RandomState,
+    RandomStateLike,
+)
 from ..typing._numpy import ArrayLike
 
 
@@ -255,3 +262,11 @@ def validate_domain_range(domain_range: DomainRangeLike) -> DomainRange:
     domain_range = cast(Sequence[Sequence[float]], domain_range)
 
     return tuple(_validate_domain_range_limits(s) for s in domain_range)
+
+
+def validate_random_state(random_state: RandomStateLike) -> RandomState:
+    """Validate random state or seed."""
+    if isinstance(random_state, np.random.Generator):
+        return random_state
+
+    return _check_random_state(random_state)  # type: ignore[no-any-return]

skfda/ml/clustering/_kmeans.py

@@ -8,13 +8,15 @@
 
 import numpy as np
 from sklearn.base import BaseEstimator, ClusterMixin, TransformerMixin
-from sklearn.utils import check_random_state
 from sklearn.utils.validation import check_is_fitted
 
-from ..._utils import RandomStateLike
 from ...misc.metrics import PairwiseMetric, l2_distance
-from ...misc.validation import check_fdata_same_dimensions
+from ...misc.validation import (
+    check_fdata_same_dimensions,
+    validate_random_state,
+)
 from ...representation import FDataGrid
+from ...typing._base import RandomStateLike
 from ...typing._metric import Metric
 from ...typing._numpy import NDArrayAny, NDArrayFloat, NDArrayInt
 
@@ -309,7 +311,7 @@ def fit(
 
         """
         fdata = self._check_clustering(X)
-        random_state = check_random_state(self.random_state)
+        random_state = validate_random_state(self.random_state)
 
         self._check_params()
 

skfda/preprocessing/dim_reduction/variable_selection/mrmr.py

@@ -21,12 +21,13 @@
 )
 from typing_extensions import Final, Literal
 
-from ...._utils import RandomStateLike, _compute_dependence, _DependenceMeasure
+from ...._utils import _compute_dependence, _DependenceMeasure
 from ...._utils._sklearn_adapter import (
     BaseEstimator,
     InductiveTransformerMixin,
 )
 from ....representation.grid import FDataGrid
+from ....typing._base import RandomStateLike
 from ....typing._numpy import NDArrayFloat, NDArrayInt, NDArrayReal
 
 _Criterion = Callable[[NDArrayFloat, NDArrayFloat], NDArrayFloat]

skfda/typing/_base.py

@@ -1,6 +1,7 @@
 """Common types."""
 from typing import Optional, Sequence, Tuple, TypeVar, Union
 
+import numpy as np
 from typing_extensions import Protocol
 
 from ._numpy import ArrayLike, NDArrayFloat
@@ -23,6 +24,10 @@
 EvaluationPoints = NDArrayFloat
 
 
+RandomStateLike = Union[int, np.random.RandomState, np.random.Generator, None]
+RandomState = Union[np.random.RandomState, np.random.Generator]
+
+
 class Vector(Protocol):
     """
     Protocol representing a generic vector.