[MNT] Fix #1039: Eliminate global seeds and replace sklearn.utils.check_random_state

skpro/regression/bootstrap.py

@@ -6,11 +6,11 @@
 import numpy as np
 import pandas as pd
 from sklearn import clone
-from sklearn.utils import check_random_state
 
 from skpro.distributions.empirical import Empirical
 from skpro.regression.base import BaseProbaRegressor
 from skpro.utils.numpy import flatten_to_1D_if_colvector
+from skpro.utils.random_state import check_random_state
 from skpro.utils.sampling import _random_ss_ix
 from skpro.utils.sklearn import prep_skl_df
 
@@ -110,7 +110,6 @@ def _fit(self, X, y):
         """
         estimator = self.estimator
         n_bootstrap_samples = self.n_bootstrap_samples
-        np.random.seed(self.random_state)
 
         inst_ix = X.index
         n = len(inst_ix)

skpro/regression/enbpi.py

@@ -6,11 +6,11 @@
 import numpy as np
 import pandas as pd
 from sklearn import clone
-from sklearn.utils import check_random_state
 
 from skpro.distributions.empirical import Empirical
 from skpro.regression.base import BaseProbaRegressor
 from skpro.utils.numpy import flatten_to_1D_if_colvector
+from skpro.utils.random_state import check_random_state
 from skpro.utils.sampling import _random_ss_ix
 from skpro.utils.sklearn import prep_skl_df
 
@@ -160,7 +160,6 @@ def _fit(self, X, y):
         """
         estimator = self.estimator
         n_bootstrap_samples = self.n_bootstrap_samples
-        np.random.seed(self.random_state)
 
         inst_ix = X.index
         n = len(inst_ix)

skpro/regression/ensemble/_bagging.py

@@ -10,6 +10,7 @@
 
 from skpro.distributions.mixture import Mixture
 from skpro.regression.base import BaseProbaRegressor
+from skpro.utils.random_state import check_random_state
 from skpro.utils.sampling import _random_ss_ix
 
 
@@ -98,6 +99,7 @@ def __init__(
         self.bootstrap = bootstrap
         self.bootstrap_features = bootstrap_features
         self.random_state = random_state
+        self._random_state = check_random_state(random_state)
 
         super().__init__()
 
@@ -129,8 +131,7 @@ def _fit(self, X, y, C=None):
         n_features = self.n_features
         bootstrap = self.bootstrap
         bootstrap_ft = self.bootstrap_features
-        random_state = self.random_state
-        np.random.seed(random_state)
+        # removed np.random.seed(random_state) mutation
 
         inst_ix = X.index
         col_ix = X.columns
@@ -154,14 +155,17 @@ def _fit(self, X, y, C=None):
             esti = estimator.clone()
             row_iloc = pd.RangeIndex(n)
             row_ss = _random_ss_ix(
-                row_iloc, size=n_samples_, replace=bootstrap, random_state=random_state
+                row_iloc,
+                size=n_samples_,
+                replace=bootstrap,
+                random_state=self._random_state,
             )
             inst_ix_i = inst_ix[row_ss]
             col_ix_i = _random_ss_ix(
                 col_ix,
                 size=n_features_,
                 replace=bootstrap_ft,
-                random_state=random_state,
+                random_state=self._random_state,
             )
 
             # store column subset for use in predict

skpro/regression/mdn.py

@@ -8,6 +8,7 @@
 from skpro.distributions.normal_mixture import NormalMixture
 from skpro.regression._bandwidth import bw_isj_1d, bw_silverman_1d
 from skpro.regression.base import BaseProbaRegressor
+from skpro.utils.random_state import check_random_state
 
 
 def _noise_scale_method_factor(n_samples, total_dim, method="silverman"):
@@ -241,6 +242,7 @@ def __init__(
         self.batch_size = batch_size
         self.device = device
         self.random_state = random_state
+        self._random_state = check_random_state(random_state)
 
         super().__init__()
 
@@ -438,7 +440,7 @@ def _fit(self, X, y, C=None):
 
         if self.random_state is not None:
             torch.manual_seed(self.random_state)
-            np.random.seed(self.random_state)
+            # removed np.random.seed(self.random_state) mutation
 
         input_noise_std = float(self.input_noise_std)
         target_noise_std = float(self.target_noise_std)

skpro/utils/random_state.py

@@ -2,7 +2,28 @@
 # copied from scikit-learn to avoid dependency on sklearn private methods
 
 import numpy as np
-from sklearn.utils import check_random_state
+
+
+def check_random_state(seed):
+    """Turn seed into a np.random.RandomState instance.
+
+    Parameters
+    ----------
+    seed : None, int or instance of RandomState
+        If seed is None, return the RandomState singleton used by np.random.
+        If seed is an int, return a new RandomState instance seeded with seed.
+        If seed is already a RandomState instance, return it.
+        Otherwise raise ValueError.
+    """
+    if seed is None or seed is np.random:
+        return np.random.mtrand._rand
+    if isinstance(seed, (int, np.integer)):
+        return np.random.RandomState(seed)
+    if isinstance(seed, np.random.RandomState):
+        return seed
+    raise ValueError(
+        "%r cannot be used to seed a numpy.random.RandomState instance" % seed
+    )
 
 
 def set_random_state(estimator, random_state=0):

skpro/utils/sampling.py

@@ -3,7 +3,7 @@
 
 __author__ = ["fkiraly"]
 
-from sklearn.utils import check_random_state
+from skpro.utils.random_state import check_random_state
 
 
 def _random_ss_ix(ix, size, replace=True, random_state=None):