scikit-learn · adrinjalali · Mar 28, 2022 · Apr 1, 2022 · Apr 1, 2022 · Apr 1, 2022
diff --git a/doc/whats_new/v1.1.rst b/doc/whats_new/v1.1.rst
@@ -130,6 +130,12 @@ Changelog
   error message when setting invalid hyper-parameters with `set_params`.
   :pr:`21542` by :user:`Olivier Grisel <ogrisel>`.
 
+:mod:`sklearn.base`
+...................
+- |Fix| :func:`base.clone` now clones estimators which store a copy of their
+  constructor arguments as long as they are not arrays. :pr:`22973` by `Adrin
+  Jalali`_.
+
 :mod:`sklearn.calibration`
 ..........................
 

diff --git a/sklearn/base.py b/sklearn/base.py
@@ -92,10 +92,20 @@ def clone(estimator, *, safe=True):
     for name in new_object_params:
         param1 = new_object_params[name]
         param2 = params_set[name]
-        if param1 is not param2:
+        init_param_error = (
+            f"Cannot clone object {estimator}, as the constructor "
+            f"either does not set or modifies parameter {name}."
+        )
+        if param1 is param2:
+            continue
+        try:
+            if param1 != param2:
+                raise RuntimeError(init_param_error)
+        except ValueError as e:
             raise RuntimeError(
-                "Cannot clone object %s, as the constructor "
-                "either does not set or modifies parameter %s" % (estimator, name)
+                init_param_error
+                + " The following error occurred while comparing old and new"
+                f" parameters: {str(e)}"
             )
     return new_object
 

diff --git a/sklearn/conftest.py b/sklearn/conftest.py
@@ -19,10 +19,7 @@
 from sklearn.datasets import fetch_kddcup99
 from sklearn.datasets import fetch_olivetti_faces
 from sklearn.datasets import fetch_rcv1
-
-
-# This plugin is necessary to define the random seed fixture
-pytest_plugins = ("sklearn.tests.random_seed",)
+from sklearn.tests import random_seed
 
 
 if parse_version(pytest.__version__) < parse_version(PYTEST_MIN_VERSION):
@@ -249,3 +246,7 @@ def pytest_configure(config):
         matplotlib.use("agg")
     except ImportError:
         pass
+
+    # Register global_random_seed plugin if it is not already registered
+    if not config.pluginmanager.hasplugin("sklearn.tests.random_seed"):
+        config.pluginmanager.register(random_seed)
diff --git a/sklearn/tests/test_base.py b/sklearn/tests/test_base.py
@@ -2,6 +2,7 @@
 # License: BSD 3 clause
 
 import re
+import copy
 import numpy as np
 import scipy.sparse as sp
 import pytest
@@ -76,8 +77,8 @@ class ModifyInitParams(BaseEstimator):
     Doesn't fulfill a is a
     """
 
-    def __init__(self, a=np.array([0])):
-        self.a = a.copy()
+    def __init__(self, a=np.array([0, 1], dtype=int)):
+        self.a = a.astype(float)
 
 
 class Buggy(BaseEstimator):
@@ -210,6 +211,20 @@ def test_clone_class_rather_than_instance():
         clone(MyEstimator)
 
 
+def test_clone_eq_on_param():
+    # test that copying given parameters in constructor is a valid operation
+    # for `clone`
+    # regression test for #22857
+    class Est(BaseEstimator):
+        def __init__(self, param=None):
+            self.param = copy.deepcopy(param)
+
+    est1 = Est(param={"key": "value"})
+    # this should not raise
+    est2 = clone(est1)
+    assert est1.param == est2.param
+
+
 def test_repr():
     # Smoke test the repr of the base estimator.
     my_estimator = MyEstimator()

diff --git a/sklearn/utils/_pprint.py b/sklearn/utils/_pprint.py
@@ -98,8 +98,14 @@ def _changed_params(estimator):
     def has_changed(k, v):
         if k not in init_params:  # happens if k is part of a **kwargs
             return True
-        if init_params[k] == inspect._empty:  # k has no default value
-            return True
+        try:
+            if init_params[k] == inspect._empty:  # k has no default value
+                return True
+        except ValueError:
+            # skip if "ValueError: The truth value of an array with more than
+            # one element is ambiguous. Use a.any() or a.all()" is raised
+            # during comparison.
+            pass
         # try to avoid calling repr on nested estimators
         if isinstance(v, BaseEstimator) and v.__class__ != init_params[k].__class__:
             return True