scikit-learn-intelex 2025.4.0__py313-none-manylinux_2_28_x86_64.whl

sklearnex/covariance/incremental_covariance.py

@@ -0,0 +1,405 @@
+# ===============================================================================
+# Copyright 2024 Intel Corporation
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ===============================================================================
+
+import numbers
+import warnings
+
+import numpy as np
+from scipy import linalg
+from sklearn.base import BaseEstimator, clone
+from sklearn.covariance import EmpiricalCovariance as _sklearn_EmpiricalCovariance
+from sklearn.covariance import log_likelihood
+from sklearn.utils import check_array, gen_batches
+from sklearn.utils.validation import _num_features, check_is_fitted
+
+from daal4py.sklearn._n_jobs_support import control_n_jobs
+from daal4py.sklearn._utils import daal_check_version, sklearn_check_version
+from onedal.covariance import (
+    IncrementalEmpiricalCovariance as onedal_IncrementalEmpiricalCovariance,
+)
+from sklearnex import config_context
+
+from .._device_offload import dispatch, wrap_output_data
+from .._utils import IntelEstimator, PatchingConditionsChain, register_hyperparameters
+from ..metrics import pairwise_distances
+from ..utils._array_api import get_namespace
+
+if sklearn_check_version("1.2"):
+    from sklearn.utils._param_validation import Interval
+
+if sklearn_check_version("1.6"):
+    from sklearn.utils.validation import validate_data
+else:
+    validate_data = BaseEstimator._validate_data
+
+
+@control_n_jobs(decorated_methods=["partial_fit", "fit", "_onedal_finalize_fit"])
+class IncrementalEmpiricalCovariance(IntelEstimator, BaseEstimator):
+    """
+    Maximum likelihood covariance estimator that allows for the estimation when the data are split into
+    batches. The user can use the ``partial_fit`` method to provide a single batch of data or use the ``fit`` method to provide
+    the entire dataset.
+
+    Parameters
+    ----------
+    store_precision : bool, default=False
+        Specifies if the estimated precision is stored.
+
+    assume_centered : bool, default=False
+        If True, data are not centered before computation.
+        Useful when working with data whose mean is almost, but not exactly
+        zero.
+        If False (default), data are centered before computation.
+
+    batch_size : int, default=None
+        The number of samples to use for each batch. Only used when calling
+        ``fit``. If ``batch_size`` is ``None``, then ``batch_size``
+        is inferred from the data and set to ``5 * n_features``, to provide a
+        balance between approximation accuracy and memory consumption.
+
+    copy : bool, default=True
+        If False, X will be overwritten. ``copy=False`` can be used to
+        save memory but is unsafe for general use.
+
+    Attributes
+    ----------
+    location_ : ndarray of shape (n_features,)
+        Estimated location, i.e. the estimated mean.
+
+    covariance_ : ndarray of shape (n_features, n_features)
+        Estimated covariance matrix
+
+    n_samples_seen_ : int
+        The number of samples processed by the estimator. Will be reset on
+        new calls to ``fit``, but increments across ``partial_fit`` calls.
+
+    batch_size_ : int
+        Inferred batch size from ``batch_size``.
+
+    n_features_in_ : int
+        Number of features seen during ``fit`` or ``partial_fit``.
+
+    Note
+    ----
+    Serializing instances of this class will trigger a forced finalization of calculations.
+    Since finalize_fit can't be dispatched without directly provided queue
+    and the dispatching policy can't be serialized, the computation is finalized
+    during serialization and the policy is not saved in serialized data.
+
+    Examples
+    --------
+    >>> import numpy as np
+    >>> from sklearnex.covariance import IncrementalEmpiricalCovariance
+    >>> inccov = IncrementalEmpiricalCovariance(batch_size=1)
+    >>> X = np.array([[1, 2], [3, 4]])
+    >>> inccov.partial_fit(X[:1])
+    >>> inccov.partial_fit(X[1:])
+    >>> inccov.covariance_
+    np.array([[1., 1.],[1., 1.]])
+    >>> inccov.location_
+    np.array([2., 3.])
+    >>> inccov.fit(X)
+    >>> inccov.covariance_
+    np.array([[1., 1.],[1., 1.]])
+    >>> inccov.location_
+    np.array([2., 3.])
+    """
+
+    _onedal_incremental_covariance = staticmethod(onedal_IncrementalEmpiricalCovariance)
+
+    if sklearn_check_version("1.2"):
+        _parameter_constraints: dict = {
+            "store_precision": ["boolean"],
+            "assume_centered": ["boolean"],
+            "batch_size": [Interval(numbers.Integral, 1, None, closed="left"), None],
+            "copy": ["boolean"],
+        }
+
+    get_precision = _sklearn_EmpiricalCovariance.get_precision
+    error_norm = wrap_output_data(_sklearn_EmpiricalCovariance.error_norm)
+
+    def __init__(
+        self, *, store_precision=False, assume_centered=False, batch_size=None, copy=True
+    ):
+        self.assume_centered = assume_centered
+        self.store_precision = store_precision
+        self.batch_size = batch_size
+        self.copy = copy
+
+    def _onedal_supported(self, method_name, *data):
+        patching_status = PatchingConditionsChain(
+            f"sklearn.covariance.{self.__class__.__name__}.{method_name}"
+        )
+        return patching_status
+
+    def _onedal_finalize_fit(self, queue=None):
+        assert hasattr(self, "_onedal_estimator")
+        self._onedal_estimator.finalize_fit(queue=queue)
+        self._need_to_finalize = False
+
+        if not daal_check_version((2024, "P", 400)) and self.assume_centered:
+            location = self._onedal_estimator.location_[None, :]
+            self._onedal_estimator.covariance_ += np.dot(location.T, location)
+            self._onedal_estimator.location_ = np.zeros_like(np.squeeze(location))
+        if self.store_precision:
+            self.precision_ = linalg.pinvh(
+                self._onedal_estimator.covariance_, check_finite=False
+            )
+        else:
+            self.precision_ = None
+
+    @property
+    def covariance_(self):
+        if hasattr(self, "_onedal_estimator"):
+            if self._need_to_finalize:
+                self._onedal_finalize_fit()
+            return self._onedal_estimator.covariance_
+        else:
+            raise AttributeError(
+                f"'{self.__class__.__name__}' object has no attribute 'covariance_'"
+            )
+
+    @property
+    def location_(self):
+        if hasattr(self, "_onedal_estimator"):
+            if self._need_to_finalize:
+                self._onedal_finalize_fit()
+            return self._onedal_estimator.location_
+        else:
+            raise AttributeError(
+                f"'{self.__class__.__name__}' object has no attribute 'location_'"
+            )
+
+    def _onedal_partial_fit(self, X, queue=None, check_input=True):
+
+        first_pass = not hasattr(self, "n_samples_seen_") or self.n_samples_seen_ == 0
+
+        # finite check occurs on onedal side
+        if check_input:
+            if sklearn_check_version("1.2"):
+                self._validate_params()
+
+            if sklearn_check_version("1.0"):
+                X = validate_data(
+                    self,
+                    X,
+                    dtype=[np.float64, np.float32],
+                    reset=first_pass,
+                    copy=self.copy,
+                    force_all_finite=False,
+                )
+            else:
+                X = check_array(
+                    X,
+                    dtype=[np.float64, np.float32],
+                    copy=self.copy,
+                    force_all_finite=False,
+                )
+
+        onedal_params = {
+            "method": "dense",
+            "bias": True,
+            "assume_centered": self.assume_centered,
+        }
+        if not hasattr(self, "_onedal_estimator"):
+            self._onedal_estimator = self._onedal_incremental_covariance(**onedal_params)
+        try:
+            if first_pass:
+                self.n_samples_seen_ = X.shape[0]
+                self.n_features_in_ = X.shape[1]
+            else:
+                self.n_samples_seen_ += X.shape[0]
+
+            self._onedal_estimator.partial_fit(X, queue=queue)
+        finally:
+            self._need_to_finalize = True
+
+        return self
+
+    @wrap_output_data
+    def score(self, X_test, y=None):
+        xp, _ = get_namespace(X_test)
+
+        check_is_fitted(self)
+        location = self.location_
+        if sklearn_check_version("1.0"):
+            X = validate_data(
+                self,
+                X_test,
+                dtype=[np.float64, np.float32],
+                reset=False,
+            )
+        else:
+            X = check_array(
+                X_test,
+                dtype=[np.float64, np.float32],
+            )
+
+        if "numpy" not in xp.__name__:
+            location = xp.asarray(location, device=X_test.device)
+            # depending on the sklearn version, check_array
+            # and validate_data will return only numpy arrays
+            # which will break dpnp/dpctl support. If the
+            # array namespace isn't from numpy and the data
+            # is now a numpy array, it has been validated and
+            # the original can be used.
+            if isinstance(X, np.ndarray):
+                X = X_test
+
+        est = clone(self)
+        est.set_params(**{"assume_centered": True})
+
+        # test_cov is a numpy array, but calculated on device
+        test_cov = est.fit(X - location).covariance_
+        res = log_likelihood(test_cov, self.get_precision())
+
+        return res
+
+    def partial_fit(self, X, y=None, check_input=True):
+        """
+        Incremental fit with X. All of X is processed as a single batch.
+
+        Parameters
+        ----------
+        X : array-like of shape (n_samples, n_features)
+            Training data, where `n_samples` is the number of samples and
+            `n_features` is the number of features.
+
+        y : Ignored
+            Not used, present for API consistency by convention.
+
+        check_input : bool, default=True
+            Run check_array on X.
+
+        Returns
+        -------
+        self : object
+            Returns the instance itself.
+        """
+        return dispatch(
+            self,
+            "partial_fit",
+            {
+                "onedal": self.__class__._onedal_partial_fit,
+                "sklearn": None,
+            },
+            X,
+            check_input=check_input,
+        )
+
+    def fit(self, X, y=None):
+        """
+        Fit the model with X, using minibatches of size batch_size.
+
+        Parameters
+        ----------
+        X : array-like of shape (n_samples, n_features)
+            Training data, where `n_samples` is the number of samples and
+            `n_features` is the number of features.
+
+        y : Ignored
+            Not used, present for API consistency by convention.
+
+        Returns
+        -------
+        self : object
+            Returns the instance itself.
+        """
+
+        return dispatch(
+            self,
+            "fit",
+            {
+                "onedal": self.__class__._onedal_fit,
+                "sklearn": None,
+            },
+            X,
+        )
+
+    def _onedal_fit(self, X, queue=None):
+        self.n_samples_seen_ = 0
+        if hasattr(self, "_onedal_estimator"):
+            self._onedal_estimator._reset()
+
+        if sklearn_check_version("1.2"):
+            self._validate_params()
+
+        # finite check occurs on onedal side
+        if sklearn_check_version("1.0"):
+            X = validate_data(
+                self,
+                X,
+                dtype=[np.float64, np.float32],
+                copy=self.copy,
+                force_all_finite=False,
+            )
+        else:
+            X = check_array(
+                X, dtype=[np.float64, np.float32], copy=self.copy, force_all_finite=False
+            )
+            self.n_features_in_ = X.shape[1]
+
+        self.batch_size_ = self.batch_size if self.batch_size else 5 * self.n_features_in_
+
+        if X.shape[0] == 1:
+            warnings.warn(
+                "Only one sample available. You may want to reshape your data array"
+            )
+
+        for batch in gen_batches(X.shape[0], self.batch_size_):
+            X_batch = X[batch]
+            self._onedal_partial_fit(X_batch, queue=queue, check_input=False)
+
+        self._onedal_finalize_fit(queue=queue)
+
+        return self
+
+    # expose sklearnex pairwise_distances if mahalanobis distance eventually supported
+    def mahalanobis(self, X):
+        if sklearn_check_version("1.0"):
+            self._check_feature_names(X, reset=False)
+
+        xp, _ = get_namespace(X)
+        precision = self.get_precision()
+        # compute mahalanobis distances
+        # pairwise_distances will check n_features (via n_feature matching with
+        # self.location_) , and will check for finiteness via check array
+        # check_feature_names will match _validate_data functionally
+        location = self.location_[np.newaxis, :]
+        if "numpy" not in xp.__name__:
+            # Guarantee that inputs to pairwise_distances match in type and location
+            location = xp.asarray(location, device=X.device)
+
+        try:
+            dist = pairwise_distances(X, location, metric="mahalanobis", VI=precision)
+        except ValueError as e:
+            # Throw the expected sklearn error in an n_feature length violation
+            if "Incompatible dimension for X and Y matrices: X.shape[1] ==" in str(e):
+                raise ValueError(
+                    f"X has {_num_features(X)} features, but {self.__class__.__name__} "
+                    f"is expecting {self.n_features_in_} features as input."
+                )
+            else:
+                raise e
+
+        return (xp.reshape(dist, (-1,))) ** 2
+
+    _onedal_cpu_supported = _onedal_supported
+    _onedal_gpu_supported = _onedal_supported
+
+    mahalanobis.__doc__ = _sklearn_EmpiricalCovariance.mahalanobis.__doc__
+    error_norm.__doc__ = _sklearn_EmpiricalCovariance.error_norm.__doc__
+    score.__doc__ = _sklearn_EmpiricalCovariance.score.__doc__

sklearnex/covariance/tests/test_incremental_covariance.py

@@ -0,0 +1,287 @@
+# ===============================================================================
+# Copyright 2024 Intel Corporation
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ===============================================================================
+
+from os import environ
+
+from daal4py.sklearn._utils import sklearn_check_version
+
+# sklearn requires manual enabling of Scipy array API support
+# if `array-api-compat` package is present in environment
+# TODO: create generic approach to handle this for all tests
+if sklearn_check_version("1.6"):
+    environ["SCIPY_ARRAY_API"] = "1"
+
+
+import numpy as np
+import pytest
+from numpy.linalg import slogdet
+from numpy.testing import assert_allclose
+from scipy.linalg import pinvh
+from sklearn.covariance.tests.test_covariance import (
+    test_covariance,
+    test_EmpiricalCovariance_validates_mahalanobis,
+)
+from sklearn.datasets import load_diabetes
+from sklearn.decomposition import PCA
+
+from daal4py.sklearn._utils import daal_check_version
+from onedal.tests.utils._dataframes_support import (
+    _as_numpy,
+    _convert_to_dataframe,
+    get_dataframes_and_queues,
+)
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+@pytest.mark.parametrize("assume_centered", [True, False])
+def test_sklearnex_partial_fit_on_gold_data(dataframe, queue, dtype, assume_centered):
+    is_gpu = queue is not None and queue.sycl_device.is_gpu
+    if assume_centered and is_gpu and not daal_check_version((2025, "P", 0)):
+        pytest.skip(
+            "Due to a bug on oneDAL side, means are not set to zero when assume_centered=True"
+        )
+    from sklearnex.covariance import IncrementalEmpiricalCovariance
+
+    X = np.array([[0, 1], [0, 1]])
+    X = X.astype(dtype)
+    X_split = np.array_split(X, 2)
+    inccov = IncrementalEmpiricalCovariance(assume_centered=assume_centered)
+
+    for i in range(2):
+        X_split_df = _convert_to_dataframe(
+            X_split[i], sycl_queue=queue, target_df=dataframe
+        )
+        result = inccov.partial_fit(X_split_df)
+
+    if assume_centered:
+        expected_covariance = np.array([[0, 0], [0, 1]])
+        expected_means = np.array([0, 0])
+    else:
+        expected_covariance = np.array([[0, 0], [0, 0]])
+        expected_means = np.array([0, 1])
+
+    assert_allclose(expected_covariance, result.covariance_)
+    assert_allclose(expected_means, result.location_)
+
+    X = np.array([[1, 2], [3, 6]])
+    X = X.astype(dtype)
+    X_split = np.array_split(X, 2)
+    inccov = IncrementalEmpiricalCovariance(assume_centered=assume_centered)
+
+    for i in range(2):
+        X_split_df = _convert_to_dataframe(
+            X_split[i], sycl_queue=queue, target_df=dataframe
+        )
+        result = inccov.partial_fit(X_split_df)
+
+    if assume_centered:
+        expected_covariance = np.array([[5, 10], [10, 20]])
+        expected_means = np.array([0, 0])
+    else:
+        expected_covariance = np.array([[1, 2], [2, 4]])
+        expected_means = np.array([2, 4])
+
+    assert_allclose(expected_covariance, result.covariance_)
+    assert_allclose(expected_means, result.location_)
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+@pytest.mark.parametrize("batch_size", [2, 4])
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+def test_sklearnex_fit_on_gold_data(dataframe, queue, batch_size, dtype):
+    from sklearnex.covariance import IncrementalEmpiricalCovariance
+
+    X = np.array([[0, 1, 2, 3], [0, -1, -2, -3], [0, 1, 2, 3], [0, 1, 2, 3]])
+    X = X.astype(dtype)
+    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+    inccov = IncrementalEmpiricalCovariance(batch_size=batch_size)
+
+    result = inccov.fit(X_df)
+
+    expected_covariance = np.array(
+        [[0, 0, 0, 0], [0, 0.75, 1.5, 2.25], [0, 1.5, 3, 4.5], [0, 2.25, 4.5, 6.75]]
+    )
+    expected_means = np.array([0, 0.5, 1, 1.5])
+
+    assert_allclose(expected_covariance, result.covariance_)
+    assert_allclose(expected_means, result.location_)
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+@pytest.mark.parametrize("num_batches", [2, 10])
+@pytest.mark.parametrize("row_count", [100, 1000])
+@pytest.mark.parametrize("column_count", [10, 100])
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+def test_sklearnex_partial_fit_on_random_data(
+    dataframe, queue, num_batches, row_count, column_count, dtype
+):
+    from sklearnex.covariance import IncrementalEmpiricalCovariance
+
+    seed = 77
+    gen = np.random.default_rng(seed)
+    X = gen.uniform(low=-0.3, high=+0.7, size=(row_count, column_count))
+    X = X.astype(dtype)
+    X_split = np.array_split(X, num_batches)
+    inccov = IncrementalEmpiricalCovariance()
+
+    for i in range(num_batches):
+        X_split_df = _convert_to_dataframe(
+            X_split[i], sycl_queue=queue, target_df=dataframe
+        )
+        result = inccov.partial_fit(X_split_df)
+
+    expected_covariance = np.cov(X.T, bias=1)
+    expected_means = np.mean(X, axis=0)
+
+    assert_allclose(expected_covariance, result.covariance_, atol=1e-6)
+    assert_allclose(expected_means, result.location_, atol=1e-6)
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+@pytest.mark.parametrize("num_batches", [2, 10])
+@pytest.mark.parametrize("row_count", [100, 1000])
+@pytest.mark.parametrize("column_count", [10, 100])
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+@pytest.mark.parametrize("assume_centered", [True, False])
+def test_sklearnex_fit_on_random_data(
+    dataframe, queue, num_batches, row_count, column_count, dtype, assume_centered
+):
+    is_gpu = queue is not None and queue.sycl_device.is_gpu
+    if assume_centered and is_gpu and not daal_check_version((2025, "P", 0)):
+        pytest.skip(
+            "Due to a bug on oneDAL side, means are not set to zero when assume_centered=True"
+        )
+    from sklearnex.covariance import IncrementalEmpiricalCovariance
+
+    seed = 77
+    gen = np.random.default_rng(seed)
+    X = gen.uniform(low=-0.3, high=+0.7, size=(row_count, column_count))
+    X = X.astype(dtype)
+    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+    batch_size = row_count // num_batches
+    inccov = IncrementalEmpiricalCovariance(
+        batch_size=batch_size, assume_centered=assume_centered
+    )
+
+    result = inccov.fit(X_df)
+
+    if assume_centered:
+        expected_covariance = np.dot(X.T, X) / X.shape[0]
+        expected_means = np.zeros_like(X[0])
+    else:
+        expected_covariance = np.cov(X.T, bias=1)
+        expected_means = np.mean(X, axis=0)
+
+    assert_allclose(expected_covariance, result.covariance_, atol=1e-6)
+    assert_allclose(expected_means, result.location_, atol=1e-6)
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+def test_whitened_toy_score(dataframe, queue):
+    from sklearnex.covariance import IncrementalEmpiricalCovariance
+
+    # Load a sklearn toy dataset with sufficient data
+    X, _ = load_diabetes(return_X_y=True)
+    n = X.shape[1]
+
+    # Transform the data into uncorrelated, unity variance components
+    X = PCA(whiten=True).fit_transform(X)
+
+    # change dataframe
+    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+
+    # fit data
+    est = IncrementalEmpiricalCovariance()
+    est.fit(X_df)
+    # location_ attribute approximately zero (10,), covariance_ identity (10,10)
+
+    # The log-likelihood can be calculated simply due to covariance_
+    # use of scipy.linalg.pinvh, np.linalg.sloget and np.cov for estimator
+    # independence
+    expected_result = (
+        -(n - slogdet(pinvh(np.cov(X.T, bias=1)))[1] + n * np.log(2 * np.pi)) / 2
+    )
+    # expected_result = -14.1780602988
+    result = _as_numpy(est.score(X_df))
+    assert_allclose(expected_result, result, atol=1e-6)
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+def test_sklearnex_incremental_estimatior_pickle(dataframe, queue, dtype):
+    import pickle
+
+    from sklearnex.covariance import IncrementalEmpiricalCovariance
+
+    inccov = IncrementalEmpiricalCovariance()
+
+    # Check that estimator can be serialized without any data.
+    dump = pickle.dumps(inccov)
+    inccov_loaded = pickle.loads(dump)
+
+    seed = 77
+    gen = np.random.default_rng(seed)
+    X = gen.uniform(low=-0.3, high=+0.7, size=(10, 10))
+    X = X.astype(dtype)
+    X_split = np.array_split(X, 2)
+    X_split_df = _convert_to_dataframe(X_split[0], sycl_queue=queue, target_df=dataframe)
+    inccov.partial_fit(X_split_df)
+    inccov_loaded.partial_fit(X_split_df)
+
+    # Check that estimator can be serialized after partial_fit call.
+    dump = pickle.dumps(inccov_loaded)
+    inccov_loaded = pickle.loads(dump)
+
+    assert inccov.batch_size == inccov_loaded.batch_size
+    assert inccov.n_features_in_ == inccov_loaded.n_features_in_
+    assert inccov.n_samples_seen_ == inccov_loaded.n_samples_seen_
+    if hasattr(inccov, "_parameter_constraints"):
+        assert inccov._parameter_constraints == inccov_loaded._parameter_constraints
+    assert inccov.n_jobs == inccov_loaded.n_jobs
+
+    X_split_df = _convert_to_dataframe(X_split[1], sycl_queue=queue, target_df=dataframe)
+    inccov.partial_fit(X_split_df)
+    inccov_loaded.partial_fit(X_split_df)
+    dump = pickle.dumps(inccov)
+    inccov_loaded = pickle.loads(dump)
+
+    assert_allclose(inccov.location_, inccov_loaded.location_, atol=1e-6)
+    assert_allclose(inccov.covariance_, inccov_loaded.covariance_, atol=1e-6)
+
+    # Check that finalized estimator can be serialized.
+    dump = pickle.dumps(inccov_loaded)
+    inccov_loaded = pickle.loads(dump)
+
+    assert_allclose(inccov.location_, inccov_loaded.location_, atol=1e-6)
+    assert_allclose(inccov.covariance_, inccov_loaded.covariance_, atol=1e-6)
+
+
+# Monkeypatch IncrementalEmpiricalCovariance into relevant sklearn.covariance tests
+@pytest.mark.allow_sklearn_fallback
+@pytest.mark.parametrize(
+    "sklearn_test",
+    [
+        test_covariance,
+        test_EmpiricalCovariance_validates_mahalanobis,
+    ],
+)
+def test_IncrementalEmpiricalCovariance_against_sklearn(monkeypatch, sklearn_test):
+    from sklearnex.covariance import IncrementalEmpiricalCovariance
+
+    class_name = ".".join([sklearn_test.__module__, "EmpiricalCovariance"])
+    monkeypatch.setattr(class_name, IncrementalEmpiricalCovariance)
+    sklearn_test()