scikit-learn-intelex 2024.0.1__py310-none-win_amd64.whl → 2025.1.0__py310-none-win_amd64.whl

{scikit_learn_intelex-2024.0.1.data/data/Lib/site-packages/sklearnex/basic_statistics → scikit_learn_intelex-2025.1.0.data/data/Lib/site-packages/onedal/decomposition}/__init__.py

@@ -1,4 +1,3 @@
-#!/usr/bin/env python
 # ==============================================================================
 # Copyright 2023 Intel Corporation
 #
@@ -15,6 +14,7 @@
 # limitations under the License.
 # ==============================================================================
 
-from .basic_statistics import BasicStatistics
+from .incremental_pca import IncrementalPCA
+from .pca import PCA
 
-__all__ = ["BasicStatistics"]
+__all__ = ["IncrementalPCA", "PCA"]

scikit_learn_intelex-2025.1.0.data/data/Lib/site-packages/onedal/decomposition/incremental_pca.py

@@ -0,0 +1,204 @@
+# ==============================================================================
+# 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 numpy as np
+
+from daal4py.sklearn._utils import get_dtype
+
+from ..datatypes import _convert_to_supported, from_table, to_table
+from ..utils import _check_array
+from .pca import BasePCA
+
+
+class IncrementalPCA(BasePCA):
+    """
+    Incremental estimator for PCA based on oneDAL implementation.
+    Allows to compute PCA if data are splitted into batches.
+
+    Parameters
+    ----------
+    n_components : int, default=None
+        Number of components to keep. If ``n_components`` is ``None``,
+        then ``n_components`` is set to ``min(n_samples, n_features)``.
+
+    is_deterministic : bool, default=True
+        When True the ``components_`` vectors are chosen in deterministic
+        way, otherwise some of them can be oppositely directed.
+
+    method : string, default='cov'
+        Method used on oneDAL side to compute result.
+
+    whiten : bool, default=False
+        When True the ``components_`` vectors are divided
+        by ``n_samples`` times ``components_`` to ensure uncorrelated outputs
+        with unit component-wise variances.
+
+        Whitening will remove some information from the transformed signal
+        (the relative variance scales of the components) but can sometimes
+        improve the predictive accuracy of the downstream estimators by
+        making data respect some hard-wired assumptions.
+
+    Attributes
+    ----------
+        components_ : ndarray of shape (n_components, n_features)
+        Principal axes in feature space, representing the directions of
+        maximum variance in the data. Equivalently, the right singular
+        vectors of the centered input data, parallel to its eigenvectors.
+        The components are sorted by decreasing ``explained_variance_``.
+
+        explained_variance_ : ndarray of shape (n_components,)
+            Variance explained by each of the selected components.
+
+        explained_variance_ratio_ : ndarray of shape (n_components,)
+            Percentage of variance explained by each of the selected components.
+            If all components are stored, the sum of explained variances is equal
+            to 1.0.
+
+        singular_values_ : ndarray of shape (n_components,)
+            The singular values corresponding to each of the selected components.
+            The singular values are equal to the 2-norms of the ``n_components``
+            variables in the lower-dimensional space.
+
+        mean_ : ndarray of shape (n_features,)
+            Per-feature empirical mean, aggregate over calls to ``partial_fit``.
+
+        var_ : ndarray of shape (n_features,)
+            Per-feature empirical variance, aggregate over calls to
+            ``partial_fit``.
+
+        noise_variance_ : float
+            Equal to the average of (min(n_features, n_samples) - n_components)
+            smallest eigenvalues of the covariance matrix of X.
+
+    """
+
+    def __init__(
+        self,
+        n_components=None,
+        is_deterministic=True,
+        method="cov",
+        whiten=False,
+    ):
+        self.n_components = n_components
+        self.method = method
+        self.is_deterministic = is_deterministic
+        self.whiten = whiten
+        self._reset()
+
+    def _reset(self):
+        self._partial_result = self._get_backend(
+            "decomposition", "dim_reduction", "partial_train_result"
+        )
+        if hasattr(self, "components_"):
+            del self.components_
+
+    def partial_fit(self, X, queue):
+        """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.
+
+        Returns
+        -------
+        self : object
+            Returns the instance itself.
+        """
+        X = _check_array(X)
+        n_samples, n_features = X.shape
+
+        first_pass = not hasattr(self, "components_")
+        if first_pass:
+            self.components_ = None
+            self.n_samples_seen_ = n_samples
+            self.n_features_in_ = n_features
+        else:
+            self.n_samples_seen_ += n_samples
+
+        if self.n_components is None:
+            if self.components_ is None:
+                self.n_components_ = min(n_samples, n_features)
+            else:
+                self.n_components_ = self.components_.shape[0]
+        else:
+            self.n_components_ = self.n_components
+
+        self._queue = queue
+
+        policy = self._get_policy(queue, X)
+        X = _convert_to_supported(policy, X)
+
+        if not hasattr(self, "_dtype"):
+            self._dtype = get_dtype(X)
+            self._params = self._get_onedal_params(X)
+
+        X_table = to_table(X)
+        self._partial_result = self._get_backend(
+            "decomposition",
+            "dim_reduction",
+            "partial_train",
+            policy,
+            self._params,
+            self._partial_result,
+            X_table,
+        )
+        return self
+
+    def finalize_fit(self, queue=None):
+        """
+        Finalizes principal components computation and obtains resulting
+        attributes from the current `_partial_result`.
+
+        Parameters
+        ----------
+        queue : dpctl.SyclQueue
+            Not used here, added for API conformance
+
+        Returns
+        -------
+        self : object
+            Returns the instance itself.
+        """
+        if queue is not None:
+            policy = self._get_policy(queue)
+        else:
+            policy = self._get_policy(self._queue)
+        result = self._get_backend(
+            "decomposition",
+            "dim_reduction",
+            "finalize_train",
+            policy,
+            self._params,
+            self._partial_result,
+        )
+        self.mean_ = from_table(result.means).ravel()
+        self.var_ = from_table(result.variances).ravel()
+        self.components_ = from_table(result.eigenvectors)
+        self.singular_values_ = np.nan_to_num(from_table(result.singular_values).ravel())
+        self.explained_variance_ = np.maximum(from_table(result.eigenvalues).ravel(), 0)
+        self.explained_variance_ratio_ = from_table(
+            result.explained_variances_ratio
+        ).ravel()
+        self.noise_variance_ = self._compute_noise_variance(
+            self.n_components_, min(self.n_samples_seen_, self.n_features_in_)
+        )
+
+        return self

scikit_learn_intelex-2025.1.0.data/data/Lib/site-packages/onedal/decomposition/pca.py

@@ -0,0 +1,186 @@
+# ==============================================================================
+# Copyright 2023 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
+from abc import ABCMeta
+
+import numpy as np
+from sklearn.decomposition._pca import _infer_dimension
+from sklearn.utils.extmath import stable_cumsum
+
+from ..common._base import BaseEstimator
+from ..datatypes import _convert_to_supported, from_table, to_table
+
+
+class BasePCA(BaseEstimator, metaclass=ABCMeta):
+    """
+    Base class for PCA oneDAL implementation.
+    """
+
+    def __init__(
+        self,
+        n_components=None,
+        is_deterministic=True,
+        method="cov",
+        whiten=False,
+    ):
+        self.n_components = n_components
+        self.method = method
+        self.is_deterministic = is_deterministic
+        self.whiten = whiten
+
+    def _get_onedal_params(self, data, stage=None):
+        if stage is None:
+            n_components = self._resolve_n_components_for_training(data.shape)
+        elif stage == "predict":
+            n_components = self.n_components_
+        return {
+            "fptype": "float" if data.dtype == np.float32 else "double",
+            "method": self.method,
+            "n_components": n_components,
+            "is_deterministic": self.is_deterministic,
+            "whiten": self.whiten,
+        }
+
+    def _validate_n_components(self, n_components, n_samples, n_features):
+        if n_components is None:
+            n_components = min(n_samples, n_features)
+        if n_components == "mle":
+            if n_samples < n_features:
+                raise ValueError(
+                    "n_components='mle' is only supported if n_samples >= n_features"
+                )
+        elif not 0 <= n_components <= min(n_samples, n_features):
+            raise ValueError(
+                "n_components=%r must be between 0 and "
+                "min(n_samples, n_features)=%r with "
+                "svd_solver='full'" % (n_components, min(n_samples, n_features))
+            )
+        elif n_components >= 1:
+            if not isinstance(n_components, numbers.Integral):
+                raise ValueError(
+                    "n_components=%r must be of type int "
+                    "when greater than or equal to 1, "
+                    "was of type=%r" % (n_components, type(n_components))
+                )
+
+    def _resolve_n_components_for_training(self, shape_tuple):
+        if self.n_components is None or self.n_components == "mle":
+            return min(shape_tuple)
+        elif (
+            isinstance(self.n_components, float)
+            and self.n_components > 0.0
+            and self.n_components <= 1.0
+        ):
+            return min(shape_tuple)
+        else:
+            return self.n_components
+
+    def _resolve_n_components_for_result(self, shape_tuple):
+        if self.n_components is None:
+            return min(shape_tuple)
+        elif self.n_components == "mle":
+            return _infer_dimension(self.explained_variance_, shape_tuple[0])
+        elif 0.0 < self.n_components < 1.0:
+            ratio_cumsum = stable_cumsum(self.explained_variance_ratio_)
+            return np.searchsorted(ratio_cumsum, self.n_components, side="right") + 1
+        elif isinstance(self.n_components, float) and self.n_components == 1.0:
+            return min(shape_tuple)
+        else:
+            return self.n_components
+
+    def _compute_noise_variance(self, n_components, n_sf_min):
+        if n_components < n_sf_min:
+            if len(self.explained_variance_) == n_sf_min:
+                return self.explained_variance_[n_components:].mean()
+            elif len(self.explained_variance_) < n_sf_min:
+                # TODO Rename variances_ to var_ to align with sklearn/sklearnex IncrementalPCA
+                if hasattr(self, "variances_"):
+                    resid_var = self.variances_.sum()
+                elif hasattr(self, "var_"):
+                    resid_var = self.var_.sum()
+
+                resid_var -= self.explained_variance_.sum()
+                return resid_var / (n_sf_min - n_components)
+        else:
+            return 0.0
+
+    def _create_model(self):
+        m = self._get_backend("decomposition", "dim_reduction", "model")
+        m.eigenvectors = to_table(self.components_)
+        m.means = to_table(self.mean_)
+        if self.whiten:
+            m.eigenvalues = to_table(self.explained_variance_)
+        self._onedal_model = m
+        return m
+
+    def predict(self, X, queue=None):
+        policy = self._get_policy(queue, X)
+        model = self._create_model()
+        X = _convert_to_supported(policy, X)
+        params = self._get_onedal_params(X, stage="predict")
+
+        result = self._get_backend(
+            "decomposition", "dim_reduction", "infer", policy, params, model, to_table(X)
+        )
+        return from_table(result.transformed_data)
+
+
+class PCA(BasePCA):
+
+    def fit(self, X, y=None, queue=None):
+        n_samples, n_features = X.shape
+        n_sf_min = min(n_samples, n_features)
+        self._validate_n_components(self.n_components, n_samples, n_features)
+
+        policy = self._get_policy(queue, X)
+        # TODO: investigate why np.ndarray with OWNDATA=FALSE flag
+        # fails to be converted to oneDAL table
+        if isinstance(X, np.ndarray) and not X.flags["OWNDATA"]:
+            X = X.copy()
+        X = _convert_to_supported(policy, X)
+
+        params = self._get_onedal_params(X)
+        result = self._get_backend(
+            "decomposition", "dim_reduction", "train", policy, params, to_table(X)
+        )
+
+        self.mean_ = from_table(result.means).ravel()
+        self.variances_ = from_table(result.variances)
+        self.components_ = from_table(result.eigenvectors)
+        self.singular_values_ = from_table(result.singular_values).ravel()
+        self.explained_variance_ = np.maximum(from_table(result.eigenvalues).ravel(), 0)
+        self.explained_variance_ratio_ = from_table(
+            result.explained_variances_ratio
+        ).ravel()
+        self.n_samples_ = n_samples
+        self.n_features_ = n_features
+
+        U = None
+        S = self.singular_values_
+        Vt = self.components_
+
+        n_components = self._resolve_n_components_for_result(X.shape)
+        self.n_components_ = n_components
+        self.noise_variance_ = self._compute_noise_variance(n_components, n_sf_min)
+
+        if n_components < params["n_components"]:
+            self.explained_variance_ = self.explained_variance_[:n_components]
+            self.components_ = self.components_[:n_components]
+            self.singular_values_ = self.singular_values_[:n_components]
+            self.explained_variance_ratio_ = self.explained_variance_ratio_[:n_components]
+
+        return self

scikit_learn_intelex-2025.1.0.data/data/Lib/site-packages/onedal/decomposition/tests/test_incremental_pca.py

@@ -0,0 +1,198 @@
+# ==============================================================================
+# 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 numpy as np
+import pytest
+from numpy.testing import assert_allclose
+
+from daal4py.sklearn._utils import daal_check_version
+from onedal.decomposition import IncrementalPCA
+from onedal.tests.utils._device_selection import get_queues
+
+
+@pytest.mark.parametrize("queue", get_queues())
+@pytest.mark.parametrize("is_deterministic", [True, False])
+@pytest.mark.parametrize("whiten", [True, False])
+@pytest.mark.parametrize("num_blocks", [1, 2, 3])
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+def test_on_gold_data(queue, is_deterministic, whiten, num_blocks, dtype):
+    X = np.array([[-1, -1], [-2, -1], [-3, -2], [1, 1], [2, 1], [3, 2]])
+    X = X.astype(dtype=dtype)
+    X_split = np.array_split(X, num_blocks)
+    incpca = IncrementalPCA(is_deterministic=is_deterministic, whiten=whiten)
+
+    for i in range(num_blocks):
+        incpca.partial_fit(X_split[i], queue=queue)
+
+    result = incpca.finalize_fit()
+
+    transformed_data = incpca.predict(X, queue=queue)
+
+    expected_n_components_ = 2
+    expected_components_ = np.array([[0.83849224, 0.54491354], [-0.54491354, 0.83849224]])
+    expected_singular_values_ = np.array([6.30061232, 0.54980396])
+    expected_mean_ = np.array([0, 0])
+    expected_var_ = np.array([5.6, 2.4])
+    expected_explained_variance_ = np.array([7.93954312, 0.06045688])
+    expected_explained_variance_ratio_ = np.array([0.99244289, 0.00755711])
+    expected_transformed_data = (
+        np.array(
+            [
+                [-0.49096647, -1.19399271],
+                [-0.78854479, 1.02218579],
+                [-1.27951125, -0.17180692],
+                [0.49096647, 1.19399271],
+                [0.78854479, -1.02218579],
+                [1.27951125, 0.17180692],
+            ]
+        )
+        if whiten
+        else np.array(
+            [
+                [-1.38340578, -0.2935787],
+                [-2.22189802, 0.25133484],
+                [-3.6053038, -0.04224385],
+                [1.38340578, 0.2935787],
+                [2.22189802, -0.25133484],
+                [3.6053038, 0.04224385],
+            ]
+        )
+    )
+
+    tol = 1e-7
+    if transformed_data.dtype == np.float32:
+        tol = 7e-6 if whiten else 1e-6
+
+    assert result.n_components_ == expected_n_components_
+
+    assert_allclose(result.singular_values_, expected_singular_values_, atol=tol)
+    assert_allclose(result.mean_, expected_mean_, atol=tol)
+    assert_allclose(result.var_, expected_var_, atol=tol)
+    assert_allclose(result.explained_variance_, expected_explained_variance_, atol=tol)
+    assert_allclose(
+        result.explained_variance_ratio_, expected_explained_variance_ratio_, atol=tol
+    )
+    if is_deterministic and daal_check_version((2024, "P", 500)):
+        assert_allclose(result.components_, expected_components_, atol=tol)
+        assert_allclose(transformed_data, expected_transformed_data, atol=tol)
+    else:
+        for i in range(result.n_components_):
+            abs_dot_product = np.abs(
+                np.dot(result.components_[i], expected_components_[i])
+            )
+            assert np.abs(abs_dot_product - 1.0) < tol
+
+            if np.dot(result.components_[i], expected_components_[i]) < 0:
+                assert_allclose(
+                    -transformed_data[i], expected_transformed_data[i], atol=tol
+                )
+            else:
+                assert_allclose(
+                    transformed_data[i], expected_transformed_data[i], atol=tol
+                )
+
+
+@pytest.mark.parametrize("queue", get_queues())
+@pytest.mark.parametrize("n_components", [None, 1, 5])
+@pytest.mark.parametrize("whiten", [True, False])
+@pytest.mark.parametrize("num_blocks", [1, 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_on_random_data(
+    queue, n_components, whiten, num_blocks, row_count, column_count, dtype
+):
+    seed = 78
+    gen = np.random.default_rng(seed)
+    X = gen.uniform(low=-0.3, high=+0.7, size=(row_count, column_count))
+    X = X.astype(dtype=dtype)
+    X_split = np.array_split(X, num_blocks)
+    incpca = IncrementalPCA(n_components=n_components, whiten=whiten)
+
+    for i in range(num_blocks):
+        incpca.partial_fit(X_split[i], queue=queue)
+
+    incpca.finalize_fit()
+
+    transformed_data = incpca.predict(X, queue=queue)
+    tol = 3e-3 if transformed_data.dtype == np.float32 else 2e-6
+
+    n_components = incpca.n_components_
+    expected_n_samples_seen = X.shape[0]
+    expected_n_features_in = X.shape[1]
+    n_samples_seen = incpca.n_samples_seen_
+    n_features_in = incpca.n_features_in_
+    assert n_samples_seen == expected_n_samples_seen
+    assert n_features_in == expected_n_features_in
+
+    components = incpca.components_
+    singular_values = incpca.singular_values_
+    centered_data = X - np.mean(X, axis=0)
+    cov_eigenvalues, cov_eigenvectors = np.linalg.eig(
+        centered_data.T @ centered_data / (n_samples_seen - 1)
+    )
+    cov_eigenvalues = np.nan_to_num(cov_eigenvalues)
+    cov_eigenvalues[cov_eigenvalues < 0] = 0
+    eigenvalues_order = np.argsort(cov_eigenvalues)[::-1]
+    sorted_eigenvalues = cov_eigenvalues[eigenvalues_order]
+    sorted_eigenvectors = cov_eigenvectors[:, eigenvalues_order]
+    expected_singular_values = np.sqrt(sorted_eigenvalues * (n_samples_seen - 1))[
+        :n_components
+    ]
+    expected_components = sorted_eigenvectors.T[:n_components]
+
+    assert_allclose(singular_values, expected_singular_values, atol=tol)
+    for i in range(n_components):
+        component_length = np.dot(components[i], components[i])
+        assert np.abs(component_length - 1.0) < tol
+        abs_dot_product = np.abs(np.dot(components[i], expected_components[i]))
+        assert np.abs(abs_dot_product - 1.0) < tol
+
+    expected_mean = np.mean(X, axis=0)
+    assert_allclose(incpca.mean_, expected_mean, atol=tol)
+
+    expected_var_ = np.var(X, ddof=1, axis=0)
+    assert_allclose(incpca.var_, expected_var_, atol=tol)
+
+    expected_explained_variance = sorted_eigenvalues[:n_components]
+    assert_allclose(incpca.explained_variance_, expected_explained_variance, atol=tol)
+
+    expected_explained_variance_ratio = expected_explained_variance / np.sum(
+        sorted_eigenvalues
+    )
+    assert_allclose(
+        incpca.explained_variance_ratio_, expected_explained_variance_ratio, atol=tol
+    )
+
+    expected_noise_variance = (
+        np.mean(sorted_eigenvalues[n_components:])
+        if len(sorted_eigenvalues) > n_components
+        else 0.0
+    )
+    # TODO Fix noise variance computation (It is necessary to update C++ side)
+    # assert np.abs(incpca.noise_variance_ - expected_noise_variance) < tol
+
+    expected_transformed_data = centered_data @ components.T
+    if whiten:
+        scale = np.sqrt(incpca.explained_variance_)
+        min_scale = np.finfo(scale.dtype).eps
+        scale[scale < min_scale] = np.inf
+        expected_transformed_data /= scale
+
+    if daal_check_version((2024, "P", 500)) or not (
+        whiten and queue is not None and queue.sycl_device.device_type.name == "gpu"
+    ):
+        assert_allclose(transformed_data, expected_transformed_data, atol=tol)

scikit_learn_intelex-2025.1.0.data/data/Lib/site-packages/onedal/ensemble/__init__.py

@@ -0,0 +1,29 @@
+# ==============================================================================
+# Copyright 2023 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 .forest import (
+    ExtraTreesClassifier,
+    ExtraTreesRegressor,
+    RandomForestClassifier,
+    RandomForestRegressor,
+)
+
+__all__ = [
+    "RandomForestClassifier",
+    "RandomForestRegressor",
+    "ExtraTreesClassifier",
+    "ExtraTreesRegressor",
+]