scikit-learn-intelex 2025.0.0__py311-none-manylinux_2_28_x86_64.whl

sklearnex/basic_statistics/incremental_basic_statistics.py

@@ -0,0 +1,288 @@
+# ==============================================================================
+# 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 sklearn.base import BaseEstimator
+from sklearn.utils import check_array, gen_batches
+from sklearn.utils.validation import _check_sample_weight
+
+from daal4py.sklearn._n_jobs_support import control_n_jobs
+from daal4py.sklearn._utils import sklearn_check_version
+from onedal.basic_statistics import (
+    IncrementalBasicStatistics as onedal_IncrementalBasicStatistics,
+)
+
+from .._device_offload import dispatch
+from .._utils import PatchingConditionsChain
+
+if sklearn_check_version("1.2"):
+    from sklearn.utils._param_validation import Interval, StrOptions
+
+import numbers
+
+
+@control_n_jobs(decorated_methods=["partial_fit", "_onedal_finalize_fit"])
+class IncrementalBasicStatistics(BaseEstimator):
+    """
+    Incremental estimator for basic statistics.
+    Allows to compute basic statistics if data are splitted into batches.
+    Parameters
+    ----------
+    result_options: string or list, default='all'
+        List of statistics to compute
+
+    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.
+
+    Attributes (are existing only if corresponding result option exists)
+    ----------
+        min : ndarray of shape (n_features,)
+            Minimum of each feature over all samples.
+
+        max : ndarray of shape (n_features,)
+            Maximum of each feature over all samples.
+
+        sum : ndarray of shape (n_features,)
+            Sum of each feature over all samples.
+
+        mean : ndarray of shape (n_features,)
+            Mean of each feature over all samples.
+
+        variance : ndarray of shape (n_features,)
+            Variance of each feature over all samples.
+
+        variation : ndarray of shape (n_features,)
+            Variation of each feature over all samples.
+
+        sum_squares : ndarray of shape (n_features,)
+            Sum of squares for each feature over all samples.
+
+        standard_deviation : ndarray of shape (n_features,)
+            Standard deviation of each feature over all samples.
+
+        sum_squares_centered : ndarray of shape (n_features,)
+            Centered sum of squares for each feature over all samples.
+
+        second_order_raw_moment : ndarray of shape (n_features,)
+            Second order moment of each feature over all samples.
+    """
+
+    _onedal_incremental_basic_statistics = staticmethod(onedal_IncrementalBasicStatistics)
+
+    if sklearn_check_version("1.2"):
+        _parameter_constraints: dict = {
+            "result_options": [
+                StrOptions(
+                    {
+                        "all",
+                        "min",
+                        "max",
+                        "sum",
+                        "mean",
+                        "variance",
+                        "variation",
+                        "sum_squares",
+                        "standard_deviation",
+                        "sum_squares_centered",
+                        "second_order_raw_moment",
+                    }
+                ),
+                list,
+            ],
+            "batch_size": [Interval(numbers.Integral, 1, None, closed="left"), None],
+        }
+
+    def __init__(self, result_options="all", batch_size=None):
+        if result_options == "all":
+            self.result_options = (
+                self._onedal_incremental_basic_statistics.get_all_result_options()
+            )
+        else:
+            self.result_options = result_options
+        self._need_to_finalize = False
+        self.batch_size = batch_size
+
+    def _onedal_supported(self, method_name, *data):
+        patching_status = PatchingConditionsChain(
+            f"sklearn.basic_statistics.{self.__class__.__name__}.{method_name}"
+        )
+        return patching_status
+
+    _onedal_cpu_supported = _onedal_supported
+    _onedal_gpu_supported = _onedal_supported
+
+    def _get_onedal_result_options(self, options):
+        if isinstance(options, list):
+            onedal_options = "|".join(self.result_options)
+        else:
+            onedal_options = options
+        assert isinstance(onedal_options, str)
+        return options
+
+    def _onedal_finalize_fit(self, queue=None):
+        assert hasattr(self, "_onedal_estimator")
+        self._onedal_estimator.finalize_fit(queue=queue)
+        self._need_to_finalize = False
+
+    def _onedal_partial_fit(self, X, sample_weight=None, queue=None):
+        first_pass = not hasattr(self, "n_samples_seen_") or self.n_samples_seen_ == 0
+
+        if sklearn_check_version("1.0"):
+            X = self._validate_data(
+                X,
+                dtype=[np.float64, np.float32],
+                reset=first_pass,
+            )
+        else:
+            X = check_array(
+                X,
+                dtype=[np.float64, np.float32],
+            )
+
+        if sample_weight is not None:
+            sample_weight = _check_sample_weight(sample_weight, X)
+
+        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]
+
+        onedal_params = {
+            "result_options": self._get_onedal_result_options(self.result_options)
+        }
+        if not hasattr(self, "_onedal_estimator"):
+            self._onedal_estimator = self._onedal_incremental_basic_statistics(
+                **onedal_params
+            )
+        self._onedal_estimator.partial_fit(X, weights=sample_weight, queue=queue)
+        self._need_to_finalize = True
+
+    def _onedal_fit(self, X, sample_weight=None, queue=None):
+        if sklearn_check_version("1.0"):
+            X = self._validate_data(X, dtype=[np.float64, np.float32])
+        else:
+            X = check_array(X, dtype=[np.float64, np.float32])
+
+        if sample_weight is not None:
+            sample_weight = _check_sample_weight(sample_weight, X)
+
+        n_samples, n_features = X.shape
+        if self.batch_size is None:
+            self.batch_size_ = 5 * n_features
+        else:
+            self.batch_size_ = self.batch_size
+
+        self.n_samples_seen_ = 0
+        if hasattr(self, "_onedal_estimator"):
+            self._onedal_estimator._reset()
+
+        for batch in gen_batches(X.shape[0], self.batch_size_):
+            X_batch = X[batch]
+            weights_batch = sample_weight[batch] if sample_weight is not None else None
+            self._onedal_partial_fit(X_batch, weights_batch, queue=queue)
+
+        if sklearn_check_version("1.2"):
+            self._validate_params()
+
+        self.n_features_in_ = X.shape[1]
+
+        self._onedal_finalize_fit(queue=queue)
+
+        return self
+
+    def __getattr__(self, attr):
+        result_options = self.__dict__["result_options"]
+        is_statistic_attr = (
+            isinstance(result_options, str) and (attr == result_options)
+        ) or (isinstance(result_options, list) and (attr in result_options))
+        if is_statistic_attr:
+            if self._need_to_finalize:
+                self._onedal_finalize_fit()
+            return getattr(self._onedal_estimator, attr)
+        if attr in self.__dict__:
+            return self.__dict__[attr]
+
+        raise AttributeError(
+            f"'{self.__class__.__name__}' object has no attribute '{attr}'"
+        )
+
+    def partial_fit(self, X, sample_weight=None):
+        """Incremental fit with X. All of X is processed as a single batch.
+
+        Parameters
+        ----------
+        X : array-like of shape (n_samples, n_features)
+            Data for compute, 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.
+
+        sample_weight : array-like of shape (n_samples,), default=None
+            Weights for compute weighted statistics, where `n_samples` is the number of samples.
+
+        Returns
+        -------
+        self : object
+            Returns the instance itself.
+        """
+        dispatch(
+            self,
+            "partial_fit",
+            {
+                "onedal": self.__class__._onedal_partial_fit,
+                "sklearn": None,
+            },
+            X,
+            sample_weight,
+        )
+        return self
+
+    def fit(self, X, y=None, sample_weight=None):
+        """Compute statistics with X, using minibatches of size batch_size.
+
+        Parameters
+        ----------
+        X : array-like of shape (n_samples, n_features)
+            Data for compute, 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.
+
+        sample_weight : array-like of shape (n_samples,), default=None
+            Weights for compute weighted statistics, where `n_samples` is the number of samples.
+
+        Returns
+        -------
+        self : object
+            Returns the instance itself.
+        """
+        dispatch(
+            self,
+            "fit",
+            {
+                "onedal": self.__class__._onedal_fit,
+                "sklearn": None,
+            },
+            X,
+            sample_weight,
+        )
+        return self

sklearnex/basic_statistics/tests/test_basic_statistics.py

@@ -0,0 +1,251 @@
+# ==============================================================================
+# 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 numpy as np
+import pytest
+from numpy.testing import assert_allclose
+
+from onedal.basic_statistics.tests.test_basic_statistics import (
+    expected_max,
+    expected_mean,
+    expected_sum,
+    options_and_tests,
+)
+from onedal.tests.utils._dataframes_support import (
+    _convert_to_dataframe,
+    get_dataframes_and_queues,
+)
+from sklearnex.basic_statistics import BasicStatistics
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+def test_sklearnex_import_basic_statistics(dataframe, queue):
+    X = np.array([[0, 0], [1, 1]])
+    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+
+    weights = np.array([1, 0.5])
+    weights_df = _convert_to_dataframe(weights, sycl_queue=queue, target_df=dataframe)
+
+    result = BasicStatistics().fit(X_df)
+
+    expected_mean = np.array([0.5, 0.5])
+    expected_min = np.array([0, 0])
+    expected_max = np.array([1, 1])
+
+    assert_allclose(expected_mean, result.mean)
+    assert_allclose(expected_max, result.max)
+    assert_allclose(expected_min, result.min)
+
+    result = BasicStatistics().fit(X_df, sample_weight=weights_df)
+
+    expected_weighted_mean = np.array([0.25, 0.25])
+    expected_weighted_min = np.array([0, 0])
+    expected_weighted_max = np.array([0.5, 0.5])
+
+    assert_allclose(expected_weighted_mean, result.mean)
+    assert_allclose(expected_weighted_min, result.min)
+    assert_allclose(expected_weighted_max, result.max)
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+@pytest.mark.parametrize("weighted", [True, False])
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+def test_multiple_options_on_gold_data(dataframe, queue, weighted, dtype):
+    X = np.array([[0, 0], [1, 1]])
+    X = X.astype(dtype=dtype)
+    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+    if weighted:
+        weights = np.array([1, 0.5])
+        weights = weights.astype(dtype=dtype)
+        weights_df = _convert_to_dataframe(weights, sycl_queue=queue, target_df=dataframe)
+    basicstat = BasicStatistics()
+
+    if weighted:
+        result = basicstat.fit(X_df, sample_weight=weights_df)
+    else:
+        result = basicstat.fit(X_df)
+
+    if weighted:
+        expected_weighted_mean = np.array([0.25, 0.25])
+        expected_weighted_min = np.array([0, 0])
+        expected_weighted_max = np.array([0.5, 0.5])
+        assert_allclose(expected_weighted_mean, result.mean)
+        assert_allclose(expected_weighted_max, result.max)
+        assert_allclose(expected_weighted_min, result.min)
+    else:
+        expected_mean = np.array([0.5, 0.5])
+        expected_min = np.array([0, 0])
+        expected_max = np.array([1, 1])
+        assert_allclose(expected_mean, result.mean)
+        assert_allclose(expected_max, result.max)
+        assert_allclose(expected_min, result.min)
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+@pytest.mark.parametrize("option", options_and_tests)
+@pytest.mark.parametrize("row_count", [100, 1000])
+@pytest.mark.parametrize("column_count", [10, 100])
+@pytest.mark.parametrize("weighted", [True, False])
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+def test_single_option_on_random_data(
+    dataframe, queue, option, row_count, column_count, weighted, dtype
+):
+    result_option, function, tols = option
+    fp32tol, fp64tol = tols
+    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=dtype)
+    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+    if weighted:
+        weights = gen.uniform(low=-0.5, high=1.0, size=row_count)
+        weights = weights.astype(dtype=dtype)
+        weights_df = _convert_to_dataframe(weights, sycl_queue=queue, target_df=dataframe)
+    basicstat = BasicStatistics(result_options=result_option)
+
+    if weighted:
+        result = basicstat.fit(X_df, sample_weight=weights_df)
+    else:
+        result = basicstat.fit(X_df)
+
+    res = getattr(result, result_option)
+    if weighted:
+        weighted_data = np.diag(weights) @ X
+        gtr = function(weighted_data)
+    else:
+        gtr = function(X)
+
+    tol = fp32tol if res.dtype == np.float32 else fp64tol
+    assert_allclose(gtr, res, atol=tol)
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+@pytest.mark.parametrize("row_count", [100, 1000])
+@pytest.mark.parametrize("column_count", [10, 100])
+@pytest.mark.parametrize("weighted", [True, False])
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+def test_multiple_options_on_random_data(
+    dataframe, queue, row_count, column_count, weighted, dtype
+):
+    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=dtype)
+    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+    if weighted:
+        weights = gen.uniform(low=-0.5, high=1.0, size=row_count)
+        weights = weights.astype(dtype=dtype)
+        weights_df = _convert_to_dataframe(weights, sycl_queue=queue, target_df=dataframe)
+    basicstat = BasicStatistics(result_options=["mean", "max", "sum"])
+
+    if weighted:
+        result = basicstat.fit(X_df, sample_weight=weights_df)
+    else:
+        result = basicstat.fit(X_df)
+
+    res_mean, res_max, res_sum = result.mean, result.max, result.sum
+    if weighted:
+        weighted_data = np.diag(weights) @ X
+        gtr_mean, gtr_max, gtr_sum = (
+            expected_mean(weighted_data),
+            expected_max(weighted_data),
+            expected_sum(weighted_data),
+        )
+    else:
+        gtr_mean, gtr_max, gtr_sum = (
+            expected_mean(X),
+            expected_max(X),
+            expected_sum(X),
+        )
+
+    tol = 5e-4 if res_mean.dtype == np.float32 else 1e-7
+    assert_allclose(gtr_mean, res_mean, atol=tol)
+    assert_allclose(gtr_max, res_max, atol=tol)
+    assert_allclose(gtr_sum, res_sum, atol=tol)
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+@pytest.mark.parametrize("row_count", [100, 1000])
+@pytest.mark.parametrize("column_count", [10, 100])
+@pytest.mark.parametrize("weighted", [True, False])
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+def test_all_option_on_random_data(
+    dataframe, queue, row_count, column_count, weighted, dtype
+):
+    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=dtype)
+    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+    if weighted:
+        weights = gen.uniform(low=-0.5, high=+1.0, size=row_count)
+        weights = weights.astype(dtype=dtype)
+        weights_df = _convert_to_dataframe(weights, sycl_queue=queue, target_df=dataframe)
+    basicstat = BasicStatistics(result_options="all")
+
+    if weighted:
+        result = basicstat.fit(X_df, sample_weight=weights_df)
+    else:
+        result = basicstat.fit(X_df)
+
+    if weighted:
+        weighted_data = np.diag(weights) @ X
+
+    for option in options_and_tests:
+        result_option, function, tols = option
+        fp32tol, fp64tol = tols
+        res = getattr(result, result_option)
+        if weighted:
+            gtr = function(weighted_data)
+        else:
+            gtr = function(X)
+        tol = fp32tol if res.dtype == np.float32 else fp64tol
+        assert_allclose(gtr, res, atol=tol)
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+@pytest.mark.parametrize("option", options_and_tests)
+@pytest.mark.parametrize("data_size", [100, 1000])
+@pytest.mark.parametrize("weighted", [True, False])
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+def test_1d_input_on_random_data(dataframe, queue, option, data_size, weighted, dtype):
+    result_option, function, tols = option
+    fp32tol, fp64tol = tols
+    seed = 77
+    gen = np.random.default_rng(seed)
+    X = gen.uniform(low=-0.3, high=+0.7, size=data_size)
+    X = X.astype(dtype=dtype)
+    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+    if weighted:
+        weights = gen.uniform(low=-0.5, high=1.0, size=data_size)
+        weights = weights.astype(dtype=dtype)
+        weights_df = _convert_to_dataframe(weights, sycl_queue=queue, target_df=dataframe)
+    basicstat = BasicStatistics(result_options=result_option)
+
+    if weighted:
+        result = basicstat.fit(X_df, sample_weight=weights_df)
+    else:
+        result = basicstat.fit(X_df)
+
+    res = getattr(result, result_option)
+    if weighted:
+        weighted_data = weights * X
+        gtr = function(weighted_data)
+    else:
+        gtr = function(X)
+
+    tol = fp32tol if res.dtype == np.float32 else fp64tol
+    assert_allclose(gtr, res, atol=tol)