.. DO NOT EDIT.
.. THIS FILE WAS AUTOMATICALLY GENERATED BY SPHINX-GALLERY.
.. TO MAKE CHANGES, EDIT THE SOURCE PYTHON FILE:
.. "examples/Builtin/spatial_perturbation_logistic_regression.py"
.. LINE NUMBERS ARE GIVEN BELOW.

.. only:: html

    .. note::
        :class: sphx-glr-download-link-note

        :ref:`Go to the end <sphx_glr_download_examples_Builtin_spatial_perturbation_logistic_regression.py>`
        to download the full example code.

.. rst-class:: sphx-glr-example-title

.. _sphx_glr_examples_Builtin_spatial_perturbation_logistic_regression.py:


Spatial perturbations and logistic regression
=============================================

This example illustrates the use of the :class:`skwdro.linear_models.LogisticRegression` class on datasets that are shifted at test time.

.. GENERATED FROM PYTHON SOURCE LINES 8-19

.. code-block:: Python

    import numpy as np
    import matplotlib.pyplot as plt

    from sklearn.datasets import make_blobs, make_moons
    from sklearn.metrics import accuracy_score
    from sklearn.model_selection import train_test_split

    from skwdro.linear_models import LogisticRegression

    from utils.classifier_comparison import plot_classifier_comparison








.. GENERATED FROM PYTHON SOURCE LINES 20-22

Setup
~~~~~

.. GENERATED FROM PYTHON SOURCE LINES 22-40

.. code-block:: Python


    n = 500 # Total number of samples
    n_train = (3 * n) // 4 # Number of training samples
    n_test = n - n_train # Number of test samples

    sdevs = [(2.5, 5), (1, 5)]

    # Fix centers for blobs dataset
    pos = 4
    centers = [np.array([-pos,-pos]), np.array([pos,pos])]

    # Create datasets with variance that is shifted at test time
    datasets = []
    for (sdev_1, sdev_2) in sdevs:
        train_dataset = make_blobs(n_samples=n_train, centers=centers, cluster_std=(sdev_1, sdev_2)) # type: ignore
        test_dataset = make_blobs(n_samples=n_test, centers=centers, cluster_std=(sdev_2, sdev_1)) # type: ignore
        datasets.append((train_dataset, test_dataset))








.. GENERATED FROM PYTHON SOURCE LINES 41-43

WDRO classifiers
~~~~~~~~~~~~~~~~

.. GENERATED FROM PYTHON SOURCE LINES 43-62

.. code-block:: Python


    # Rho chosen analytically
    rhos = [0, 2*4**2]

    # Kappa: weight of label shift
    kappa = 1000

    # Cost:
    # t: torch backend
    # NLC: norm cost that takes labels into account
    # 2 2 : squared 2-norm
    # kappa: weight of label shift
    cost = f"t-NLC-2-2-{kappa}"

    # WDRO classifier
    classifiers = []
    for rho in rhos:
        classifiers.append(LogisticRegression(rho=rho, cost=cost))








.. GENERATED FROM PYTHON SOURCE LINES 63-65

Make plot
~~~~~~~~~

.. GENERATED FROM PYTHON SOURCE LINES 65-69

.. code-block:: Python


    names = ["Logistic Regression", "WDRO Logistic Regression"]
    levels = [0., 0.25, 0.45, 0.5, 0.55, 0.75, 1.]
    plot_classifier_comparison(names, classifiers, datasets, levels=levels) # type: ignore



.. image-sg:: /examples/Builtin/images/sphx_glr_spatial_perturbation_logistic_regression_001.png
   :alt: Training data, Testing data, Logistic Regression, WDRO Logistic Regression
   :srcset: /examples/Builtin/images/sphx_glr_spatial_perturbation_logistic_regression_001.png
   :class: sphx-glr-single-img


.. rst-class:: sphx-glr-script-out

 .. code-block:: none

    Dataset 0, Classifier Logistic Regression: Test Acc. 95%, F1 95%
    Dataset 0, Classifier WDRO Logistic Regression: Test Acc. 85%, F1 83%
    Dataset 1, Classifier Logistic Regression: Test Acc. 81%, F1 84%
    Dataset 1, Classifier WDRO Logistic Regression: Test Acc. 90%, F1 91%





.. rst-class:: sphx-glr-timing

   **Total running time of the script:** (0 minutes 33.304 seconds)


.. _sphx_glr_download_examples_Builtin_spatial_perturbation_logistic_regression.py:

.. only:: html

  .. container:: sphx-glr-footer sphx-glr-footer-example

    .. container:: sphx-glr-download sphx-glr-download-jupyter

      :download:`Download Jupyter notebook: spatial_perturbation_logistic_regression.ipynb <spatial_perturbation_logistic_regression.ipynb>`

    .. container:: sphx-glr-download sphx-glr-download-python

      :download:`Download Python source code: spatial_perturbation_logistic_regression.py <spatial_perturbation_logistic_regression.py>`

    .. container:: sphx-glr-download sphx-glr-download-zip

      :download:`Download zipped: spatial_perturbation_logistic_regression.zip <spatial_perturbation_logistic_regression.zip>`


.. only:: html

 .. rst-class:: sphx-glr-signature

    `Gallery generated by Sphinx-Gallery <https://sphinx-gallery.github.io>`_