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

.. only:: html

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

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

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

.. _sphx_glr_tutorials_getting_started.py:


Getting Started with TabICL
============================

This example demonstrates the basic usage of TabICL for classification and
regression tasks using the scikit-learn compatible API.

.. GENERATED FROM PYTHON SOURCE LINES 10-15

Classification with TabICLClassifier
-------------------------------------

TabICL provides a scikit-learn compatible classifier that works out of the
box. Let's use it on a synthetic classification dataset.

.. GENERATED FROM PYTHON SOURCE LINES 15-29

.. code-block:: Python


    from sklearn.datasets import make_classification
    from sklearn.model_selection import cross_val_score

    from tabicl import TabICLClassifier

    X, y = make_classification(
        n_samples=300, n_features=10, n_informative=5, random_state=42
    )

    clf = TabICLClassifier(n_estimators=4, device="cpu")
    scores = cross_val_score(clf, X, y, cv=5, scoring="accuracy")
    print(f"Classification accuracy: {scores.mean():.3f} (+/- {scores.std():.3f})")





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

 .. code-block:: none

    Classification accuracy: 0.880 (+/- 0.019)




.. GENERATED FROM PYTHON SOURCE LINES 30-34

Regression with TabICLRegressor
--------------------------------

TabICLRegressor follows the same interface for regression tasks.

.. GENERATED FROM PYTHON SOURCE LINES 34-48

.. code-block:: Python


    from sklearn.datasets import make_regression
    from sklearn.model_selection import cross_val_score

    from tabicl import TabICLRegressor

    X, y = make_regression(
        n_samples=300, n_features=10, n_informative=5, noise=0.5, random_state=42
    )

    reg = TabICLRegressor(n_estimators=4, device="cpu")
    scores = cross_val_score(reg, X, y, cv=5, scoring="r2")
    print(f"Regression R² score: {scores.mean():.3f} (+/- {scores.std():.3f})")





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

 .. code-block:: none

    Regression R² score: 1.000 (+/- 0.000)




.. GENERATED FROM PYTHON SOURCE LINES 49-55

Using KV caching for faster repeated inference
------------------------------------------------

When you need to call ``predict`` multiple times on the same training data
(e.g., during evaluation), enable KV caching to speed up inference. The
cache is built once during ``fit`` and reused across ``predict`` calls.

.. GENERATED FROM PYTHON SOURCE LINES 55-74

.. code-block:: Python


    from sklearn.model_selection import train_test_split

    X, y = make_classification(
        n_samples=300, n_features=10, n_informative=5, random_state=42
    )
    X_train, X_test, y_train, y_test = train_test_split(
        X, y, test_size=0.2, random_state=42
    )

    clf = TabICLClassifier(n_estimators=4, kv_cache=True, device="cpu")
    clf.fit(X_train, y_train)

    # Subsequent predict calls reuse the cached context
    predictions = clf.predict(X_test)
    probabilities = clf.predict_proba(X_test)
    print(f"Predictions shape: {predictions.shape}")
    print(f"Probabilities shape: {probabilities.shape}")





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

 .. code-block:: none

    Predictions shape: (60,)
    Probabilities shape: (60, 2)




.. GENERATED FROM PYTHON SOURCE LINES 75-79

Saving the model
----------------

Save and load a fitted classifier or regressor:

.. GENERATED FROM PYTHON SOURCE LINES 79-88

.. code-block:: Python


    clf.save(
       "classifier.pkl",
       save_model_weights=False,  # if False, reload from checkpoint on load
       save_training_data=True,   # if True, include training data; if False, discard it (requires KV cache)
       save_kv_cache=True,        # if True and KV cache exists, save it
    )
    clf = TabICLClassifier.load("classifier.pkl")








.. GENERATED FROM PYTHON SOURCE LINES 89-92

When KV cache exists and is saved, you can set
``save_training_data=False`` to exclude cached training data, which may
be useful for data privacy.


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

   **Total running time of the script:** (0 minutes 11.455 seconds)


.. _sphx_glr_download_tutorials_getting_started.py:

.. only:: html

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

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

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

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

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

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

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


.. only:: html

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

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