:orphan:

Example gallery
================

We provide here different examples on how to use the GemClus library, from clustering to variable selection.


.. raw:: html

    <div class="sphx-glr-thumbnails">

.. thumbnail-parent-div-open

.. thumbnail-parent-div-close

.. raw:: html

    </div>

General examples
----------------


.. raw:: html

    <div class="sphx-glr-thumbnails">

.. thumbnail-parent-div-open

.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="We show in this example how to use the gemclus.mlp.MLPMMD to fit a simple mixture of Gaussian distributions. The architecture here is a two-layer neural network trained with the MMD GEMINI, an objective equivalent to kernel KMeans.">

.. only:: html

  .. image:: /auto_examples/_general/images/thumb/sphx_glr_plot_intro_mlp_mmd_thumb.png
    :alt:

  :ref:`sphx_glr_auto_examples__general_plot_intro_mlp_mmd.py`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">An introducing example to clustering with an MLP and the MMD GEMINI</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="This example illustrates how we can run nonparametric clustering using GEMINI. The specificity of this model is that the decision of model is not dependent on the position of the inputs, but only on the parameters associated to the input. Consequently, this model cannot be used for unseen samples as it will always return the same predictions.">

.. only:: html

  .. image:: /auto_examples/_general/images/thumb/sphx_glr_plot_nonparametric_clustering_thumb.png
    :alt:

  :ref:`sphx_glr_auto_examples__general_plot_nonparametric_clustering.py`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Non parametric clustering</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="Since the MMD GEMINI objective is equivalent in OvO mode to a kernel KMeans objective, we can use it with the nonparametric model that directly associates a cluster to each sample. The overall model would thus behave as a kernel KMeans algorithm. However, its training is done by gradient descent.">

.. only:: html

  .. image:: /auto_examples/_general/images/thumb/sphx_glr_plot_kernel_kmeans_thumb.png
    :alt:

  :ref:`sphx_glr_auto_examples__general_plot_kernel_kmeans.py`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Kernel KMeans clustering with GEMINI</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="This example is a retake from the experiment in the original GEMINI paper where we want to find the true clusters in a mixture of Gaussian that incorporates a low-degree-of-freedom student t-distribution. Consequently, this distribution generates sample that may seem like outliers if we are to expect only Gaussian distribution.">

.. only:: html

  .. image:: /auto_examples/_general/images/thumb/sphx_glr_plot_gstm_thumb.png
    :alt:

  :ref:`sphx_glr_auto_examples__general_plot_gstm.py`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Example of decision boundary map for a mixture of Gaussian and low-degree Student distributions</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="We show here a simple dataset consisting in two centred circle that can be challenging for some clustering algorithms. We solve this case using the kernel RIM algorithm. This algorithm fits a logistic regression on the kernel matrix derived from the dataset.">

.. only:: html

  .. image:: /auto_examples/_general/images/thumb/sphx_glr_plot_kernel_rim_thumb.png
    :alt:

  :ref:`sphx_glr_auto_examples__general_plot_kernel_rim.py`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Clustering circles with kernel RIM</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="The squared-loss mutual information (SMI) Is a variant of mutual information proposed in [1]_">

.. only:: html

  .. image:: /auto_examples/_general/images/thumb/sphx_glr_plot_squared_loss_mi_thumb.png
    :alt:

  :ref:`sphx_glr_auto_examples__general_plot_squared_loss_mi.py`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Clustering with the squared-loss mutual information</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="This example is a retake from the experiment in the original GEMINI paper where we want to find the true clusters in between two facing moons. To do so, the trick is to use a specific distance using the &quot;precomputed&quot; option which will guide the clustering algorithm to the desired solution.">

.. only:: html

  .. image:: /auto_examples/_general/images/thumb/sphx_glr_plot_moon_clustering_thumb.png
    :alt:

  :ref:`sphx_glr_auto_examples__general_plot_moon_clustering.py`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Drawing a decision boundary between two interlacing moons</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="RIM (regularised mutual information) is a proposal of model by Krause et al. (2010) which consists in maximising for a linear model under \ell_2 penalty. In this example, we show how to do clustering of a Gaussian mixture using RIM.">

.. only:: html

  .. image:: /auto_examples/_general/images/thumb/sphx_glr_plot_rim_thumb.png
    :alt:

  :ref:`sphx_glr_auto_examples__general_plot_rim.py`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Simple logistic regression with RIM</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="We create here a random graph by following a simplified version of the Latent Position Model generative procedure. To create the graph, we sample latent positions from a Gaussian Mixture Model and create a graph with as many nodes as samples. Edges are then determined according to probability depending only on the distance between samples.">

.. only:: html

  .. image:: /auto_examples/_general/images/thumb/sphx_glr_plot_graph_node_clustering_thumb.png
    :alt:

  :ref:`sphx_glr_auto_examples__general_plot_graph_node_clustering.py`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Graph node clustering with a nonparametric model</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="We show here a simple dataset consisting in two centred circle that can be challenging for some clustering algorithms. This dataset can be challenging for GEMINI as well, unless we change the kernel adequately.">

.. only:: html

  .. image:: /auto_examples/_general/images/thumb/sphx_glr_plot_circles_thumb.png
    :alt:

  :ref:`sphx_glr_auto_examples__general_plot_circles.py`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Comparative clustering of circles dataset with kernel change</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="It is possible that the model we wish to use is not available among the choices of GemClus. To that end, it is possible to extend the framework and define custom models that can still be trained by GEMINI.">

.. only:: html

  .. image:: /auto_examples/_general/images/thumb/sphx_glr_plot_custom_model_thumb.png
    :alt:

  :ref:`sphx_glr_auto_examples__general_plot_custom_model.py`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Extending GemClus to build your own discriminative clustering model</div>
    </div>


.. thumbnail-parent-div-close

.. raw:: html

    </div>


.. toctree::
   :hidden:

   /auto_examples/_general/plot_intro_mlp_mmd
   /auto_examples/_general/plot_nonparametric_clustering
   /auto_examples/_general/plot_kernel_kmeans
   /auto_examples/_general/plot_gstm
   /auto_examples/_general/plot_kernel_rim
   /auto_examples/_general/plot_squared_loss_mi
   /auto_examples/_general/plot_moon_clustering
   /auto_examples/_general/plot_rim
   /auto_examples/_general/plot_graph_node_clustering
   /auto_examples/_general/plot_circles
   /auto_examples/_general/plot_custom_model


Feature selection
-----------------


.. raw:: html

    <div class="sphx-glr-thumbnails">

.. thumbnail-parent-div-open

.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="In this example, we ask the gemclus.sparse.SparseLinearMMD to perform a path where the regularisation penalty is progressively increased until all features but 2 are discarded. The model then keeps the best weights with the minimum number of features that maintains a GEMINI score close to 90% of the maximum GEMINI value encountered during the path.">

.. only:: html

  .. image:: /auto_examples/feature_selection/images/thumb/sphx_glr_plot_feature_selection_linear_thumb.png
    :alt:

  :ref:`sphx_glr_auto_examples_feature_selection_plot_feature_selection_linear.py`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Feature selection using the Sparse MMD OvO (Logistic regression)</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="In this example, we ask the gemclus.sparse.SparseLinearMI to perform a path where the regularisation penalty is progressively increased until all features but 2 are discarded. The model then keeps the best weights with the minimum number of features that maintains a GEMINI score close to 50% of the maximum GEMINI value encountered during the path.">

.. only:: html

  .. image:: /auto_examples/feature_selection/images/thumb/sphx_glr_plot_feature_selection_logreg_mi_thumb.png
    :alt:

  :ref:`sphx_glr_auto_examples_feature_selection_plot_feature_selection_logreg_mi.py`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Feature selection using the Sparse Linear MI (Logistic regression)</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="In this example, we ask the gemclus.sparse.SparseLinearMMD to perform a path where the regularisation penalty is progressively increased until all features but 2 are discarded. Moreover, we will produce some categorical variables that are one-hot-encoded and constrain the model to consider these features altogether using the groups option of the model.">

.. only:: html

  .. image:: /auto_examples/feature_selection/images/thumb/sphx_glr_plot_grouped_selection_thumb.png
    :alt:

  :ref:`sphx_glr_auto_examples_feature_selection_plot_grouped_selection.py`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Grouped Feature selection with a linear model</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="In this example, we ask the gemclus.sparse.SparseMLPMMD to perform a path where the regularisation penalty is progressively increased until all features but 2 are discarded. The model then keeps the best weights with the minimum number of features that maintains a GEMINI score close to 90% of the maximum GEMINI value encountered during the path.">

.. only:: html

  .. image:: /auto_examples/feature_selection/images/thumb/sphx_glr_plot_feature_selection_mlp_thumb.png
    :alt:

  :ref:`sphx_glr_auto_examples_feature_selection_plot_feature_selection_mlp.py`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Feature selection using the Sparse MMD OvA (MLP)</div>
    </div>


.. thumbnail-parent-div-close

.. raw:: html

    </div>


.. toctree::
   :hidden:

   /auto_examples/feature_selection/plot_feature_selection_linear
   /auto_examples/feature_selection/plot_feature_selection_logreg_mi
   /auto_examples/feature_selection/plot_grouped_selection
   /auto_examples/feature_selection/plot_feature_selection_mlp

Consensus clustering
---------------------


.. raw:: html

    <div class="sphx-glr-thumbnails">

.. thumbnail-parent-div-open

.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="We show in this example how to:">

.. only:: html

  .. image:: /auto_examples/mlcl/images/thumb/sphx_glr_plot_semi_supervised_consensus_thumb.png
    :alt:

  :ref:`sphx_glr_auto_examples_mlcl_plot_semi_supervised_consensus.py`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Consensus clustering with linking constraints on sample pairs</div>
    </div>


.. thumbnail-parent-div-close

.. raw:: html

    </div>


.. toctree::
   :hidden:

   /auto_examples/mlcl/plot_semi_supervised_consensus

Scoring with GEMINI
-------------------


.. raw:: html

    <div class="sphx-glr-thumbnails">

.. thumbnail-parent-div-open

.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="We show in this example how we can score the prediction of another model using GEMINI. We do not seek to perform clustering but rather to evaluate.">

.. only:: html

  .. image:: /auto_examples/scoring/images/thumb/sphx_glr_plot_gemini_scoring_thumb.png
    :alt:

  :ref:`sphx_glr_auto_examples_scoring_plot_gemini_scoring.py`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Scoring any model with GEMINI</div>
    </div>


.. thumbnail-parent-div-close

.. raw:: html

    </div>


.. toctree::
   :hidden:

   /auto_examples/scoring/plot_gemini_scoring

Trees
------


.. raw:: html

    <div class="sphx-glr-thumbnails">

.. thumbnail-parent-div-open

.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="This example shows how to use the Douglas tree for a dataset with few features.">

.. only:: html

  .. image:: /auto_examples/trees/images/thumb/sphx_glr_plot_douglas_example_thumb.png
    :alt:

  :ref:`sphx_glr_auto_examples_trees_plot_douglas_example.py`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Building a differentiable unsupervised tree: DOUGLAS</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="We show here how to obtain two different decision trees for clustering using two different kernels to accompanny the KAURI method.">

.. only:: html

  .. image:: /auto_examples/trees/images/thumb/sphx_glr_plot_kauri_example_thumb.png
    :alt:

  :ref:`sphx_glr_auto_examples_trees_plot_kauri_example.py`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Building an unsupervised tree with kernel-kmeans objective: KAURI</div>
    </div>


.. thumbnail-parent-div-close

.. raw:: html

    </div>


.. toctree::
   :hidden:

   /auto_examples/trees/plot_douglas_example
   /auto_examples/trees/plot_kauri_example


.. only:: html

  .. container:: sphx-glr-footer sphx-glr-footer-gallery

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

      :download:`Download all examples in Python source code: auto_examples_python.zip </auto_examples/auto_examples_python.zip>`

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

      :download:`Download all examples in Jupyter notebooks: auto_examples_jupyter.zip </auto_examples/auto_examples_jupyter.zip>`


.. only:: html

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

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