Pushing the docs to dev/ for branch: main, commit d7c84b239c18a4627e6af7c369a504cb7609bfcd

Author: sklearn-ci

dev/_downloads/07fcc19ba03226cd3d83d4e40ec44385/auto_examples_python.zip

@@ -0,0 +1,54 @@
+{
+  "cells": [
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {
+        "collapsed": false
+      },
+      "outputs": [],
+      "source": [
+        "%matplotlib inline"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "\n# GMM Initialization Methods\n\nExamples of the different methods of initialization in Gaussian Mixture Models\n\nSee `gmm` for more information on the estimator.\n\nHere we generate some sample data with four easy to identify clusters. The\npurpose of this example is to show the four different methods for the\ninitialization parameter *init_param*.\n\nThe four initializations are *kmeans* (default), *random*, *random_from_data* and\n*k-means++*.\n\nOrange diamonds represent the initialization centers for the gmm generated by\nthe *init_param*. The rest of the data is represented as crosses and the\ncolouring represents the eventual associated classification after the GMM has\nfinished.\n\nThe numbers in the top right of each subplot represent the number of\niterations taken for the GaussianMixture to converge and the relative time\ntaken for the initialization part of the algorithm to run. The shorter\ninitialization times tend to have a greater number of iterations to converge.\n\nThe initialization time is the ratio of the time taken for that method versus\nthe time taken for the default *kmeans* method. As you can see all three\nalternative methods take less time to initialize when compared to *kmeans*.\n\nIn this example, when initialized with *random_from_data* or *random* the model takes\nmore iterations to converge. Here *k-means++* does a good job of both low\ntime to initialize and low number of GaussianMixture iterations to converge.\n"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {
+        "collapsed": false
+      },
+      "outputs": [],
+      "source": [
+        "# Author: Gordon Walsh <[email protected]>\n# Data generation code from Jake Vanderplas <[email protected]>\n\nimport matplotlib.pyplot as plt\nimport numpy as np\nfrom sklearn.mixture import GaussianMixture\nfrom sklearn.utils.extmath import row_norms\nfrom sklearn.datasets._samples_generator import make_blobs\nfrom timeit import default_timer as timer\n\nprint(__doc__)\n\n# Generate some data\n\nX, y_true = make_blobs(n_samples=4000, centers=4, cluster_std=0.60, random_state=0)\nX = X[:, ::-1]\n\nn_samples = 4000\nn_components = 4\nx_squared_norms = row_norms(X, squared=True)\n\n\ndef get_initial_means(X, init_params, r):\n    # Run a GaussianMixture with max_iter=0 to output the initalization means\n    gmm = GaussianMixture(\n        n_components=4, init_params=init_params, tol=1e-9, max_iter=0, random_state=r\n    ).fit(X)\n    return gmm.means_\n\n\nmethods = [\"kmeans\", \"random_from_data\", \"k-means++\", \"random\"]\ncolors = [\"navy\", \"turquoise\", \"cornflowerblue\", \"darkorange\"]\ntimes_init = {}\nrelative_times = {}\n\nplt.figure(figsize=(4 * len(methods) // 2, 6))\nplt.subplots_adjust(\n    bottom=0.1, top=0.9, hspace=0.15, wspace=0.05, left=0.05, right=0.95\n)\n\nfor n, method in enumerate(methods):\n    r = np.random.RandomState(seed=1234)\n    plt.subplot(2, len(methods) // 2, n + 1)\n\n    start = timer()\n    ini = get_initial_means(X, method, r)\n    end = timer()\n    init_time = end - start\n\n    gmm = GaussianMixture(\n        n_components=4, means_init=ini, tol=1e-9, max_iter=2000, random_state=r\n    ).fit(X)\n\n    times_init[method] = init_time\n    for i, color in enumerate(colors):\n        data = X[gmm.predict(X) == i]\n        plt.scatter(data[:, 0], data[:, 1], color=color, marker=\"x\")\n\n    plt.scatter(\n        ini[:, 0], ini[:, 1], s=75, marker=\"D\", c=\"orange\", lw=1.5, edgecolors=\"black\"\n    )\n    relative_times[method] = times_init[method] / times_init[methods[0]]\n\n    plt.xticks(())\n    plt.yticks(())\n    plt.title(method, loc=\"left\", fontsize=12)\n    plt.title(\n        \"Iter %i | Init Time %.2fx\" % (gmm.n_iter_, relative_times[method]),\n        loc=\"right\",\n        fontsize=10,\n    )\nplt.suptitle(\"GMM iterations and relative time taken to initialize\")\nplt.show()"
+      ]
+    }
+  ],
+  "metadata": {
+    "kernelspec": {
+      "display_name": "Python 3",
+      "language": "python",
+      "name": "python3"
+    },
+    "language_info": {
+      "codemirror_mode": {
+        "name": "ipython",
+        "version": 3
+      },
+      "file_extension": ".py",
+      "mimetype": "text/x-python",
+      "name": "python",
+      "nbconvert_exporter": "python",
+      "pygments_lexer": "ipython3",
+      "version": "3.9.12"
+    }
+  },
+  "nbformat": 4,
+  "nbformat_minor": 0
+}

dev/_downloads/4969738e4f46f5753249be090ab7b3f5/plot_gmm_init.ipynb

@@ -0,0 +1,109 @@
+"""
+==========================
+GMM Initialization Methods
+==========================
+
+Examples of the different methods of initialization in Gaussian Mixture Models
+
+See :ref:`gmm` for more information on the estimator.
+
+Here we generate some sample data with four easy to identify clusters. The
+purpose of this example is to show the four different methods for the
+initialization parameter *init_param*.
+
+The four initializations are *kmeans* (default), *random*, *random_from_data* and
+*k-means++*.
+
+Orange diamonds represent the initialization centers for the gmm generated by
+the *init_param*. The rest of the data is represented as crosses and the
+colouring represents the eventual associated classification after the GMM has
+finished.
+
+The numbers in the top right of each subplot represent the number of
+iterations taken for the GaussianMixture to converge and the relative time
+taken for the initialization part of the algorithm to run. The shorter
+initialization times tend to have a greater number of iterations to converge.
+
+The initialization time is the ratio of the time taken for that method versus
+the time taken for the default *kmeans* method. As you can see all three
+alternative methods take less time to initialize when compared to *kmeans*.
+
+In this example, when initialized with *random_from_data* or *random* the model takes
+more iterations to converge. Here *k-means++* does a good job of both low
+time to initialize and low number of GaussianMixture iterations to converge.
+"""
+
+
+# Author: Gordon Walsh <[email protected]>
+# Data generation code from Jake Vanderplas <[email protected]>
+
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.mixture import GaussianMixture
+from sklearn.utils.extmath import row_norms
+from sklearn.datasets._samples_generator import make_blobs
+from timeit import default_timer as timer
+
+print(__doc__)
+
+# Generate some data
+
+X, y_true = make_blobs(n_samples=4000, centers=4, cluster_std=0.60, random_state=0)
+X = X[:, ::-1]
+
+n_samples = 4000
+n_components = 4
+x_squared_norms = row_norms(X, squared=True)
+
+
+def get_initial_means(X, init_params, r):
+    # Run a GaussianMixture with max_iter=0 to output the initalization means
+    gmm = GaussianMixture(
+        n_components=4, init_params=init_params, tol=1e-9, max_iter=0, random_state=r
+    ).fit(X)
+    return gmm.means_
+
+
+methods = ["kmeans", "random_from_data", "k-means++", "random"]
+colors = ["navy", "turquoise", "cornflowerblue", "darkorange"]
+times_init = {}
+relative_times = {}
+
+plt.figure(figsize=(4 * len(methods) // 2, 6))
+plt.subplots_adjust(
+    bottom=0.1, top=0.9, hspace=0.15, wspace=0.05, left=0.05, right=0.95
+)
+
+for n, method in enumerate(methods):
+    r = np.random.RandomState(seed=1234)
+    plt.subplot(2, len(methods) // 2, n + 1)
+
+    start = timer()
+    ini = get_initial_means(X, method, r)
+    end = timer()
+    init_time = end - start
+
+    gmm = GaussianMixture(
+        n_components=4, means_init=ini, tol=1e-9, max_iter=2000, random_state=r
+    ).fit(X)
+
+    times_init[method] = init_time
+    for i, color in enumerate(colors):
+        data = X[gmm.predict(X) == i]
+        plt.scatter(data[:, 0], data[:, 1], color=color, marker="x")
+
+    plt.scatter(
+        ini[:, 0], ini[:, 1], s=75, marker="D", c="orange", lw=1.5, edgecolors="black"
+    )
+    relative_times[method] = times_init[method] / times_init[methods[0]]
+
+    plt.xticks(())
+    plt.yticks(())
+    plt.title(method, loc="left", fontsize=12)
+    plt.title(
+        "Iter %i | Init Time %.2fx" % (gmm.n_iter_, relative_times[method]),
+        loc="right",
+        fontsize=10,
+    )
+plt.suptitle("GMM iterations and relative time taken to initialize")
+plt.show()