Files
2026-07-13 13:30:25 +08:00

215 lines
53 KiB
Plaintext
Executable File

{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# 9. Mixture Models and EM"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
"import matplotlib.pyplot as plt\n",
"import numpy as np\n",
"from sklearn.datasets import fetch_mldata\n",
"%matplotlib inline\n",
"\n",
"from prml.clustering import KMeans\n",
"from prml.rv import (\n",
" MultivariateGaussianMixture,\n",
" BernoulliMixture\n",
")\n",
"\n",
"np.random.seed(1111)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 9.1 K-means Clustering"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
"# training data\n",
"x1 = np.random.normal(size=(100, 2))\n",
"x1 += np.array([-5, -5])\n",
"x2 = np.random.normal(size=(100, 2))\n",
"x2 += np.array([5, -5])\n",
"x3 = np.random.normal(size=(100, 2))\n",
"x3 += np.array([0, 5])\n",
"x_train = np.vstack((x1, x2, x3))\n",
"\n",
"x0, x1 = np.meshgrid(np.linspace(-10, 10, 100), np.linspace(-10, 10, 100))\n",
"x = np.array([x0, x1]).reshape(2, -1).T"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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\n",
"text/plain": [
"<Figure size 432x288 with 1 Axes>"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"kmeans = KMeans(n_clusters=3)\n",
"kmeans.fit(x_train)\n",
"cluster = kmeans.predict(x_train)\n",
"plt.scatter(x_train[:, 0], x_train[:, 1], c=cluster)\n",
"plt.scatter(kmeans.centers[:, 0], kmeans.centers[:, 1], s=200, marker='X', lw=2, c=['purple', 'cyan', 'yellow'], edgecolor=\"white\")\n",
"plt.contourf(x0, x1, kmeans.predict(x).reshape(100, 100), alpha=0.1)\n",
"plt.xlim(-10, 10)\n",
"plt.ylim(-10, 10)\n",
"plt.gca().set_aspect('equal', adjustable='box')\n",
"plt.show()"
]
},
{
"cell_type": "markdown",
"metadata": {
"collapsed": true
},
"source": [
"## 9.2 Mixture of Gaussians"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {
"scrolled": true
},
"outputs": [
{
"data": {
"image/png": "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\n",
"text/plain": [
"<Figure size 432x288 with 1 Axes>"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"gmm = MultivariateGaussianMixture(n_components=3)\n",
"gmm.fit(x_train)\n",
"p = gmm.classify_proba(x_train)\n",
"\n",
"plt.scatter(x_train[:, 0], x_train[:, 1], c=p)\n",
"plt.scatter(gmm.mu[:, 0], gmm.mu[:, 1], s=200, marker='X', lw=2, c=['red', 'green', 'blue'], edgecolor=\"white\")\n",
"plt.xlim(-10, 10)\n",
"plt.ylim(-10, 10)\n",
"plt.gca().set_aspect(\"equal\")\n",
"plt.show()"
]
},
{
"cell_type": "markdown",
"metadata": {
"collapsed": true
},
"source": [
"### 9.3.3 Mixtures of Bernoulli distributions"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [],
"source": [
"mnist = fetch_mldata(\"MNIST original\")\n",
"x = mnist.data\n",
"y = mnist.target\n",
"x_train = []\n",
"for i in [0, 1, 2, 3, 4]:\n",
" x_train.append(x[np.random.choice(np.where(y == i)[0], 200)])\n",
"x_train = np.concatenate(x_train, axis=0)\n",
"x_train = (x_train > 127).astype(np.float)"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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\n",
"text/plain": [
"<Figure size 1440x360 with 5 Axes>"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"bmm = BernoulliMixture(n_components=5)\n",
"bmm.fit(x_train)\n",
"\n",
"plt.figure(figsize=(20, 5))\n",
"for i, mean in enumerate(bmm.mu):\n",
" plt.subplot(1, 5, i + 1)\n",
" plt.imshow(mean.reshape(28, 28), cmap=\"gray\")\n",
" plt.axis('off')\n",
"plt.show()"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"anaconda-cloud": {},
"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": "3.7.0"
}
},
"nbformat": 4,
"nbformat_minor": 1
}