Merge pull request #14 from longyangking/main

add K-means interactive

notebooks/yang/Kmeans interactive.ipynb

@@ -0,0 +1,166 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# K-means interactive\n",
+    "\n",
+    "> Yang"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 22,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "%matplotlib qt\n",
+    "import matplotlib.pyplot as plt\n",
+    "import numpy as np\n",
+    "from matplotlib.widgets import Button\n",
+    "from matplotlib.widgets import PolygonSelector\n",
+    "from sklearn.cluster import KMeans\n",
+    "\n",
+    "def colors_from_lbs(lbs, colors=None):\n",
+    "    mpl_20 = ['#1f77b4', '#ff7f0e', '#2ca02c', '#d62728', '#9467bd',\n",
+    "          '#8c564b', '#e377c2', '#7f7f7f', '#bcbd22', '#17becf',\n",
+    "          '#3397dc', '#ff993e', '#3fca3f', '#df5152', '#a985ca',\n",
+    "          '#ad7165', '#e992ce', '#999999', '#dbdc3c', '#35d8e9']\n",
+    "    \n",
+    "    if colors is None:\n",
+    "        colors = np.array(mpl_20)\n",
+    "    else:\n",
+    "        colors = np.array(colors)\n",
+    "    lbs = np.array(lbs) % len(colors)\n",
+    "    return colors[lbs]\n",
+    "\n",
+    "rng = np.random.RandomState(0)\n",
+    "n_samples = 1000\n",
+    "cov = [[1, 0], [0, 1]]\n",
+    "X = np.concatenate([\n",
+    "    rng.multivariate_normal(mean=[-2, 0], cov=cov, size=n_samples), \n",
+    "    rng.multivariate_normal(mean=[2, 0], cov=cov, size=n_samples)])\n",
+    "\n",
+    "kmeans = KMeans(n_clusters=2, random_state=0, n_init=\"auto\")\n",
+    "labels = kmeans.fit_predict(X)\n",
+    "\n",
+    "#kmeans.cluster_centers_ = np.array([[1,2],[2,1]],dtype=kmeans.cluster_centers_.dtype)\n",
+    "#centers = kmeans.predict(X)\n",
+    "centers = kmeans.cluster_centers_\n",
+    "\n",
+    "fig, (ax_orig, ax_redim) = plt.subplots(1, 2, figsize=(12, 6))\n",
+    "\n",
+    "def plot_figure(axe_list, X, centers):\n",
+    "    ax_orig, ax_redim = axe_list\n",
+    "\n",
+    "    #labels = [np.argmax([np.linalg.norm(x-center) for center in centers]) for x in X]\n",
+    "    kmeans.cluster_centers_ = np.array(centers, dtype=np.float64)\n",
+    "    labels = kmeans.predict(X)    \n",
+    "\n",
+    "    ax_orig.clear()\n",
+    "    ax_orig.scatter(X[:, 0], X[:, 1], alpha=0.3, label=\"samples\", c=colors_from_lbs(labels))\n",
+    "    ax_orig.scatter(centers[:,0], centers[:,1], s=50, c='black', edgecolors='r')\n",
+    "    ax_orig.set(\n",
+    "        aspect=\"auto\", \n",
+    "        title=\"K-means\",\n",
+    "        xlabel=\"first feature\",\n",
+    "        ylabel=\"second feature\",\n",
+    "    )\n",
+    "\n",
+    "    ax_redim.clear()\n",
+    "    class_name = ['class {0}'.format(i) for i in range(len(centers))]\n",
+    "\n",
+    "    # update labels\n",
+    "    counts = [np.sum(labels==i) for i in range(len(centers))]\n",
+    "    \n",
+    "\n",
+    "    ax_redim.bar(class_name, counts, label=class_name)\n",
+    "    #ax_redim.bar((X @ _component.T - _x_center @ _component.T),50)\n",
+    "    ax_redim.set(\n",
+    "        aspect=\"auto\",\n",
+    "        title=\"Clustering results\",\n",
+    "        xlabel=\"Main feature\",\n",
+    "        ylabel=\"Number of samples\",\n",
+    "    )\n",
+    "    fig.canvas.draw_idle()\n",
+    "\n",
+    "plot_figure((ax_orig, ax_redim), X, centers)\n",
+    "\n",
+    "def onselect(verts):\n",
+    "    centers = np.array(verts)\n",
+    "    plot_figure((ax_orig, ax_redim), X, centers)\n",
+    "\n",
+    "selector = PolygonSelector(ax_orig, onselect=onselect, \n",
+    "                           props=dict(color='r', linestyle='', linewidth=3, alpha=0.6, label=f\"Component\"))\n",
+    "selector.verts = centers\n",
+    "\n",
+    "\n",
+    "# ax_redim.hist((X @ component.T - x_center @ component.T),50)\n",
+    "# ax_redim.set(\n",
+    "#     aspect=\"auto\",\n",
+    "#     title=\"1-dimensional dataset after dimension reduction\",\n",
+    "#     xlabel=\"Main feature\",\n",
+    "#     ylabel=\"Number of samples\",\n",
+    "# )\n",
+    "#_asp = np.diff(ax_orig.get_ylim())[0] / np.diff(ax_orig.get_xlim())[0]\n",
+    "#ax_redim.set_aspect(_asp)\n",
+    "\n",
+    "plt.tight_layout()\n",
+    "plt.show()\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "\n",
+    "\n",
+    "# pca = PCA(n_components=1).fit(X)\n",
+    "# component = pca.components_.reshape(-1)\n",
+    "\n",
+    "# # print(pca.components_)\n",
+    "# # print(pca.explained_variance_)\n",
+    "# # print(list(zip(pca.components_, pca.explained_variance_)))\n",
+    "\n",
+    "# # fig, (ax_orig, ax_redim) = plt.subplots(1, 2, figsize=(12, 6))\n",
+    "# # ax_orig.scatter(X[:, 0], X[:, 1], alpha=0.3, label=\"samples\")\n",
+    "# # x_center = np.mean(X, axis=0)\n",
+    "\n",
+    "# comp_vector = [component, x_center]\n",
+    "\n",
+    "# ax_orig.set(\n",
+    "#     aspect=\"auto\", \n",
+    "#     title=\"2-dimensional dataset with principal components\",\n",
+    "#     xlabel=\"first feature\",\n",
+    "#     ylabel=\"second feature\",\n",
+    "# )\n",
+    "\n",
+    "\n"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "base",
+   "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": "3.9.19"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}