src/examples/iris_svm_classification.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Iris dataset classification with support vector machine"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "from sklearn.datasets import load_iris\n",
    "from sklearn.preprocessing import LabelBinarizer\n",
    "import pyprind\n",
    "from pydeeptoy.networks import *\n",
    "from pydeeptoy.optimizers import *\n",
    "from pydeeptoy.losses import *\n",
    "from sklearn.metrics import accuracy_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "iris = load_iris()\n",
    "X = iris.data[:, :2]\n",
    "y = iris.target\n",
    "# standardize\n",
    "mean = X.mean(axis=0)\n",
    "std = X.std(axis=0)\n",
    "X = (X - mean) / std\n",
    "one_hot_y = np.array(LabelBinarizer().fit_transform(y).T)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0%                                                        100%\n",
      "[████████████████████████████████████████████████████████████] | ETA: 00:00:00 | Item ID: loss = 0.70\n",
      "Total time elapsed: 00:00:00\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.76666666666666672"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cg = ComputationalGraph()\n",
    "x_in = cg.constant(name=\"X.T\")\n",
    "W = cg.variable(\"W\", 0.01 * np.random.randn(3, X.shape[1]))\n",
    "svm_output = cg.matrix_multiply(W, x_in)\n",
    "svm_output.name = \"svm_output\"\n",
    "\n",
    "y_train = cg.constant(name=\"one_hot_y\")\n",
    "batch_size=256\n",
    "loss = hinge(cg, svm_output, y_train, \"loss_hinge\")\n",
    "\n",
    "ctx = SimulationContext()        \n",
    "\n",
    "sgd = MomentumSgdOptimizer(learning_rate=0.01)\n",
    "epochs = 1000\n",
    "bar = pyprind.ProgBar(epochs, bar_char='█', width=60, track_time=True, stream=1)\n",
    "for epoch in range(0, epochs):\n",
    "    indexes = np.arange(0, len(X))\n",
    "    np.random.shuffle(indexes)\n",
    "    train_x = X[indexes]\n",
    "    train_y = one_hot_y[:, indexes]    \n",
    "    for batch in range(0, len(train_x), batch_size):\n",
    "        batch_x = train_x[batch:batch + batch_size]\n",
    "        batch_y = train_y[:, batch:batch + batch_size]\n",
    "        sgd.minimize(ctx, cg, {x_in: batch_x.T, y_train: batch_y})    \n",
    "    bar.update(item_id = \"loss = {0:.2f}\".format(ctx[loss].value))\n",
    "\n",
    "ctx.forward(cg, {x_in: X.T}, out=[svm_output])\n",
    "y_pred = np.argmax(ctx[svm_output].value, axis=0)\n",
    "\n",
    "accuracy = accuracy_score(y, y_pred)\n",
    "accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/kirill/anaconda/lib/python3.5/site-packages/matplotlib/collections.py:590: FutureWarning: elementwise comparison failed; returning scalar instead, but in the future will perform elementwise comparison\n",
      "  if self._edgecolors == str('face'):\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAk0AAAJKCAYAAAAx/3HgAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nHwFQEiKDs84qAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x108572518>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "colors = \"bry\"\n",
    "fig = plt.figure()\n",
    "   \n",
    "h = 0.01\n",
    "x_min, x_max = X[:, 0].min() - 1, X[:, 0].max() + 1\n",
    "y_min, y_max = X[:, 1].min() - 1, X[:, 1].max() + 1\n",
    "xx, yy = np.meshgrid(np.arange(x_min, x_max, h),\n",
    "                     np.arange(y_min, y_max, h))\n",
    "\n",
    "mps = np.c_[xx.ravel(), yy.ravel()]\n",
    "ctx.forward(cg, {x_in: mps.T, y_train: 1})\n",
    "Z = np.argmax(ctx[svm_output].value, axis=0)\n",
    "# Put the result into a color plot\n",
    "Z = Z.reshape(xx.shape)\n",
    "cs = plt.contourf(xx, yy, Z, cmap=plt.cm.Paired)\n",
    "\n",
    "for i, color in zip(np.unique(iris.target), colors):\n",
    "    idx = np.where(y == i)\n",
    "    plt.scatter(X[y == i, 0], X[y == i, 1], c=color, label=iris.target_names[i], cmap=plt.cm.Paired)\n",
    "\n",
    "plt.axis('tight')\n",
    "fig.set_size_inches(10, 10)"
   ]
  }
 ],
 "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}