Classification using MNIST dataset.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# from six.moves import urllib\n",
    "# from sklearn.datasets import fetch_mldata\n",
    "# try:\n",
    "#     mnist = fetch_mldata('MNIST original')\n",
    "# except urllib.error.HTTPError as ex:\n",
    "#     print(\"Could not download MNIST data from mldata.org, trying alternative...\")\n",
    "\n",
    "#     # Alternative method to load MNIST, if mldata.org is down\n",
    "#     from scipy.io import loadmat\n",
    "#     mnist_alternative_url = \"https://github.com/amplab/datascience-sp14/raw/master/lab7/mldata/mnist-original.mat\"\n",
    "#     mnist_path = \"./mnist-original.mat\"\n",
    "#     response = urllib.request.urlopen(mnist_alternative_url)\n",
    "#     with open(mnist_path, \"wb\") as f:\n",
    "#         content = response.read()\n",
    "#         f.write(content)\n",
    "#     mnist_raw = loadmat(mnist_path)\n",
    "#     mnist = {\n",
    "#         \"data\": mnist_raw[\"data\"].T,\n",
    "#         \"target\": mnist_raw[\"label\"][0],\n",
    "#         \"COL_NAMES\": [\"label\", \"data\"],\n",
    "#         \"DESCR\": \"mldata.org dataset: mnist-original\",\n",
    "#     }\n",
    "#     print(\"Success!\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "# Common imports\n",
    "import numpy as np\n",
    "import os\n",
    "\n",
    "# to make this notebook's output stable across runs\n",
    "np.random.seed(42)\n",
    "\n",
    "# To plot pretty figures\n",
    "%matplotlib inline\n",
    "import matplotlib as mpl\n",
    "import matplotlib.pyplot as plt\n",
    "mpl.rc('axes', labelsize=14)\n",
    "mpl.rc('xtick', labelsize=12)\n",
    "mpl.rc('ytick', labelsize=12)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Success!\n"
     ]
    }
   ],
   "source": [
    "from scipy.io import loadmat\n",
    "mnist_path = './mnist-original.mat'\n",
    "mnist_raw = loadmat(mnist_path)\n",
    "mnist = {\n",
    "    \"data\": mnist_raw[\"data\"].T,\n",
    "    \"target\": mnist_raw[\"label\"][0],\n",
    "    \"COL_NAMES\": [\"label\", \"data\"],\n",
    "    \"DESCR\": \"mldata.org dataset: mnist-original\",\n",
    "}\n",
    "print(\"Success!\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'data': array([[0, 0, 0, ..., 0, 0, 0],\n",
       "        [0, 0, 0, ..., 0, 0, 0],\n",
       "        [0, 0, 0, ..., 0, 0, 0],\n",
       "        ...,\n",
       "        [0, 0, 0, ..., 0, 0, 0],\n",
       "        [0, 0, 0, ..., 0, 0, 0],\n",
       "        [0, 0, 0, ..., 0, 0, 0]], dtype=uint8),\n",
       " 'target': array([0., 0., 0., ..., 9., 9., 9.]),\n",
       " 'COL_NAMES': ['label', 'data'],\n",
       " 'DESCR': 'mldata.org dataset: mnist-original'}"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mnist"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "X, y = mnist[\"data\"], mnist[\"target\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(70000, 784)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(70000,)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib \n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_digit(digit_image):\n",
    "    plt.imshow(digit_image, cmap=matplotlib.cm.binary,\n",
    "          interpolation=\"nearest\")\n",
    "    plt.axis(\"off\")\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# EXTRA\n",
    "def plot_digits(instances, images_per_row=10, **options):\n",
    "    size = 28\n",
    "    images_per_row = min(len(instances), images_per_row)\n",
    "    images = [instance.reshape(size,size) for instance in instances]\n",
    "    n_rows = (len(instances) - 1) // images_per_row + 1\n",
    "    row_images = []\n",
    "    n_empty = n_rows * images_per_row - len(instances)\n",
    "    images.append(np.zeros((size, size * n_empty)))\n",
    "    for row in range(n_rows):\n",
    "        rimages = images[row * images_per_row : (row + 1) * images_per_row]\n",
    "        row_images.append(np.concatenate(rimages, axis=1))\n",
    "    image = np.concatenate(row_images, axis=0)\n",
    "    plt.imshow(image, cmap = mpl.cm.binary, **options)\n",
    "    plt.axis(\"off\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "some_digit = X[36000]\n",
    "some_digit_image = some_digit.reshape(28,28)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "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": [
    "plot_digits(X[100:200])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP0AAAD8CAYAAAC8aaJZAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAABZlJREFUeJzt3T1vT28cwOFWhBBTQ2IgweghMWkMHSV9ARIRmzfBJJEYEEYSi9XD0IjJIEbsdg+jsBiEkLT/N+B3178/9OFzXWO/Oedu03xyD3fOObMrKyszQMe29f4FgH9L9BAjeogRPcSIHmJEDzHb12ld54Tw983+6od2eogRPcSIHmJEDzGihxjRQ4zoIUb0ECN6iBE9xIgeYkQPMaKHGNFDjOghRvQQI3qIET3EiB5iRA8xoocY0UOM6CFG9BAjeogRPcSIHmJEDzGihxjRQ8x6faoaftvPnz+H88XFxeF83759E2cPHjwYXjs7+8uvPW9qdnqIET3EiB5iRA8xoocY0UOM6CHGOT0b3r1794bzFy9erPneZ8+enWq+GdnpIUb0ECN6iBE9xIgeYkQPMaKHmNmVlZX1WHddFmVz2r9//3D+8ePH4Xz0TPznz5+H187NzQ3nG9wv/3A7PcSIHmJEDzGihxjRQ4zoIcajtay7a9euDeefPn2a6v537tyZONvkR3JrYqeHGNFDjOghRvQQI3qIET3EiB5inNPzTzx58mTi7MaNG8Nrl5eXh/P5+fnh/OTJk8N5jZ0eYkQPMaKHGNFDjOghRvQQI3qI8Qps/ogvX74M5wsLCxNnb968GV67Z8+e4fzVq1fD+fHjx4fzLcwrsAHRQ47oIUb0ECN6iBE9xIgeYjxPzx9x6dKl4Xy1s/iR06dPD+fhc/g1sdNDjOghRvQQI3qIET3EiB5iRA8xnqfnt7x//344P3r06HD+7du3Na/99evX4Xz37t1rvvcW53l6QPSQI3qIET3EiB5iRA8xHq1lZmZmZubly5fD+eLi4nC+2pHcrl27Js6ePXu25mv5/+z0ECN6iBE9xIgeYkQPMaKHGNFDjEdrI1Z7PHW110iv9mjtah4+fDhxdu7cuanuzUQerQVEDzmihxjRQ4zoIUb0ECN6iPE8fcRqn4qe9hx+586dw/mhQ4emuj9/jp0eYkQPMaKHGNFDjOghRvQQI3qIcU6/hXz//n3i7OrVq3917StXrgzn8/Pzf3V9fp+dHmJEDzGihxjRQ4zoIUb0ECN6iPHe+y1k9G758+fPT3XvhYWF4fz58+fD+Y4dO6ZanzXx3ntA9JAjeogRPcSIHmJEDzEerd1ElpaWhvPLly+v+d579+4dzm/dujWcO5LbPOz0ECN6iBE9xIgeYkQPMaKHGNFDjHP6DWR5eXk4f/To0XD+4cOHibPVztGfPn06nJ86dWo438hGn+k+duzY8Npt27bevrj1/iJgSPQQI3qIET3EiB5iRA8xoocY5/QbyMWLF4fzx48fr/nehw8fHs4PHDiw5ntvdCdOnFjvX2FDsdNDjOghRvQQI3qIET3EiB5iRA8xzun/obdv3w7n7969m+r+R44cmTi7efPm8NqDBw9OtTabh50eYkQPMaKHGNFDjOghRvQQ48juD/rx48dwfuHCheH89evXU60/OnY7c+bMVPdm67DTQ4zoIUb0ECN6iBE9xIgeYkQPMc7p/6Dbt28P59Oew6/2GuulpaWJs127dk21NluHnR5iRA8xoocY0UOM6CFG9BAjeohxTv8/3b9/f+Ls+vXrU917+/bxv+Pu3bvD+dzc3FTr02CnhxjRQ4zoIUb0ECN6iBE9xIgeYmZXVlbWY911WRRiZn/1Qzs9xIgeYkQPMaKHGNFDjOghRvQQI3qIET3EiB5iRA8xoocY0UOM6CFG9BAjeogRPcSIHmJEDzGihxjRQ4zoIUb0ECN6iBE9xIgeYkQPMaKHGNFDjOghZvs6rfvLT+gCf5+dHmJEDzGihxjRQ4zoIUb0ECN6iBE9xIgeYkQPMaKHGNFDjOghRvQQI3qIET3EiB5iRA8xoocY0UOM6CFG9BAjeogRPcSIHmJEDzH/ASvjqt88MmTEAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "digit_1 = X[10000]\n",
    "digit_1_image = digit_1.reshape(28, 28)\n",
    "plot_digit(digit_1_image)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1.0"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y[10000]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "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": [
    "plt.imshow(some_digit_image, cmap = matplotlib.cm.binary,\n",
    "          interpolation=\"nearest\")\n",
    "plt.axis(\"off\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5.0"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y[36000]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = X[:60000], X[60000:], y[:60000], y[60000:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "shuffle_index = np.random.permutation(60000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, y_train = X_train[shuffle_index], y_train[shuffle_index]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*We train a binary classifier for the digit 5*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train_5 = (y_train == 5) # Only true for 5s and False for all other digits"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_test_5 = (y_test == 5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Now we train a Stochastic Gradient Classifier*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.linear_model import SGDClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "sgd_clf = SGDClassifier(random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "SGDClassifier(alpha=0.0001, average=False, class_weight=None,\n",
       "       early_stopping=False, epsilon=0.1, eta0=0.0, fit_intercept=True,\n",
       "       l1_ratio=0.15, learning_rate='optimal', loss='hinge', max_iter=None,\n",
       "       n_iter=None, n_iter_no_change=5, n_jobs=None, penalty='l2',\n",
       "       power_t=0.5, random_state=42, shuffle=True, tol=None,\n",
       "       validation_fraction=0.1, verbose=0, warm_start=False)"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sgd_clf.fit(X_train, y_train_5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ True] == 0.0\n",
      "[ True] == 4.0\n",
      "[ True] == 5.0\n",
      "[ True] == 5.0\n",
      "[ True] == 5.0\n",
      "[ True] == 5.0\n",
      "[ True] == 5.0\n",
      "[ True] == 5.0\n"
     ]
    }
   ],
   "source": [
    "for index in range(1000, 70000, 1000):\n",
    "    digit = X[index]\n",
    "    predicted = sgd_clf.predict([digit])\n",
    "    if predicted:\n",
    "        print(\"{} == {}\".format(predicted, y[index]))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ True])"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sgd_clf.predict([some_digit])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0,   0,   0,   0,   0,  86, 131, 225, 225, 225,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0,   0,  13,  73, 197, 253, 252, 252, 252, 252,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   4,\n",
       "         29,  29, 154, 187, 252, 252, 253, 252, 252, 233, 145,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,  29,\n",
       "        252, 253, 252, 252, 252, 252, 253, 204, 112,  37,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0, 169,\n",
       "        253, 255, 253, 228, 126,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,  98, 243,\n",
       "        252, 253, 252, 246, 130,  38,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,  98, 240, 252,\n",
       "        252, 253, 252, 252, 252, 221,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0, 225, 252, 252,\n",
       "        236, 225, 223, 230, 252, 252,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0, 146, 252, 157,\n",
       "         50,   0,   0,  25, 205, 252,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0,   0,  26, 207, 253,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,  29,  19,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0,   0,  73, 205, 252,  79,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0, 120, 215, 209, 175,   0,   0,   0,   0,   0,\n",
       "          0,   0,  19, 209, 252, 220,  79,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0, 174, 252, 252, 239, 140,   0,   0,   0,   0,\n",
       "          0,  29, 104, 252, 249, 177,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0, 174, 252, 252, 223,   0,   0,   0,   0,   0,\n",
       "          0, 174, 252, 252, 223,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0, 141, 241, 253, 146,   0,   0,   0,   0, 169,\n",
       "        253, 255, 253, 253,  84,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0, 178, 252, 154,  85,  85, 210, 225, 243,\n",
       "        252, 215, 121,  27,   9,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,  66, 208, 220, 252, 253, 252, 252, 214,\n",
       "        195,  31,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,  19,  37,  84, 146, 223, 114,  28,\n",
       "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "          0,   0]], dtype=uint8)"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "some_digit_image"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Implementing cross-validation*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import StratifiedKFold"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.base import clone"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "skfolds = StratifiedKFold(n_splits=3, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9502\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.96565\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.96495\n"
     ]
    }
   ],
   "source": [
    "for train_index, test_index in skfolds.split(X_train, y_train_5):\n",
    "    \n",
    "    clone_clf = clone(sgd_clf)\n",
    "    X_train_folds = X_train[train_index]\n",
    "    y_train_folds = (y_train_5[train_index])\n",
    "    X_test_fold = X_train[test_index]\n",
    "    y_test_fold = (y_train_5[test_index])\n",
    "    \n",
    "    clone_clf.fit(X_train_folds, y_train_folds)\n",
    "    y_pred = clone_clf.predict(X_test_fold)\n",
    "    n_correct = sum(y_pred == y_test_fold)\n",
    "    print(n_correct / len(y_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Evaluate our model using cross_val_score()*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import cross_val_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([0.9502 , 0.96565, 0.96495])"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cross_val_score(sgd_clf, X_train, y_train_5, cv=3, scoring=\"accuracy\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Dumb classifier for not-5 class*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.base import BaseEstimator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Never5Classifier(BaseEstimator):\n",
    "    \n",
    "    def fit(self, X, y=None):\n",
    "        pass\n",
    "\n",
    "    def predict(self, X):\n",
    "        return np.zeros((len(X), 1), dtype=bool)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "# let's try out the accuracy of our dumb classifier\n",
    "never_5_clf = Never5Classifier()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.909  , 0.90715, 0.9128 ])"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cross_val_score(never_5_clf, X_train, y_train_5, cv=3, scoring=\"accuracy\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Confusion Matrix*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import cross_val_predict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n"
     ]
    }
   ],
   "source": [
    "y_train_pred = cross_val_predict(sgd_clf, X_train, y_train_5, cv=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import confusion_matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[53272,  1307],\n",
       "       [ 1077,  4344]])"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "confusion_matrix(y_train_5, y_train_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[54579,     0],\n",
       "       [    0,  5421]])"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# ideal confusion matrix has non-zero elements only in the diagnol like this...\n",
    "confusion_matrix(y_train_5, y_train_5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Let's find out the precision and recall*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import precision_score, recall_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.7687135020350381"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "precision_score(y_train_5, y_train_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.801328168234643"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "recall_score(y_train_5, y_train_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import f1_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Veer the F1 score is =  0.7846820809248555\n"
     ]
    }
   ],
   "source": [
    "print(\"Veer the F1 score is = \", f1_score(y_train_5, y_train_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_scores = sgd_clf.decision_function([some_digit])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([161855.74572176])"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ True])"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "threshold = 0\n",
    "\n",
    "y_some_digit_pred = (y_scores > threshold)\n",
    "\n",
    "y_some_digit_pred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([False])"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "threshold = 200000\n",
    "y_some_digit_pred = (y_scores > threshold)\n",
    "y_some_digit_pred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n"
     ]
    }
   ],
   "source": [
    "\n",
    "\n",
    "y_scores = cross_val_predict(sgd_clf, X_train, y_train_5, cv=3,\n",
    "                            method=\"decision_function\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ -434076.49813641, -1825667.15281624,  -767086.76186905, ...,\n",
       "        -867191.25267994,  -565357.11420164,  -366599.16018198])"
      ]
     },
     "execution_count": 71,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import precision_recall_curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "precisions, recalls, thresholds = precision_recall_curve(y_train_5, y_scores)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_precision_recall_vs_threshold(precisions, recalls, thresholds):\n",
    "    \n",
    "    plt.plot(thresholds, precisions[:-1], \"b--\", label=\"Precision\")\n",
    "    plt.plot(thresholds, recalls[:-1], \"g-\", label=\"Recall\")\n",
    "    plt.xlabel(\"Threshold\")\n",
    "    plt.legend(loc=\"upper left\")\n",
    "    plt.ylim([0, 1])\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "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": [
    "plot_precision_recall_vs_threshold(precisions, recalls, thresholds)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "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": [
    "plt.plot(precisions, recalls, \"b--\", label=\"Precision\")\n",
    "plt.xlabel(\"Precision\")\n",
    "plt.ylabel(\"Recall\")\n",
    "plt.legend(loc=\"upper left\")\n",
    "plt.ylim([0, 1])\n",
    "plt.tight_layout()\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train_pred_90 = (y_scores > 70000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8659205116491548"
      ]
     },
     "execution_count": 78,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "precision_score(y_train_5, y_train_pred_90)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6993174691016417"
      ]
     },
     "execution_count": 80,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "recall_score(y_train_5, y_train_pred_90)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "If someone says “let’s reach 99% precision,” you should ask, “at what recall?”"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import roc_curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [],
   "source": [
    "fpr, tpr, thresholds = roc_curve(y_train_5, y_scores)\n",
    "# fpr = false positive rate = (1 - true negative rate)\n",
    "# tpr = true positive rate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_roc_curve(fpr, tpr, label=None):\n",
    "    plt.plot(fpr, tpr, linewidth=2, label=label)\n",
    "    plt.plot([0, 1], [0, 1], 'k--')\n",
    "    plt.axis([0, 1, 0, 1])\n",
    "    plt.xlabel('False Positive Rate')\n",
    "    plt.ylabel('True Positive Rate')\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "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": [
    "plot_roc_curve(fpr, tpr, \"Roc Curve\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import roc_auc_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9624496555967156"
      ]
     },
     "execution_count": 89,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "roc_auc_score(y_train_5, y_scores)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Random Forest Classifier*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.ensemble import RandomForestClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [],
   "source": [
    "forest_clf = RandomForestClassifier(random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/ensemble/forest.py:246: FutureWarning: The default value of n_estimators will change from 10 in version 0.20 to 100 in 0.22.\n",
      "  \"10 in version 0.20 to 100 in 0.22.\", FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/ensemble/forest.py:246: FutureWarning: The default value of n_estimators will change from 10 in version 0.20 to 100 in 0.22.\n",
      "  \"10 in version 0.20 to 100 in 0.22.\", FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/ensemble/forest.py:246: FutureWarning: The default value of n_estimators will change from 10 in version 0.20 to 100 in 0.22.\n",
      "  \"10 in version 0.20 to 100 in 0.22.\", FutureWarning)\n"
     ]
    }
   ],
   "source": [
    "y_probas_forest = cross_val_predict(forest_clf, X_train, y_train_5, cv=3,\n",
    "                                   method=\"predict_proba\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[1. , 0. ],\n",
       "       [0.9, 0.1],\n",
       "       [1. , 0. ],\n",
       "       ...,\n",
       "       [1. , 0. ],\n",
       "       [1. , 0. ],\n",
       "       [1. , 0. ]])"
      ]
     },
     "execution_count": 97,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_probas_forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_scores_forest = y_probas_forest[:, 1] # select 1'st column as scores as it is the probability \n",
    "# of the positive class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0. , 0.1, 0. , ..., 0. , 0. , 0. ])"
      ]
     },
     "execution_count": 99,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_scores_forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {},
   "outputs": [],
   "source": [
    "fpr_forest, tpr_forest, thresholds_forest = roc_curve(y_train_5, y_scores_forest)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "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": [
    "plt.plot(fpr, tpr, \"b:\", label=\"SGD\")\n",
    "plot_roc_curve(fpr_forest, tpr_forest, \"Random Forest\")\n",
    "plt.legend(loc=\"lower right\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9931243366003829"
      ]
     },
     "execution_count": 102,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "roc_auc_score(y_train_5, y_scores_forest)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "MultiClass Classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "SGDClassifier(alpha=0.0001, average=False, class_weight=None,\n",
       "       early_stopping=False, epsilon=0.1, eta0=0.0, fit_intercept=True,\n",
       "       l1_ratio=0.15, learning_rate='optimal', loss='hinge', max_iter=None,\n",
       "       n_iter=None, n_iter_no_change=5, n_jobs=None, penalty='l2',\n",
       "       power_t=0.5, random_state=42, shuffle=True, tol=None,\n",
       "       validation_fraction=0.1, verbose=0, warm_start=False)"
      ]
     },
     "execution_count": 105,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sgd_clf.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([5.])"
      ]
     },
     "execution_count": 104,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sgd_clf.predict([some_digit])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [],
   "source": [
    "some_digit_scores = sgd_clf.decision_function([some_digit])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[-311402.62954431, -363517.28355739, -446449.5306454 ,\n",
       "        -183226.61023518, -414337.15339485,  161855.74572176,\n",
       "        -452576.39616343, -471957.14962573, -518542.33997148,\n",
       "        -536774.63961222]])"
      ]
     },
     "execution_count": 107,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "some_digit_scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5"
      ]
     },
     "execution_count": 108,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.argmax(some_digit_scores)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0., 1., 2., 3., 4., 5., 6., 7., 8., 9.])"
      ]
     },
     "execution_count": 110,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sgd_clf.classes_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5.0"
      ]
     },
     "execution_count": 111,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sgd_clf.classes_[5]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Forcefully making One Vs One classifiers*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.multiclass import OneVsOneClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {},
   "outputs": [],
   "source": [
    "ovo_clf = OneVsOneClassifier(SGDClassifier(random_state=42))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "OneVsOneClassifier(estimator=SGDClassifier(alpha=0.0001, average=False, class_weight=None,\n",
       "       early_stopping=False, epsilon=0.1, eta0=0.0, fit_intercept=True,\n",
       "       l1_ratio=0.15, learning_rate='optimal', loss='hinge', max_iter=None,\n",
       "       n_iter=None, n_iter_no_change=5, n_jobs=None, penalty='l2',\n",
       "       power_t=0.5, random_state=42, shuffle=True, tol=None,\n",
       "       validation_fraction=0.1, verbose=0, warm_start=False),\n",
       "          n_jobs=None)"
      ]
     },
     "execution_count": 114,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ovo_clf.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([5.])"
      ]
     },
     "execution_count": 115,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ovo_clf.predict([some_digit])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "45"
      ]
     },
     "execution_count": 116,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(ovo_clf.estimators_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([5.])"
      ]
     },
     "execution_count": 120,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "forest_clf.fit(X_train, y_train)\n",
    "\n",
    "\n",
    "forest_clf.predict([some_digit])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.1, 0. , 0. , 0.1, 0. , 0.8, 0. , 0. , 0. , 0. ]])"
      ]
     },
     "execution_count": 122,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "forest_clf.predict_proba([some_digit])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We evaluate SGDClassifier's accuracy using cross_val_score()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([0.84063187, 0.84899245, 0.86652998])"
      ]
     },
     "execution_count": 123,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cross_val_score(sgd_clf, X_train, y_train, cv=3, scoring=\"accuracy\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We scale the inputs to improve the accuracy of our SGD classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import StandardScaler\n",
    "\n",
    "scaler = StandardScaler()\n",
    "\n",
    "X_train_scaled = scaler.fit_transform(X_train.astype(np.float64))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([0.91011798, 0.90874544, 0.906636  ])"
      ]
     },
     "execution_count": 134,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cross_val_score(sgd_clf, X_train_scaled, y_train, cv=3, scoring=\"accuracy\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Error Analysis*\n",
    "To find the kinds of errors our model makes we analyze the confusion matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "/home/gurupratap/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/stochastic_gradient.py:166: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n"
     ]
    }
   ],
   "source": [
    "y_train_pred = cross_val_predict(sgd_clf, X_train_scaled, y_train, cv=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "metadata": {},
   "outputs": [],
   "source": [
    "conf_mx = confusion_matrix(y_train, y_train_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[5725,    3,   24,    9,   10,   49,   50,   10,   39,    4],\n",
       "       [   2, 6493,   43,   25,    7,   40,    5,   10,  109,    8],\n",
       "       [  51,   41, 5321,  104,   89,   26,   87,   60,  166,   13],\n",
       "       [  47,   46,  141, 5342,    1,  231,   40,   50,  141,   92],\n",
       "       [  19,   29,   41,   10, 5366,    9,   56,   37,   86,  189],\n",
       "       [  73,   45,   36,  193,   64, 4582,  111,   30,  193,   94],\n",
       "       [  29,   34,   44,    2,   42,   85, 5627,   10,   45,    0],\n",
       "       [  25,   24,   74,   32,   54,   12,    6, 5787,   15,  236],\n",
       "       [  52,  161,   73,  156,   10,  163,   61,   25, 5027,  123],\n",
       "       [  43,   35,   26,   92,  178,   28,    2,  223,   82, 5240]])"
      ]
     },
     "execution_count": 136,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "conf_mx"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 288x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.matshow(conf_mx, cmap=plt.cm.gray)\n",
    "plt.show()\n",
    "\n",
    "# the more gray the higher the error"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 143,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Figure size 4000x2000 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 288x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "row_sums = conf_mx.sum(axis=1, keepdims=True)\n",
    "\n",
    "norm_conf_mx = conf_mx / row_sums\n",
    "\n",
    "np.fill_diagonal(norm_conf_mx, 0)\n",
    "\n",
    "from matplotlib.pyplot import figure\n",
    "figure(num=None, figsize=(20, 10), dpi=200, facecolor='w', edgecolor='k')\n",
    "\n",
    "plt.matshow(norm_conf_mx, cmap=plt.cm.gray)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "cl_a, cl_b = 3, 5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_aa = X_train[(y_train == cl_a) & (y_train_pred == cl_a)]\n",
    "X_ab = X_train[(y_train == cl_a) & (y_train_pred == cl_b)]\n",
    "X_ba = X_train[(y_train == cl_b) & (y_train_pred == cl_a)]\n",
    "X_bb = X_train[(y_train == cl_b) & (y_train_pred == cl_b)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_digits(instances, images_per_row=10, **options):\n",
    "    size = 28\n",
    "    images_per_row = min(len(instances), images_per_row)\n",
    "    images = [instance.reshape(size,size) for instance in instances]\n",
    "    n_rows = (len(instances) - 1) // images_per_row + 1\n",
    "    row_images = []\n",
    "    n_empty = n_rows * images_per_row - len(instances)\n",
    "    images.append(np.zeros((size, size * n_empty)))\n",
    "    for row in range(n_rows):\n",
    "        rimages = images[row * images_per_row : (row + 1) * images_per_row]\n",
    "        row_images.append(np.concatenate(rimages, axis=1))\n",
    "    image = np.concatenate(row_images, axis=0)\n",
    "    plt.imshow(image, cmap = matplotlib.cm.binary, **options)\n",
    "    plt.axis(\"off\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.figure(figsize=(8,8))\n",
    "plt.subplot(221); plot_digits(X_aa[:25], images_per_row=5)\n",
    "plt.subplot(222); plot_digits(X_ab[:25], images_per_row=5)\n",
    "plt.subplot(223); plot_digits(X_ba[:25], images_per_row=5)\n",
    "plt.subplot(224); plot_digits(X_bb[:25], images_per_row=5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The two 5X5 blocks on the left show digits classified as 3s and the two 5X5 blocks on the right show images classified as 5s."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Multilabel Classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.neighbors import KNeighborsClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train_large = (y_train >= 7)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train_odd = (y_train % 2 == 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_multilabel = np.c_[y_train_large, y_train_odd]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# we have mixed two categories in one\n",
    "y_multilabel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "knn_clf = KNeighborsClassifier()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "knn_clf.fit(X_train, y_multilabel)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "knn_clf.predict([some_digit])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train_knn_pred = cross_val_predict(knn_clf, X_train, y_train, cv=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "f1_score(y_train, y_train_knn_pred, average=\"macro\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It takes long time to do this KNN training on my old computer"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Multioutput Classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# creates an array of 60000 X 784 with random numbers between 0,100\n",
    "noise = np.random.randint(0, 100, (len(X_train), 784))\n",
    "\n",
    "# we add this noise to our training dataset\n",
    "X_train_mod = X_train + noise\n",
    "\n",
    "# we repeat for our test dataset\n",
    "noise = np.random.randint(0, 100, (len(X_test), 784))\n",
    "X_test_mod = X_test + noise\n",
    "\n",
    "# our labels will be our original training and test dataset\n",
    "y_train_mod = X_train\n",
    "y_test_mod = X_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "PROJECT_ROOT_DIR = \".\"\n",
    "CHAPTER_ID = \"classification\"\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def save_fig(fig_id, tight_layout=True):\n",
    "    path = os.path.join(PROJECT_ROOT_DIR, \"images\", fig_id + \".png\")\n",
    "    print(\"Saving figure\", fig_id)\n",
    "    if tight_layout:\n",
    "        plt.tight_layout()\n",
    "    plt.savefig(path, format='png', dpi=300)\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_digit(data):\n",
    "    image = data.reshape(28, 28)\n",
    "    plt.imshow(image, cmap = matplotlib.cm.binary,\n",
    "               interpolation=\"nearest\")\n",
    "    plt.axis(\"off\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "some_index = 5500\n",
    "plt.subplot(121); plot_digit(X_test_mod[some_index])\n",
    "plt.subplot(122); plot_digit(y_test_mod[some_index])\n",
    "save_fig(\"noisy_digit_example_plot\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "knn_clf.fit(X_train_mod, y_train_mod)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "clean_digit = knn_clf.predict([X_test_mod[some_index]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plot_digit(clean_digitn_digit)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.7.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}