23-EX-CNN-v1.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Creative Commons CC BY 4.0 Lynd Bacon & Associates, Ltd. Not warranted to be suitable for any particular purpose. (You're on your own!)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Image Classification with a CNN\n",
    "\n",
    "In the following we're going to train CNN's to predict digit images using the famous [MNIST dataset](https://en.wikipedia.org/wiki/MNIST_database).  This [Wikipedia Page](https://en.wikipedia.org/wiki/MNIST_database) includes results from applying different kinds of ML algorithms to this problem. For more information, see [Yann LeCun's MNIST](http://yann.lecun.com/exdb/mnist/index.html) web page."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Packages"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "from IPython.core.interactiveshell import InteractiveShell\n",
    "InteractiveShell.ast_node_interactivity = \"all\"\n",
    "from IPython.display import display, Markdown\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import os\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import scikitplot as skplt\n",
    "from sklearn import linear_model\n",
    "from sklearn.model_selection import KFold, StratifiedShuffleSplit\n",
    "from sklearn.model_selection import cross_val_predict, train_test_split\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn import preprocessing\n",
    "from sklearn.metrics import accuracy_score, confusion_matrix, roc_auc_score\n",
    "from sklearn.metrics import roc_curve, precision_score, recall_score\n",
    "from sklearn.metrics import classification_report\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras.datasets import mnist\n",
    "%matplotlib inline\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Getting and Munging the Data\n",
    "\n",
    "The data are already divided into train and test splits. 60,000 training, 10,000 test."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "(X_train, y_train),(X_test, y_test) = mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(60000, 28, 28)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "(60000,)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "(10000, 28, 28)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "(10000,)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape\n",
    "y_train.shape\n",
    "X_test.shape\n",
    "y_test.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Each X record is a 28 X 28 array. The elements are grayscale values ranging from 0 to 255.\n",
    "\n",
    "In order speed training, we're going to use just one 10<sup>th</sup> of the data.\n",
    "\n",
    "It's common practice to rescale the pixel values.  Here we'll rescale them to be in the range from 0 to 1."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train10=X_train[:6001,:]/255\n",
    "y_train10=y_train[:6001]\n",
    "X_test10=X_test[:6001,:]/255\n",
    "y_test10=y_test[:6001]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## What Do the Pixel Data \"Look\" like?\n",
    "\n",
    "Here's the 101<sup>th</sup> digit image in the training data, and the corresponding label:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "oneOone_image=X_train10[100].reshape(28,28)\n",
    "plt.imshow(oneOone_image,cmap=\"binary\")\n",
    "plt.axis('off')\n",
    "plt.show();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The corresponding label is  5\n"
     ]
    }
   ],
   "source": [
    "print(\"The corresponding label is \",y_train10[100])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(99)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Fully Connected NN Classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "keras.backend.clear_session()  # this resets\n",
    "mod1=keras.models.Sequential([\n",
    "    keras.layers.Flatten(input_shape=(28,28)),\n",
    "    keras.layers.Dense(200,activation='relu'),\n",
    "    keras.layers.Dense(100,activation='relu'),\n",
    "    keras.layers.Dense(10,activation='softmax')\n",
    "    ])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Model Summary \n",
    "\n",
    "Note the number of parameters."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "flatten (Flatten)            (None, 784)               0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 200)               157000    \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 100)               20100     \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                1010      \n",
      "=================================================================\n",
      "Total params: 178,110\n",
      "Trainable params: 178,110\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "mod1.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Initial Weights for a Layer\n",
    "\n",
    "For example, for the 3<sup>rd</sup> layer.  The first array is the connection weights, and the second are for the biases."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[array([[ 0.00147645,  0.14065756,  0.07329823, ..., -0.03142992,\n",
       "         -0.02209172,  0.03629407],\n",
       "        [ 0.06085843,  0.02788216,  0.10406539, ..., -0.12599929,\n",
       "          0.09513302, -0.13601147],\n",
       "        [-0.06026858,  0.0214259 , -0.08136399, ...,  0.09297666,\n",
       "          0.03103964,  0.00906086],\n",
       "        ...,\n",
       "        [-0.11357392, -0.06161104,  0.03145878, ..., -0.13660033,\n",
       "         -0.08003778, -0.0528652 ],\n",
       "        [-0.0884479 ,  0.06783147, -0.0064802 , ...,  0.11409123,\n",
       "          0.13330792,  0.00723165],\n",
       "        [-0.08422725, -0.06671219, -0.11176863, ..., -0.08725052,\n",
       "          0.05387481, -0.13847952]], dtype=float32),\n",
       " array([0., 0., 0., 0., 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., 0., 0., 0., 0.,\n",
       "        0., 0., 0., 0., 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., 0., 0., 0., 0.,\n",
       "        0., 0., 0., 0., 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., 0., 0.],\n",
       "       dtype=float32)]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mod1.layers[2].get_weights()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Compiling Network\n",
    "\n",
    "_Accuracy_, here, will be the classification accuracy.  \"sparse_categorical_crossentropy\" is used here because the target variable labels are single digits. \"categorical_crossentropy\" would be used instead if the labels were indicate by a 10 element 0/1 array, i.e. if they were \"one hot\" encoded."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "mod1.compile(loss='sparse_categorical_crossentropy',\n",
    "            optimizer='sgd',\n",
    "            metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "mod1Training=mod1.fit(X_train10, y_train10, epochs=50,\n",
    "                     validation_data=(X_test10,y_test10),\n",
    "                     verbose=0)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "mod1historyDF=pd.DataFrame(mod1Training.history)\n",
    "mod1historyDF.plot(figsize=(8,5))\n",
    "plt.grid(True)\n",
    "plt.gca().set_ylim(0,1)\n",
    "plt.title(\"Training and Test Metrics by Epoch\")\n",
    "plt.xlabel('Epoch Number')\n",
    "plt.show();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "val_loss    0.268667\n",
       "val_acc     0.922013\n",
       "loss        0.071022\n",
       "acc         0.985502\n",
       "Name: 47, dtype: float64"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mod1historyDF.iloc[mod1historyDF.val_acc.idxmax(),:]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Recall that the curves for training should be shifted to the left.  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "More epochs."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "mod1Training2=mod1.fit(X_train10, y_train10, epochs=100,\n",
    "                       initial_epoch=51,\n",
    "                     validation_data=(X_test10,y_test10),\n",
    "                     verbose=0)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeMAAAFNCAYAAAAzYQemAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAgAElEQVR4nO3deZwldX3v/9fn1Nm6p2eG2VkGYUAcFIYBWUSjQys3gkbFKBEUEbkqIa4xiRf3EJcYxeVKJJK5/giiRCCoESPBq0IzkKAMeEGEERwGYYZ19plezv75/VF1TlefPt19enqpmZ738/Gorqpv1an6nu+pqs+3vlVdZe6OiIiIJCeVdAZERET2dwrGIiIiCVMwFhERSZiCsYiISMIUjEVERBKmYCwiIpIwBWPZa5lZYGa9Zva8yZw3SWb2fDPbr/+f0MzWmNmbJ3F5N5rZRydredPJzI41s0LS+ZDkKRjLpImCYb2rmdlAbPy88S7P3avu3uXuT0zmvHujWGVipPI7ZwLLvsfMzh1l+rFm5ma2pin9UDOrmtlv21zPl83syrHmc/dV7v79dpY5ncysKyqHvqbf4r1J501mvnTSGZCZw9276sNm9gfg3e7+85HmN7O0u1emI297O3evAvHy2wS83d17pikLJeBQM1vm7o9FaecD6ydrBWaWAtz3/icNHenuzySdCdm/6MxYpo2Zfc7Mrjez75nZbuDtZvZSM/ulme0ws6fN7HIzy0Tzp6MzlcOj8e9G0//TzHab2V1mtmy880bTX2Nmj5jZTjP7RzP7LzN75wj5biePf25m681su5ldHvtsYGZfM7OtZvYocOYEyi9tZpea2WNmtsXMvmNmc6JpXWZ2g5lti/LwSzOba2b/GzgBuDo6y/viCIt34FrgHbG084FrmvJwmJn9OFr/o2b2nij9zcAHgXdF6/nvKP2eKM93A/3AkviZuoU+YGYPR7/Tb8zsRdG0S6Py3mVm68zsZaMUz0Fmdnu0jJ+Z2cHRMr5tZn/X9B1uM7N3j13iQ0Vn/t81s3+P1vMrM3thbPpKM7sz2k7uN7NXx6Z1mdk3zGxjNL0nqpzUp7/bzJ40s+fM7MPjzZvMAO6uTt2kd8AfgP/RlPY5wjOw1xNWBDuAk4GXELbSHAE8Arw/mj9NGCQOj8a/C2wBTgIywPXAd/dg3sXAbuCsaNpfAWXgnSN8l3by+CNgLnA4sK3+3YH3Aw8CS4EFwJpwtxuz/DYB3U1pnwR6gAOjsvsO8H+iaX8dfcd8lKdTgI5o2j3AuaOs61igALwA+H2UdgpwL/BG4Lex7/pQtK4McHSUz5dH078MXNm07HsIz66PArJAEM8PcCGwAVgJWLTMQ4ATgd8Di6L0I4HDRsj/jVGZvyT6/t8CbommvQpYH5v3UKAPmNtiOV3Rb3ngCOv5MlAE/iT6/n8blUd9W94EfCia9lqgt55n4NvAfwJLonJcFX2vY4EacDmQA04l3Edafld1M7fTmbFMtzvd/cfuXnP3AXdf6+6/cveKu28AVgOnjfL5G939HncvE57JHb8H874OuM/dfxRN+xph4G6pzTx+wd13uvsfCANmfV1vAb7m7pvcfSvwD6Pkdyx/Dlzi7s+4+wDwGaB+LbhMGLiOiPJ5dzRP29z9EWCrmf0R4RnyNU2znAaYu3/F3cvu/jvCIDPW9ezV7v57dy952Bwf927g8+5+v4d+5+5PAhWgEzgGCNz9UXd/fJR1/CD6jQrAx4AzzGw+cBuQMbOXRvOdB/zE3XeOsqyHo7PXevfy2LQ73P0n0Xbz98DBhBWJbqDk7l+PyubmaN1/ZmZ54O2EFbhno99njbvXm+sN+LS7F939l4SVlxWj5E9mIF0zlum2MT5iZkcDXyE8E+ok3CZ/Ncrn49fy+oldZx3HvAfH8+HubuE12pbazGNb6wJGCygjMrOA8Izx/9rQu7EzZnYAYQVhMfBDM+skDJKfdvfaOFd1DfAu4DWEQSbeNHwY8Hwz2xFLC4BbxljmxlGmHQo82pzo7veb2SeBLwDLzewnwF+5++ax1uHumy28Q/kgd99mZt8hDIZ3Rf2PjZHf5T7yNeP4espm9jThb3wgw3/bxwl/s4MJA+5jtFZ093iZjrVdywykM2OZbs037/wz8Fvg+e4+B/g04YFrKj1N2GwMhNctCQ+aI5lIHp8mDDh1e/SvV9EZ5dPAKnc/INbl3X2Huxfc/ZPuvhx4JfA24M/qHx/Hqq6LPnuPuz/XNG0jYZN1fP2z3X2s9Yy2/o2ETdDDP+T+L+7+0mj6bMKWgJE0ytjMFhE2Vz8dJV0DvMXMTiFsJh6r8jCa+HrSwEHAU1HX/Ns+D3gy6hxYhsgIFIwlabOBnUBfdDPMn0/DOv8DeLGZvT46oH6IsIl3KvJ4A/CXZnaImS0ALtnTTANXAl80s0MAzGyJmb0uGv5jM3thdFPQLsJm3nqT8LOE17rH5O7bCJuj399i8u1Azszeb2a56IaylWZWb5J/FlgWVW7a9S3g42Z2XHQz19FRWR1rZqvMLEd4pliIfZ9W3mRmJ0fzfx74WfRd6s3v66N1fS9qYt5Tr7Dw5r8M4Rn208D9hGWTj25GS5vZGYTXq2909yLhPQyXm9liC2/qWzXOcpIZTsFYkvbXwAWEN1T9M+FNSFPK3Z8lvM75VWAr4ZnX/yO8OWey8/hN4BfAA8BawpuN9tQXCG8Au93MdgF3Et4pDeEZ24+jPN5PeEPZD6JpXyG8y3mHmX1hrJVE116HNae7e4mw+bobeAJ4DriCwSbVfwUOALaZ2R1tfqergX8Evk9YibgemEN4Q9TXCH+fpwlvbrp0lOV8h/AGqy2Ev+f/bJr+bcLrsN9pI0+P2tD/M/58bNq/EVbGtgNvAM6O7n/oJ7wX4a1Rnr8M/Fl0DwHA+wjL7P5o+qVMfQuQ7ENs8B4Ckf1TdD32KcIDa7tBRPYhZvZa4KvufvQElvFloMvdL568nImEdGYs+yUzO9PC/8PNAZ8ibNa9O+FsyRSIfuMPEN7kJrJXGjMYm9lV0T+it3wkXnSd53ILH3jwGzN78eRnU2TSvZzw/1u3ED6I443RtT2ZQczsJML/Qc4TXjIQ2SuN2UxtZqsI/3n9Gnc/tsX01xLWOl9L+E/3X3f3l0xBXkVERGakMc+M3X0NYc1yJGcRBmqP/mH9ADM7aLIyKCIiMtNNxjXjQxj6T/2bGP1/NkVERCRmMp7A1er2/JZt32Z2EXARQD6fP/F5z9urXz0749RqNVKpGXDPnju4Y7VaY7jR1Y31L5zxed3Djbie1jRttM+NtXx3Z9i/k8Y+Z81p3vjTOj/1T8UX2eq7jrq85pU3i5YfW7/F8zJivtoQz8+w9BGmjfA5a+dzE6V/NpG2I1x7HiwWtrj7sOcaTEYw3sTQJwwtJfw3kWHcfTXRHY3Lly/3hx9+eBJWP/W8VsMrFbxUxsulqB8Nl8uD6eUovVRq70AVHdi8Vgt/XHfAY+k+JG3IfO7gNbw+XHO8VoWaQ60azluthfNE/cc2bGDZstEfAuS1Gl4s4cUitWIBL0T9YgkvFKgVi3ixCLWxn7LohPkant/oO0TBdPh8Pvi9ovlqpRI+MBCWbX3nCMYu4r2SWdilUmAWfp10GgsCLAig3o/SCFJYKgg/U62GZVOt4h79xrVa+HvXKyctlxWEy0gHNF4W1FzOPri9ec3DZaTTkElj6QyWTje6RlqqzX+VtVTje4efsVg5ROkWpadS4W8cT4vNO3S+sOJgqVRjvsefeILDDjtscn6qIBWWZ/z7Z9JD04I2K7ixPDbybDaksyHjqWj2WBqxMqxvQ7RYXlibGtzHWvzOXvOWn2ssu1U+6vM2zXf32rWccsopk1Lmk6vV92nKfyrA0rF9LpUa3PdSqeGV6YjHjk+NMoUh463KPn3AAS0fiTsZwfgm4P1mdh3hDVw73f3pMT6DFYv03X03VCphoKtU8HIFr5RjadEDd1LxjTS2A0Ybh1eqg58rx5YXTyuVqJWKeKGIFwvUGsElFmhKxVigjQXWysx45W4XMNKDfYdIp0lls1g+j+VzpHJ5LJcjlcth+TypWZ3hgb0d9YDT4gDc8qA65DcenM+yOVIdeSzfEfXzpIYM5yFID6/QRBWVRhpEASo6iAZpLD0YtBoBLJWCIABLhfOlUmFaqmm4vmMbse9hjSB7+x13cFp399CDrUyph3p6WNzdnXQ29ivVjRvJHdHWQ95mjEZMirU2TmTvHjMYm9n3CJ+4szB6mP7fEr4iDHe/EriZ8E7q9YSPrbuwrRU/8yxPvOOCPcv1eGUyWCYzGExyOSyXawSa1AEHkMrnsGwOy2axTGawX+/q4+n08Hmy9fmyjWHabA4eVpttUQNtVYseWoO2wQARBOG0FoHk9jVrOO200V6IRLjsYF895dwL1QO8iMgoxgzG7v7WMaY74aPexqWyeDHPu/pfBpu90hkskx7aFJZO11cSa24BPNYsUKsN+0wjaKbTg8FJwrJI60VdIiJ7m8SOzN6RZ9appya1ehERkb3GDLi1VkREZN+mYCwiIpIwBWMREZGEKRiLiIgkTMFYREQkYQrGIiIiCVMwFhERSZiCsYiISMIUjEVERBKmYCwiIpIwBWMREZGEKRiLiIgkTMFYREQkYXqf3kznDn1bYOdGDtj+AGw+GLoWQ35u9L5k2afValAZgPIAlPvDtNxsyM2B1DS/R9kdqqUwL5VC2JULUKtAkIV0brALov548uge9ie63VaKUNgF6SxkZkEwhYfBWhUGdkBxV/Td84Ndm+883+u5g1ehWglfb4tHv1XUb6TVwrKvbxeVgahf31YGwum1cvSa3FpYfvXh+Dg+dr4sBdlZ4b6Q7Yr2i9mQ6wrTcrPD3wTCvFcK4fZbKUT5LEI16teqsffLx981H3sXPR7OW9/+69+n8T2j8REkFoyD6gD84c7Yj1Zr+gGjtHIfFHuh1Ael3WG/edwCyHRAprOp3zSczg/203nI5CHdEf4gmY7wgAHRj03rvNWqMLAdep+F3ueg77mwXx+vD1fL4YaQ7Qx3+MZwZzQcdfkDoGMedET9+Hhu7tAdtlaLNo4CVGIbTbUIfZth56ZYtzHqPxlOB44HuP+T4bLSHWFQnn0gdC0Z7M9aGOa9HDvAD+nHDvqpdNQFUReNWzRuFu1A9Z2p2mIH82iDTkWfSYWfb4xbOB4/kNV/t0zswJbpCJdX6g+3mUa/b2hapTh4MI/vUM072ZADStO2Wd82zAa/twUty+LIJ5+E/ptbfH9vGq+GQas2Ur8S7ej90Y7eP7jTjyQbHXTyc4b2s52Dv3HLg0a0jXlt5ANQPa1WGXpQbecgGZdKh79fkI32r1bbSv0ATDhffm7UHRAbnhvuM/m5LN34BPziDhjYBv3bwv11YBv0R/369lsX5Frsp1G/ETyzUT7jFYoorVaB/q1Rty02vDVc90hlEmSHb8dmI2x3seF2OC2W0RwgGdy/mve3+P7otXB7qVWifjkMXrVyY7wb4Pb2f/a9RiozdPtKkLmPc+eZJCcdHPg9F3WN/4OpTFizyda7zihotwgY01XAmc4wsHUtgVmLwn6QjQWF/igo9A0Ol/uhuHv0gykWHkBrtXC+WrmNzBjMPgjmLo26Q2DuoTB3KfetW8/xRx4Mvc/A7mfCSkOj/ywUdw5fXJAdXtFJ5wcPxPWA4dWh4/WA2wiwqdYBt37wGRaoo4BVX3Y9QIxaXi2k0uEBNn5whaYae4tafKPWm2JoAIr13cPvOySQVoekVcol0pns8ANcY7xeLrGKjcUqNql0NE8Qq2C2qnhGfffwLKywK9bfGW5r9bRS/2BwaVRQc7HA0BFOt9QIZzgMpqVSTZ/LDa/wptLhQbv+GzbOPkpDx+PbyUjbS2UACjuHdgM7ouEdYdlDOG/HPOiYD53zm4ajim612FR564dS79B9dqQ8N++LqQx0Loi6+bHhqMvNDj8z2hlhZWBo5XTUbbANjc/R9PlYxcprLfa3psqQpSDIhN8xSEf9TPi7RumPPbGRZcuOHH1dlhqsVDdOjuoVkVilJJUefV+xoL0yqFWjk7fd4Ylbsd71Dk2zoKnClR1eAUvF9wWGHjvix4xW23/8uwU5LAjudfeTmrOb2Jlxf+ch8I7vjLCxpQbTG2eRUfBNZ9tbQaPJLBag480jjeHYDlEtNeWjnjeG5rNjXhhw6wE4tweVirpyITyIDOwIa9GFqD+wYzA9lR59Q0lnoXNhGHxnHzRiGe14pgeO6x45L6X+8MyhHoDTHVPbhLcn3AcPXPEmoXqQznZFrQ/R2U2728sUubOnh+7u7kTzsN9wh/IAd95+Ky8//bVT2wzcaKUqhseF3Oz2g+QM9HhPD8tO6046G8Pl5wAHJZ2LtiR2pK0GHXDEaVO3ArPB5qSOeVO3nonK5CFzYNhMnLRsFMT2ZmZRmeWTzonsbcwg20kl0zX112NTKUhFrRQik2CG3EEgIiKy71IwFhERSZiCsYiISMIUjEVERBKmYCwiIpIwBWMREZGEKRiLiIgkTMFYREQkYQrGIiIiCVMwFhERSZiCsYiISMIUjEVERBKmYCwiIpIwBWMREZGEKRiLiIgkTMFYREQkYQrGIiIiCVMwFhERSZiCsYiISMIUjEVERBKmYCwiIpIwBWMREZGEKRiLiIgkTMFYREQkYQrGIiIiCVMwFhERSZiCsYiISMLaCsZmdqaZPWxm683soy2mzzWzH5vZ/Wb2oJldOPlZFRERmZnGDMZmFgBXAK8BXgS81cxe1DTb+4CH3H0l0A18xcyyk5xXERGRGamdM+NTgPXuvsHdS8B1wFlN8zgw28wM6AK2AZVJzamIiMgMlW5jnkOAjbHxTcBLmub5BnAT8BQwGzjH3WvNCzKzi4CLABYtWkRPT88eZFn2VG9vr8p8mqnMp5/KfPqpzCeunWBsLdK8afwM4D7gVcCRwM/M7A533zXkQ+6rgdUAy5cv9+7u7nFnWPZcT08PKvPppTKffirz6acyn7h2mqk3AYfGxpcSngHHXQj8wEPrgceAoycniyIiIjNbO8F4LXCUmS2Lbso6l7BJOu4J4HQAM1sCLAc2TGZGRUREZqoxm6ndvWJm7wd+CgTAVe7+oJldHE2/EvgscLWZPUDYrH2Ju2+ZwnyLiIjMGO1cM8bdbwZubkq7Mjb8FPDqyc2aiIjI/kFP4BIREUmYgrGIiEjCFIxFREQSpmAsIiKSMAVjERGRhCkYi4iIJEzBWEREJGEKxiIiIglTMBYREUmYgrGIiEjCFIxFREQSpmAsIiKSMAVjERGRhCkYi4iIJEzBWEREJGEKxiIiIglTMBYREUmYgrGIiEjCFIxFREQSpmAsIiKSMAVjERGRhCkYi4iIJEzBWEREJGEKxiIiIglTMBYREUmYgrGIiEjCFIxFREQSpmAsIiKSMAVjERGRhCkYi4iIJEzBWEREJGEKxiIiIglTMBYREUmYgrGIiEjCFIxFREQSpmAsIiKSMAVjERGRhCkYi4iIJEzBWEREJGEKxiIiIglTMBYREUmYgrGIiEjCFIxFREQSlk46AyIism8ol8ts2rSJQqEwJH3u3LmsW7cuoVztnfL5PEuXLiWTybQ1v4KxiIi0ZdOmTcyePZvDDz8cM2uk7969m9mzZyeYs72Lu7N161Y2bdrEsmXL2vpMW83UZnammT1sZuvN7KMjzNNtZveZ2YNmdvs48i0iIvuAQqHAggULhgRiGc7MWLBgwbAWhNGMeWZsZgFwBfDHwCZgrZnd5O4PxeY5APgn4Ex3f8LMFo879yIistdTIG7PeMupnTPjU4D17r7B3UvAdcBZTfO8DfiBuz8B4O7PjSsXIiIi+7F2gvEhwMbY+KYoLe4FwDwz6zGze83sHZOVQRERkZmunRu4Wp1re4vlnAicDnQAd5nZL939kSELMrsIuAhg0aJF9PT0jDvDsud6e3tV5tNMZT79VOZTZ+7cuezevXtYerVabZm+NzjooIN4+umnW057/PHHectb3sKvfvWrKVl3oVBoe1tsJxhvAg6NjS8FnmoxzxZ37wP6zGwNsBIYEozdfTWwGmD58uXe3d3dViZlcvT09KAyn14q8+mnMp8669ata3nX9N5+N/VIeevq6iKVSk1Z3vP5PCeccEJb87YTjNcCR5nZMuBJ4FzCa8RxPwK+YWZpIAu8BPha2zkWEZF9yt/9+EEeemoXEJ4ZB0Ew4WW+6OA5/O3rjxl1nksuuYTDDjuM9773vQBceumlmBlr1qxh+/btlMtlPve5z3HWWc23No2uUCjwF3/xF9xzzz2k02m++tWv8spXvpIHH3yQCy+8kFKpRK1W4/vf/z4HH3wwb3nLW9i0aRPVapVPfepTnHPOOXv8vaGNYOzuFTN7P/BTIACucvcHzeziaPqV7r7OzG4BfgPUgG+5+28nlDMREZEm5557Ln/5l3/ZCMY33HADt9xyCx/+8IeZM2cOW7Zs4dRTT+UNb3jDuO5ovuKKKwB44IEH+N3vfserX/1qHnnkEa688ko+9KEPcd5551EqlahWq9x8880cfPDB/OQnPwFg586dE/5ebT30w91vBm5uSruyafwy4LIJ50hERPZ68TPY6WymPuGEE3juued46qmn2Lx5M/PmzeOggw7iwx/+MGvWrCGVSvHkk0/y7LPPcuCBB7a93DvvvJMPfOADABx99NEcdthhPPLII7z0pS/l85//PJs2beJNb3oTRx11FCtWrOBv/uZvuOSSS3jd617HK17xigl/Lz2bWkRE9ilnn302N954I9dffz3nnnsu1157LZs3b+bee+/lvvvuY8mSJeN64AaET81q5W1vexs33XQTHR0dnHHGGdx666284AUv4N5772XFihV87GMf4zOf+cyEv5MehykiIvuUc889l/e85z1s2bKF22+/nRtuuIHFixeTyWS47bbbePzxx8e9zFWrVnHttdfyqle9ikceeYQnnniC5cuXs2HDBo444gg++MEPsmHDBn7zm99w9NFHM3/+fN7+9rfT1dXF1VdfPeHvpGAsIiL7lGOOOYbdu3dzyCGHcNBBB3Heeefx+te/npNOOonjjz+eo48+etzLfO9738vFF1/MihUrSKfTXH311eRyOa6//nq++93vkslkOPDAA/n0pz/N2rVr+chHPkIqlSKTyfDNb35zwt/JRjo1n2rLly/3hx9+OJF176/0Lx/TT2U+/VTmU2fdunW88IUvHJa+t/9rU1JalZeZ3evuJzXPq2vGIiIiCVMztYiIzGgPPPAA559//pC0XC43ZU/e2hMKxiIiMqOtWLGC++67L+lsjErN1CIiIglTMBYREUmYgrGIiEjCFIxFREQSpmAsIiIzVldXV9JZaIuCsYiISML0r00iIjJ+//lReOYBADqqFQgmIZwcuAJe8w+jzjKZ7zPu7e3lrLPOavm5a665hi9/+cuYGccddxzf+c53ePbZZ7n44ovZsGEDAN/85jd52cteNsEvHVIwFhGRfcZkvs84n8/zwx/+cNjnHnroIT7/+c/zX//1XyxcuJBt27YB8MEPfpDTTjuNH/7wh1SrVXp7eyfteykYi4jI+MXOYAf20fcZuzsf//jHh33u1ltv5eyzz2bhwoUAzJ8/H4Bbb72Va665BoAgCJg7d+6kfS8FYxER2afU32f8zDPPDHufcSaT4fDDD2/rfcYjfc7dxzyrnmy6gUtERPYp5557Ltdddx033ngjZ599Njt37tyj9xmP9LnTTz+dG264ga1btwI0mqlPP/30xusSq9Uqu3btmrTvpGAsIiL7lFbvM77nnns46aSTuPbaa9t+n/FInzvmmGP4xCc+wWmnncbKlSv5q7/6KwC+/vWvc9ttt7FixQpOPPFEHnzwwUn7TmqmFhGRfc4DDzzQGF64cCF33XVXy/lGu8lqtM9dcMEFXHDBBUPSlixZwo9+9KM9yO3YdGYsIiKSMJ0Zi4jIjKb3GYuIiCRM7zMWERGRMSkYi4iIJEzBWEREJGEKxiIiss/YV16JOF4KxiIiIglTMBYRkX2Ou/ORj3yEY489lhUrVnD99dcD8PTTT7Nq1SqOP/54jj32WO644w6q1SrvfOc7G/N+7WtfSzj3w+lfm0REZNy+ePcX+d223wHhc5qDIJjwMo+efzSXnHJJW/P+4Ac/4L777uP+++9ny5YtnHzyyaxatYp//dd/5YwzzuATn/gE1WqV/v5+7rvvPp588kl++9vfArBjx44J53Wy6cxYRET2OXfeeSdvfetbCYKAJUuWcNppp7F27VpOPvlk/uVf/oVLL72UBx54gNmzZ3PEEUewYcMGPvCBD3DLLbcwZ86cpLM/jM6MRURk3OJnsLun8X3Gde7eMn3VqlWsWbOGn/zkJ5x//vl85CMf4R3veAf3338/P/3pT7niiiu44YYbuOqqq6Y1v2PRmbGIiOxzVq1axfXXX0+1WmXz5s2sWbOGU045hccff5zFixfznve8h3e96138+te/ZsuWLdRqNd785jfz2c9+ll//+tdJZ38YnRmLiMg+50//9E+56667WLlyJWbGl770JQ488EC+/e1vc9lll5HJZOjq6uKaa67hySef5MILL6RWqwHwhS98IeHcD6dgLCIi+4z6KxHNjMsuu4zLLrtsyPRWrz4E9sqz4Tg1U4uIiCRMwVhERCRhCsYiIiIJUzAWERFJmIKxiIhIwhSMRUREEqZgLCIikjAFYxERkYQpGIuIyD7ljW98IyeeeCLHHHMMq1evBuCWW27hxS9+MStXruT0008HwgeEXHjhhaxYsYLjjjuO73//+0lme1R6ApeIiIzbM3//9xTXha9QrFSrbJuEVyjmXng0B37842POd9VVVzF//nwGBgY4+eSTOeuss3jPe97DmjVrWLZsGdu2bQPgs5/9LHPnzuWBBx4AYPv27RPO41Rp68zYzM40s4fNbL2ZfXSU+U42s6qZnT15WRQRERl0+eWXs3LlSk499VQ2btzI6tWrWbVqFcuWLQNg/vz5APz85z/nfe97X+Nz8+bNSyS/7RjzzNjMAuAK4I+BTcBaM7vJ3R9qMd8XgZ9ORUZFRGTvET+Dnc5XKPb09K9pe8cAABo5SURBVPDzn/+cu+66i87OTrq7u1m5ciUPP/zwsHndHTOblnxNVDtnxqcA6919g7uXgOuAs1rM9wHg+8Bzk5g/ERGRhp07dzJv3jw6Ozv53e9+xy9/+UuKxSK33347jz32GECjmfrVr3413/jGNxqf3debqQ8BNsbGN0VpDWZ2CPCnwJWTlzUREZGhzjzzTCqVCscddxyf+tSnOPXUU1m0aBGrV6/mTW96EytXruScc84B4JOf/CTbt2/n2GOPZeXKldx2220J535k5u6jz2D2Z8AZ7v7uaPx84BR3/0Bsnn8DvuLuvzSzq4H/cPcbWyzrIuAigEWLFp14ww03TNoXkbH19vbS1dWVdDb2Kyrz6acynzpz587l+c9//rD0arVKMAk3cM0069evZ+fOnUPSXvnKV97r7ic1z9vO3dSbgENj40uBp5rmOQm4LmqbXwi81swq7v7v8ZncfTWwGmD58uXe3d3dxuplsvT09KAyn14q8+mnMp8669ata3lteDqvGe9L8vk8J5xwQlvzthOM1wJHmdky4EngXOBt8RncfVl9OHZmPCQQi4iISGtjBmN3r5jZ+wnvkg6Aq9z9QTO7OJqu68QiIvuJfekO5SSNdQm4WVsP/XD3m4Gbm9JaBmF3f+e4ciAiIvuEfD7P1q1bWbBggQLyKNydrVu3ks/n2/6MnsAlIiJtWbp0KZs2bWLz5s1D0guFwrgCz/4gn8+zdOnStudXMBYRkbZkMpnGU67ienp62r5RSVrTiyJEREQSpmAsIiKSMAVjERGRhCkYi4iIJEzBWEREJGEKxiIiIglTMBYREUmYgrGIiEjCFIxFREQSpmAsIiKSMAVjERGRhCkYi4iIJEzBWEREJGEKxiIiIglTMBYREUmYgrGIiEjCFIxFREQSpmAsIiKSMAVjERGRhCkYi4iIJEzBWEREJGEKxiIiIglTMBYREUmYgrGIiEjCFIxFREQSpmAsIiKSsHTSGRAR2Rdt7yux9g/bKFZqvOzIBSzoyiWdJdmHJRaMtw44n/nxQ3RmAzqyAR2ZoGk4TWcu4ICODPM6s8zpyBCkLKnsish+7pmdBe7+wzbufmwrax/bzsPP7m5MM4NjD57LK45ayKoXLOLFz5tHNq2Gx8nm7hQrNcwglw72eBm7ixW27C6yra/UWFY2nSKXTkX9wfF0yqg59JUq9BUr9BYq7I76vbHxSrXGotk5DpyTZ/GcPAfOzdOVaz/EJhaM+yvO9WufoL9cxX3s+c1gbkeG+Z1ZDugMA/QBnVnmdWbozAZkghTpIEUmMLLpFOlUOJwJUmSCFDV3BkpV+koV+ktV+opN/VKFgVKVmjs1Z7Bf88awu+MOczszLJ6dY8mc/GB/To7Fs/MsmZOjK5fGzKjWnN5ChZ0DZXYOlNlViPrR+O5CuO6BcpVCuUp/qcJAucZAqcJAucpAqUqhXMOjAjKzYWVS78+flePIRbM4clEXRy6axRGLujhsQecebbC1mlOq1iiWaxQrVYqVWtRFw+UataYfrdVvmAmMOR2ZsMunmZVNk9qLKlTVmrO1r8iW3SU29xbZsrvIlt4i2/vLzMoGzJuVZX6smxdte5lg8g6y7k65Gm5j3tjuBre3WiwtnwnomoQyrNUcs+Hb00iqNadQrsa2ycFhgI5sQD4TVqJzmRQdmXB8rHKq1ZyqO9Va2GXTqXGVbalSY0tvsdFt3l1kS2+Jdb8vsY5HmdORZnY+w+x8mjn5cBucnc8wpyNNRybAHcq1GuWqU67Uhg5Xw23+oad2RQF4G09s6wegK5fmxYfN4w3HH8zJh88nExh3/H4Ld/x+M/+8ZgP/1PMos7IBLz1yAatesIhVRy3i8IWzhuS7rxgeyBsH+GKV/mKFUjXct6q1wfKp1Dwcjo5FYT/aLmLD7uH84bYDKQuPDSkzzIxUNJyKfnuzcB2V2uBvUG0ar9S8cdxzwp28vq97bPiZZ4r8x+b7MerrG1xHPC2+bVeH5J3G9yuUawyUo2NjdHysD/eXKtSideYzKeZ2ZGJdtjF8QGeGjkzAtv5SY7/e2hcN95UoVWptb2cpo7HO8erKpVk8JwzQS6JuJOatjqLTYPny5f7www83ajoDpSr90Q5eL/T+UpUdAyW295XZ0V9iW3+J7f3h8Pa+Mtv7S2zvL1Eot1+wdemUMSuXZlY2oDOXpjM6oKRT1thw6htukBrcmAG295d5bleBZ3cVGShXhy27IxOQDozeYmXUikbKwnk7smk6sik6M2ny2YDOTNRCkA3IpwNSBvXFDO4I3kisufPc7iIbNvfxzK7CkOUfOr+TIxd1ccTCWezavIklhxzG7kIl6sIKwe5iuZHWW6yMa0Mdj5TROBjOyYc7zaxcOizzlBGYRWVNY7ieXqnVKFWccjU8UJYqNUrRcLnqlCo1HI9+s+iz0W8YRL9pKmVUazW29pbYvLvItv7SiJWIcnXkH25OPs38WVnymWBw+SkjiLaVeB62b99GR9fcRmWmUB7aL1ZqbVVG68ygK5tmdn4w0HTFhtMpo68YVjLrB/r+UpXeWL/++6Zi+Q2ick6lBsurVKlSqNT2eHtIp4yOTEAmnYod3Guxg/wIn4lax+r9eqDPZ1IMlKuNoLtzoNxyve0cPM1aVyBbmT8ry8mHz+Pkw+fzkmULeOFBs0mPUGnYVShz16NbWfPIZtb8fjMbtw0AsLArS7Xm9BWrlKpTs3/Vj1lBVMlyhlbqxvpsOvrt06kUKYN0kIq2jWgerDGvNT4XDhUKBXK5XCNA19yj4XogD9OCpopBfX9PxdKy6RSd2bB1tCMbRMMBHZl0o/UUYOdAGAvqJzs7ByrsjMb7oopiOmUs6MqysCvHwq4cC7qyLIoNz5+VxcwoRftnKTrxaB43M+bk03Tlwv2tKxfug125TGM8ZbB5d5FndhV4blfYf7bRFXlmZ4HndhdY//d/cq+7nzTsN0g6GE+Gei2rXA3P6CrRAbp+4K5EW2Ij+GbTk9KE5O70Fis8t7vIs7sKbN5d5Lld4XCl5syJamlz8ulGjW1OrD8rG7R9dtKu3mKFxzb38ejmXjZs7uXRaPixLX2N5p2uXJrZucEDePOBPZ8Oz3By6YBc1FSTywwO11semjV/lVKlxq6oRWDXQCXql9lVqDRaB3qLlfBAHavlDz0DCNOClJENUtHZU9jiUT+TykYtIimzxplWvaZdrZ8x1JyqQ2CwINoZF83Osai+o84eTJuVDShXvVEB3NYXVv629ZfY3heOb+srUShXh52p1INOfXjXrl0sXjCPfKz8GsOZgHxUnqlGEG99JoMZhVI1rEAVmypTseFKzenKpZmVC7fzrqii2ZULL/vMyoW/rzN4VtQq3zV3skGKfD0wRsGxHhjr40DjTDk8o6lSLA+e0RTKNUrVKulUKjrQW6OfaoyHB/9SJfx8vaWovoyBco1CNNyRCVg4O/zNFjX9bguj3/KuO9fwkj96RaNcdg4MLatdhTJ9xQpByoZsP+n6cDpMT6dSPH9x2Nq0J/upu/P41n7u+P1mfrNpJ/lMWP5d0e8wK5sO+7no94mOS4MVI8LAmBpeOY0HsSB2JjpWfuItLR7tV/VK2ET09PTQ3d09oWVMpnI13JZmR62Ue4tazQmCVMtgPCNu4DIz0oGRDqCDPbuOsKfrDYNYhiMXdU3bekfTlUuzYulcViydOyS9VnN+emsPZ7yqe69qKt5bZdPG4ujaz0SEB6lTJylX0g4zC+85yaZHbRacjnwcvnDWkCbqJJlFrTfM/P2/fnlybzPasXfvy61MiVTK6EhPvAYsIiKTT8FYREQkYQrGIiIiCVMwFhERSZiCsYiISMIUjEVERBKmYCwiIpIwBWMREZGEKRiLiIgkTMFYREQkYQrGIiIiCVMwFhERSVhbwdjMzjSzh81svZl9tMX088zsN1H332a2cvKzKiIiMjONGYzNLACuAF4DvAh4q5m9qGm2x4DT3P044LPA6snOqIiIyEzVzpnxKcB6d9/g7iXgOuCs+Azu/t/uvj0a/SWwdHKzKSIiMnO18z7jQ4CNsfFNwEtGmf9dwH+2mmBmFwEXASxatIienp72cimTore3V2U+zVTm009lPv1U5hPXTjBu9QJcbzmj2SsJg/HLW01399VETdjLly/37u7u9nIpkyJ80X130tnYr6jMp5/KfPqpzCeunWC8CTg0Nr4UeKp5JjM7DvgW8Bp33zo52RMREZn52rlmvBY4ysyWmVkWOBe4KT6DmT0P+AFwvrs/MvnZFBERmbnGPDN294qZvR/4KRAAV7n7g2Z2cTT9SuDTwALgn8wMoOLuJ01dtkVERGaOdpqpcfebgZub0q6MDb8bePfkZk1ERGT/oCdwiYiIJEzBWEREJGEKxiIiIglTMBYREUmYgrGIiEjCFIxFREQSpmAsIiKSMAVjERGRhCkYi4iIJEzBWEREJGEKxiIiIglTMBYREUmYgrGIiEjCFIxFREQSpmAsIiKSMAVjERGRhCkYi4iIJEzBWEREJGEKxiIiIglTMBYREUmYgrGIiEjCFIxFREQSpmAsIiKSMAVjERGRhCkYi4iIJEzBWEREJGEKxiIiIglTMBYREUmYgrGIiEjCFIxFREQSlk5qxdsq2/jy2i8zOzu70c3Jzhk23JHuwMySyqaIiMiUSywYD/gANzxyAwOVgVHnM4yOdAedmU5mZWbRme5sjHemO4ek19NmZWaF6emw35nppCvTxazMLGZlZpFOJfa1RUREhkksKh2SOYS7z7ubcrXM7vJudpfCbldpV2N4d2k3/ZV++sv99JX76K/0M1AeoL/Sz47CDp6qPNVI7y/3U/VqW+vuSHc0gnNXpouubNeQYD1S15XpGhb8U6aWfhERmZjETxEzQYb5wXzm5+dPaDnuTrFapL8SBe5yf2M43vWWeukt94bD5d7G+JaBLWHwL/fTV+mj5rW21tuR7mh5ht6R7hixy6fzjeHOdCcdmdhw1M8EmQmVh4iI7DsSD8aTxczIp/Pk0/lJCeyFamFYIK939TPx+Bl7/My9ftY+UBmgUCmE/WphXHlIp9JDA3Smc9TxfDpPJpUhG2TDLpUdNryxtJEndj3RaLbPBTldjxcR2QvMmGA8mcyscea6sGPhpCyz5jUKlUIYsCsDQ7ty2I9P6y/3t0zbVtjWqAzUp1dqlbbz8aUffqkxHFgw5Hp6/Tp7Pp0nH+TJpXPkglxjOB/kw/Go0pMLcnSkOwbTgvzQfjpPNpVVwBcRGYOC8TRJWapxM9lkK1fLFKtFSrUSpWqJcrXcGI6nrb1/LUcsP2LEM/6+ch99lT62FbZRrBYZqAxQrBYpVosUKgUc36Pv3ZHuIB9ETfNRk3xHMNhsnw2y5IJcI/g3xqPhfJBvnOHHp8f79eF8kCedSqsCICL7FAXjGSATZNq6xlz6fYnuI7v3aB3uTrlWplAtUKgUKFaKjeFGWrU4ZHhYC0CsFWCgMsDO0s5GwC9VS43AP54z/VZSlmp5Rh8P9o1m/BbN+fEgH19GfLw+PZPKkE1lyQSD/Uwqoxv7RGRcFIylLWbWCFRzsnOmdF3VWnVYgC5VSxRrxUYrQD0tPk/8LL55uFANKxDFapGd5Z2DLQa1cmNZ5VqZUrXU9l35o0lbmkyQwWrG7H+b3biuH2/Crw/nghzpVJrAAgILSFmKlKUIUsGQtOYWgXjrQXNLQrwCoX/lE9n7aS+VvU6QCuhMTU2TfjsqtQqlaolCtRD2Y8E9HuDrAbwexOvD5Wq5MfzoE4+ycMnC8Ea+anhDX6FSYEdxx5Cb+2q1GlWvNrqa19q+o38saUs3zubrXTqVblQA6sNpSw9Jr5/l11sA6l06lW5Ma3RBbFpTen25gQUEqaCxrnploz6tvp5ckCNIBZPy3UX2FQrGIk3qAWkyKgM9vT10/1H3Hn3W3RuBuerV8Cy+UhzWahBvKRitAhGfXqlVws7DftWrlGqlxg2BFa8MqVRUakPHJ6P1YDT14JwLckMuH9QDfpAKwtaHpvF0Ks22zdv42Z0/a0yrB/9MkBlS6WiuWAxJSzVVLqJKRbwSEp8v3pKhSxSyJxSMRfZSZkbaBnfRDjogm2CGYqq16tBA3dQqMCStVqZaC8/464G/WqtS8cqQ9PoNh8Xq4OWI+E2IxWqxsez650u1EgOVgTA9qljsLu3myWeeHKxwxCoXFZ/Y/QjtMKxROagH6XgFIB7Y4wG+ubWi3pIwZDzWktFIjy1/SEtHrCJSb5HIpDJDWkDiXasKSX2ZjUsnFujmyCmiYCwi4xakwkCRJ590Vobp6emhu7u75TR3bwTtRoUhCtLl6vBKxJD5muaNVzoaLRhRJaF5uHk5w4arZQqVwrAKSnPFJT69/tkk1AN0PTh71cl+L7w/IR64U5ZqpMUrFY1LIalMo6LQqEhYmlQq1ah8xO+daL6Por7s+DpazTfSeMpSo66vscyowtOY1jQer7TsaWVFwVhE9htmRsbCM9GZIh68G0G7qVWg1Xi8shGvLDQPxy+V1O9lqI/XL6U8sfEJDjrkoEbFIz5/vUJRvx8i3rJRrpUZqAwMyWN8Ga0qNvHhybqvYrLVKykpUo3KyViBWsFYRGQfFqQCAsJr7Enp6euh+yXd077e+H0V8UrAsEpELRbYo8pBc7BvHm/cUNlquKmVot5SgdNYr+NDKi/17i7uavld2grGZnYm8HUgAL7l7v/QNN2i6a8F+oF3uvuvJ1LIIiIio4nfV5FkZWQ8Ps7HW6aPedufmQXAFcBrgBcBbzWzFzXN9hrgqKi7CPjmRDIrIiKyP2nnHvxTgPXuvsHdS8B1wFlN85wFXOOhXwIHmNlBk5xXERGRGamdYHwIsDE2vilKG+88IiIi0kI714xb3f7V/MaAdubBzC4ibMYGKJrZb9tYv0yehcCWpDOxn1GZTz+V+fRTmbfvsFaJ7QTjTcChsfGlwFN7MA/uvhpYDWBm97j7SW2sXyaJynz6qcynn8p8+qnMJ66dZuq1wFFmtszMssC5wE1N89wEvMNCpwI73f3pSc6riIjIjDTmmbG7V8zs/cBPCf+16Sp3f9DMLo6mXwncTPhvTesJ/7XpwqnLsoiIyMzS1v8Zu/vNhAE3nnZlbNiB941z3avHOb9MnMp8+qnMp5/KfPqpzCfIwjgqIiIiSdG7vkRERBKWSDA2szPN7GEzW29mH00iDzOdmV1lZs/F/33MzOab2c/M7PdRf16SeZxpzOxQM7vNzNaZ2YNm9qEoXeU+Rcwsb2Z3m9n9UZn/XZSuMp9CZhaY2f8zs/+IxlXeEzTtwbjNx2vKxF0NnNmU9lHgF+5+FPCLaFwmTwX4a3d/IXAq8L5o21a5T50i8Cp3XwkcD5wZ/UeHynxqfQhYFxtXeU9QEmfG7TxeUybI3dcA25qSzwK+HQ1/G3jjtGZqhnP3p+svSHH33YQHq0NQuU+Z6BG8vdFoJuoclfmUMbOlwJ8A34olq7wnKIlgrEdnJmdJ/f+/o/7ihPMzY5nZ4cAJwK9QuU+pqMn0PuA54GfurjKfWv8b+F9A/GXCKu8JSiIYt/XoTJF9lZl1Ad8H/tLddyWdn5nO3avufjzhk/9OMbNjk87TTGVmrwOec/d7k87LTJNEMG7r0ZkyJZ6tv00r6j+XcH5mHDPLEAbia939B1Gyyn0auPsOoIfwXgmV+dT4I+ANZvYHwkuMrzKz76LynrAkgnE7j9eUqXETcEE0fAHwowTzMuOYmQH/H7DO3b8am6RynyJmtsjMDoiGO4D/AfwOlfmUcPePuftSdz+c8Nh9q7u/HZX3hCXy0A8zey3hdYf64zU/P+2ZmOHM7HtAN+HbVJ4F/hb4d+AG4HnAE8CfuXvzTV6yh8zs5cAdwAMMXk/7OOF1Y5X7FDCz4whvGAoITy5ucPfPmNkCVOZTysy6gb9x99epvCdOT+ASERFJmJ7AJSIikjAFYxERkYQpGIuIiCRMwVhERCRhCsYiIiIJUzAWmQZmVjWz+2LdpD1I38wOj7+da5T5LjWzfjNbHEvrHe0zk50HEWktnXQGRPYTA9EjG5O2Bfhr4JKkMxJnZml3rySdD5Gk6MxYJEFm9gcz+2L0Tt67zez5UfphZvYLM/tN1H9elL7EzH4Yvb/3fjN7WbSowMz+T/RO3/8bPY2qlauAc8xsflM+hpzZmtnfmNml0XCPmX3NzNZE72o+2cx+EL279nOxxaTN7NtRnm80s87o8yea2e1mdq+Z/TT22MQeM/t7M7ud8JV8IvstBWOR6dHR1Ex9TmzaLnc/BfgG4ZPpiIavcffjgGuBy6P0y4Hbo/f3vhh4MEo/CrjC3Y8BdgBvHiEfvYQBebzBr+Tuq4ArCR91+D7gWOCd0dOXAJYDq6M87wLeGz2r+x+Bs939xGjd8SfuHeDup7n7V8aZH5EZRc3UItNjtGbq78X6X4uGXwq8KRr+DvClaPhVwDsgfFsRsNPM5gGPuft90Tz3AoePkpfLgfvMbDwBsP78+AeAB+uvyzOzDYQvftkBbHT3/4rm+y7wQeAWwqD9s/DR3QTA07HlXj+OPIjMWArGIsnzEYZHmqeVYmy4CozUTI277zCzfwXeG0uuMLSlLD/C8mtN66oxeBxpzqMTvjL1QXd/6QjZ6RspnyL7EzVTiyTvnFj/rmj4vwnfigNwHnBnNPwL4C8AzCwwszl7uM6vAn/OYCB9FlhsZgvMLAe8bg+W+Twzqwfdt0Z5fhhYVE83s4yZHbOHeRaZsRSMRaZH8zXjf4hNy5nZrwiv4344SvsgcKGZ/QY4n8FrvB8CXmlmDxA2R+9RYHP3LcAPgVw0XgY+Q/iGqf8gfA3heK0DLojyPB/4pruXgLOBL5rZ/cB9wMtGWYbIfklvbRJJUPSS9pOi4Cgi+ymdGYuIiCRMZ8YiIiIJ05mxiIhIwhSMRUREEqZgLCIikjAFYxERkYQpGIuIiCRMwVhERCRh/z/hCpXDwe2/WwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "mod2historyDF=pd.DataFrame(mod1Training2.history)\n",
    "mod2historyDF.shape\n",
    "\n",
    "mod2historyDF.plot(figsize=(8,5))\n",
    "plt.grid(True)\n",
    "plt.gca().set_ylim(0,1)\n",
    "plt.title(\"Training and Test Metrics by Epoch\")\n",
    "plt.xlabel('Epoch Number')\n",
    "plt.show();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Network Evaluation and Predictions: Test Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test data loss, accuracy:\n",
      "6001/6001 [==============================] - 0s 49us/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.2669764315783034, 0.9255124146174319]"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(\"Test data loss, accuracy:\")\n",
    "mod1.evaluate(X_test10,y_test10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(6001,)"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "array([[557,   0,   1,   0,   0,   0,   6,   1,   2,   1],\n",
       "       [  0, 673,   4,   2,   0,   1,   4,   1,   1,   0],\n",
       "       [  4,   3, 575,  12,   5,   0,   6,  10,   7,   3],\n",
       "       [  4,   2,   8, 536,   0,  21,   2,  14,   6,   2],\n",
       "       [  1,   0,   1,   0, 566,   0,   6,   4,   3,  18],\n",
       "       [  8,   2,   0,  11,   4, 495,   9,   4,   9,   6],\n",
       "       [  6,   3,   7,   0,  11,   8, 520,   3,   4,   0],\n",
       "       [  0,   9,  19,   3,   6,   1,   0, 544,   1,  26],\n",
       "       [  5,   4,   3,  16,   9,   6,   4,   7, 523,  10],\n",
       "       [  3,   7,   1,   7,  22,   5,   1,   7,   4, 565]])"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred=mod1.predict_classes(X_test10)\n",
    "y_pred.shape\n",
    "confusion_matrix(y_test10,y_pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "See [scikit.metrics Classification Report](https://scikit-learn.org/stable/modules/generated/sklearn.metrics.classification_report.html) about the following."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.95      0.98      0.96       568\n",
      "           1       0.96      0.98      0.97       686\n",
      "           2       0.93      0.92      0.92       625\n",
      "           3       0.91      0.90      0.91       595\n",
      "           4       0.91      0.94      0.93       599\n",
      "           5       0.92      0.90      0.91       548\n",
      "           6       0.93      0.93      0.93       562\n",
      "           7       0.91      0.89      0.90       609\n",
      "           8       0.93      0.89      0.91       587\n",
      "           9       0.90      0.91      0.90       622\n",
      "\n",
      "    accuracy                           0.93      6001\n",
      "   macro avg       0.93      0.92      0.92      6001\n",
      "weighted avg       0.93      0.93      0.93      6001\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test10,y_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# UDU: See If you can Improve the Accuracy of the NN, Above.\n",
    "\n",
    "Try different activation functions, increasing the number of neurons in the hidden layers. How about changing the learning rate, batch size, or other hyperparameters?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# UDU: Summarize Performance using Data Not Yet Used\n",
    "\n",
    "So far we haven't yet  used all of the MNIST.  We only used the first 6,000 images. Use some of the remaining data to further assess the above model.  Use whichever version you think is the best."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# CNN\n",
    "\n",
    "We'll now train some CNNs to predict the digit labels using multiple convolution and pooling layers.\n",
    "\n",
    "Parameters to be set for a convulutional layer include:  \n",
    "*the number of kernels, or \"filters\"  \n",
    "* the kernel size  (the shape)  \n",
    "* stride size\n",
    "* \"padding\" type around the image edges  \n",
    "* the type of activation function \n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## CNN1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Reshaping the X Data for A Convolution Layer\n",
    "\n",
    "The X data needs to four dimensions:\n",
    "\n",
    "* Number of samples.\n",
    "* Image height.\n",
    "* Image width.\n",
    "* Channels (1 for grayscale pixels, 3 for RGB color).\n",
    "\n",
    "Using keras, the _order_ of these dimensions depends on what the fit() function argument \"data_format\" is set to, \"channels_first\" or \"channels_last.\" Somewhere in your environment there may be a keras.json file that provides the default as a value of \"image_data_format.\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train10r = X_train10.reshape(X_train10.shape[0], 28, 28, 1).astype('float32')\n",
    "X_test10r = X_test10.reshape(X_test10.shape[0], 28, 28, 1).astype('float32')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Layers:\n",
    "    \n",
    "* Conv - 32 filters that are 5 X 5, no strides, padding with zeros.\n",
    "* Pooling - splits each image dimension in two.\n",
    "* Flattening input; Dense layers expect 1D input\n",
    "* Fully connected \"Dense\" layer with 64 neurons\n",
    "* Softmax output layer "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [],
   "source": [
    " # del cnn1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "keras.backend.clear_session()  # this resets\n",
    "cnn1=keras.models.Sequential([\n",
    "    keras.layers.Conv2D(32,(5,5),activation='relu',padding='same',\n",
    "                        input_shape=[28,28,1]),\n",
    "    keras.layers.MaxPooling2D(pool_size=(3, 3)),\n",
    "    keras.layers.Flatten(),\n",
    "    keras.layers.Dense(32,activation='relu'),\n",
    "    keras.layers.Dense(10,activation='softmax')\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d (Conv2D)              (None, 28, 28, 32)        832       \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 9, 9, 32)          0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 2592)              0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 32)                82976     \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 10)                330       \n",
      "=================================================================\n",
      "Total params: 84,138\n",
      "Trainable params: 84,138\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "cnn1.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Compile CNN1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "cnn1.compile(loss='sparse_categorical_crossentropy',\n",
    "            optimizer='sgd',\n",
    "            metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train CNN1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "cnn1Training=cnn1.fit(X_train10r, y_train10, epochs=50,\n",
    "                     validation_data=(X_test10r,y_test10),\n",
    "                     verbose=0)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### CNN1 Training History"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "cnn1historyDF=pd.DataFrame(cnn1Training.history)\n",
    "cnn1historyDF.plot(figsize=(8,5))\n",
    "plt.grid(True)\n",
    "plt.gca().set_ylim(0,1)\n",
    "plt.title(\"Training and Test Metrics by Epoch\")\n",
    "plt.xlabel('Epoch Number')\n",
    "plt.show();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## CNN1 Evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test data loss, accuracy:\n",
      "6001/6001 [==============================] - 0s 67us/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.13913371925032889, 0.9560073321113148]"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(\"Test data loss, accuracy:\")\n",
    "cnn1.evaluate(X_test10r,y_test10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "cnn1PredClass=cnn1.predict_classes(X_test10r)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[560,   0,   2,   0,   0,   0,   5,   0,   1,   0],\n",
       "       [  0, 678,   3,   1,   0,   0,   3,   0,   1,   0],\n",
       "       [  4,   3, 599,   4,   2,   0,   2,   4,   5,   2],\n",
       "       [  1,   0,   8, 564,   0,  12,   0,   7,   1,   2],\n",
       "       [  1,   1,   1,   0, 581,   0,   6,   3,   1,   5],\n",
       "       [  4,   1,   1,   7,   1, 521,   3,   2,   5,   3],\n",
       "       [  6,   3,   2,   0,   5,   2, 540,   0,   3,   1],\n",
       "       [  0,   4,  14,   3,   4,   0,   0, 569,   2,  13],\n",
       "       [ 10,   4,   7,  12,   7,   0,   1,   5, 533,   8],\n",
       "       [  3,   7,   3,   1,  10,   1,   0,   5,   0, 592]])"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "confusion_matrix(y_test10,cnn1PredClass)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.95      0.99      0.97       568\n",
      "           1       0.97      0.99      0.98       686\n",
      "           2       0.94      0.96      0.95       625\n",
      "           3       0.95      0.95      0.95       595\n",
      "           4       0.95      0.97      0.96       599\n",
      "           5       0.97      0.95      0.96       548\n",
      "           6       0.96      0.96      0.96       562\n",
      "           7       0.96      0.93      0.95       609\n",
      "           8       0.97      0.91      0.94       587\n",
      "           9       0.95      0.95      0.95       622\n",
      "\n",
      "    accuracy                           0.96      6001\n",
      "   macro avg       0.96      0.96      0.96      6001\n",
      "weighted avg       0.96      0.96      0.96      6001\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test10,cnn1PredClass))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# CNN2: A Deeper CNN\n",
    "\n",
    "Let's add some additional Conv and Pooling Layers, an additional fully connected layer, and some dropout:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## CNN2 Network Specification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "keras.backend.clear_session()  # this resets\n",
    "cnn2=keras.models.Sequential([\n",
    "    keras.layers.Conv2D(32,(5,5),activation='relu',padding='same',\n",
    "                        input_shape=[28,28,1]),\n",
    "    keras.layers.MaxPooling2D(pool_size=(3, 3)),\n",
    "    keras.layers.Dropout(0.2),\n",
    "    keras.layers.Conv2D(64,(3,3),activation='relu',padding='same'),\n",
    "    keras.layers.MaxPool2D(pool_size=(3,3)),\n",
    "    keras.layers.Flatten(),\n",
    "    keras.layers.Dense(64,activation='relu'),\n",
    "    keras.layers.Dense(32,activation='relu'),\n",
    "    keras.layers.Dense(10,activation='softmax')\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d (Conv2D)              (None, 28, 28, 32)        832       \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 9, 9, 32)          0         \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 9, 9, 32)          0         \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 9, 9, 64)          18496     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 3, 3, 64)          0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 576)               0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 64)                36928     \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 32)                2080      \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                330       \n",
      "=================================================================\n",
      "Total params: 58,666\n",
      "Trainable params: 58,666\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "cnn2.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Compiling CNN2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "cnn2.compile(loss='sparse_categorical_crossentropy',\n",
    "            optimizer='sgd',\n",
    "            metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training CNN2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "cnn2Training=cnn2.fit(X_train10r, y_train10, epochs=50,\n",
    "                     validation_data=(X_test10r,y_test10),\n",
    "                     verbose=0)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## CNN2 Training History"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "cnn2historyDF=pd.DataFrame(cnn2Training.history)\n",
    "cnn1historyDF.plot(figsize=(8,5))\n",
    "plt.grid(True)\n",
    "plt.gca().set_ylim(0,1)\n",
    "plt.title(\"Training and Test Metrics by Epoch\")\n",
    "plt.xlabel('Epoch Number')\n",
    "plt.show();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## CNN2 Evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test data loss, accuracy:\n",
      "6001/6001 [==============================] - 0s 64us/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.08388416444584156, 0.973004499250125]"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(\"Test data loss, accuracy:\")\n",
    "cnn2.evaluate(X_test10r,y_test10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "cnn2PredClass=cnn2.predict_classes(X_test10r)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[560,   0,   2,   0,   0,   0,   5,   0,   1,   0],\n",
       "       [  0, 678,   3,   1,   0,   0,   3,   0,   1,   0],\n",
       "       [  4,   3, 599,   4,   2,   0,   2,   4,   5,   2],\n",
       "       [  1,   0,   8, 564,   0,  12,   0,   7,   1,   2],\n",
       "       [  1,   1,   1,   0, 581,   0,   6,   3,   1,   5],\n",
       "       [  4,   1,   1,   7,   1, 521,   3,   2,   5,   3],\n",
       "       [  6,   3,   2,   0,   5,   2, 540,   0,   3,   1],\n",
       "       [  0,   4,  14,   3,   4,   0,   0, 569,   2,  13],\n",
       "       [ 10,   4,   7,  12,   7,   0,   1,   5, 533,   8],\n",
       "       [  3,   7,   3,   1,  10,   1,   0,   5,   0, 592]])"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "confusion_matrix(y_test10,cnn1PredClass)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.99      0.98      0.99       568\n",
      "           1       0.98      0.99      0.99       686\n",
      "           2       0.99      0.95      0.97       625\n",
      "           3       0.99      0.95      0.97       595\n",
      "           4       0.97      0.98      0.98       599\n",
      "           5       0.97      0.99      0.98       548\n",
      "           6       0.99      0.98      0.98       562\n",
      "           7       0.96      0.97      0.96       609\n",
      "           8       0.94      0.98      0.96       587\n",
      "           9       0.96      0.96      0.96       622\n",
      "\n",
      "    accuracy                           0.97      6001\n",
      "   macro avg       0.97      0.97      0.97      6001\n",
      "weighted avg       0.97      0.97      0.97      6001\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test10,cnn2PredClass))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Tensorboard Visualization\n",
    "\n",
    "Tensorboard is a utility that allows \"peaking into\" network training.  There are a couple of ways to use it, and one is from in Jupyter.  This works, sometimes.  Here's an example use set-up."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "# set up a dir to write logs into.\n",
    "\n",
    "root_logdir=os.path.join(os.curdir,\"ts_logs\")\n",
    "\n",
    "# A function that will create a new log directory for a training run\n",
    "\n",
    "def create_log_dir():\n",
    "    import time\n",
    "    run_id = time.strftime(\"run_%Y_%m_%d-%H_%M_%S\")\n",
    "    return os.path.join(root_logdir,run_id)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'./ts_logs/run_2019_07_25-14_39_16'"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# test the fcn\n",
    "\n",
    "create_log_dir()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Try a run "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Exception ignored in: <bound method BaseSession._Callable.__del__ of <tensorflow.python.client.session.BaseSession._Callable object at 0x7fe27b3eb6a0>>\n",
      "Traceback (most recent call last):\n",
      "  File \"/home/lynd/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py\", line 1455, in __del__\n",
      "    self._session._session, self._handle, status)\n",
      "  File \"/home/lynd/anaconda3/lib/python3.6/site-packages/tensorflow/python/framework/errors_impl.py\", line 528, in __exit__\n",
      "    c_api.TF_GetCode(self.status.status))\n",
      "tensorflow.python.framework.errors_impl.InvalidArgumentError: No such callable handle: 94146870231392\n"
     ]
    }
   ],
   "source": [
    "cnn1.reset_states() # initialize the cnn1 model\n",
    "tensorboard_cb=keras.callbacks.TensorBoard(create_log_dir)\n",
    "cnn1history=cnn1.fit(X_train10r, y_train10,epochs=50,\n",
    "                    validation_data=(X_test10r,y_test10),\n",
    "                     callbacks=[tensorboard_cb],\n",
    "                    verbose=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>val_loss</th>\n",
       "      <th>val_acc</th>\n",
       "      <th>loss</th>\n",
       "      <th>acc</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>50.000000</td>\n",
       "      <td>50.000000</td>\n",
       "      <td>50.000000</td>\n",
       "      <td>50.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>0.142784</td>\n",
       "      <td>0.957997</td>\n",
       "      <td>0.008783</td>\n",
       "      <td>0.999320</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>0.002413</td>\n",
       "      <td>0.000974</td>\n",
       "      <td>0.002456</td>\n",
       "      <td>0.000413</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.137981</td>\n",
       "      <td>0.955007</td>\n",
       "      <td>0.005348</td>\n",
       "      <td>0.998334</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>0.140954</td>\n",
       "      <td>0.957507</td>\n",
       "      <td>0.006751</td>\n",
       "      <td>0.999167</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>0.142629</td>\n",
       "      <td>0.958174</td>\n",
       "      <td>0.008264</td>\n",
       "      <td>0.999417</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>0.145027</td>\n",
       "      <td>0.958632</td>\n",
       "      <td>0.010783</td>\n",
       "      <td>0.999667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>0.147284</td>\n",
       "      <td>0.959673</td>\n",
       "      <td>0.013916</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        val_loss    val_acc       loss        acc\n",
       "count  50.000000  50.000000  50.000000  50.000000\n",
       "mean    0.142784   0.957997   0.008783   0.999320\n",
       "std     0.002413   0.000974   0.002456   0.000413\n",
       "min     0.137981   0.955007   0.005348   0.998334\n",
       "25%     0.140954   0.957507   0.006751   0.999167\n",
       "50%     0.142629   0.958174   0.008264   0.999417\n",
       "75%     0.145027   0.958632   0.010783   0.999667\n",
       "max     0.147284   0.959673   0.013916   1.000000"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.DataFrame(cnn1history.history).describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Ways of Tensorboard\n",
    "\n",
    "Two ways to do it."
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "# From a system command prompt, assuming that tensorboard is installed, and that the log dir is ts_logs:\n",
    "\n",
    "> tensorboard --logdir=./ts_logs --port=6006"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "# From inside a Jupyter Notebook:\n",
    "\n",
    "%load_ext tensorboard   #iPython \"magic\"\n",
    "%tensorboard --logdir=./ts_logs --port=6006"
   ]
  }
 ],
 "metadata": {
  "hide_input": false,
  "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": "3.6.8"
  },
  "latex_envs": {
   "LaTeX_envs_menu_present": true,
   "autoclose": false,
   "autocomplete": true,
   "bibliofile": "biblio.bib",
   "cite_by": "apalike",
   "current_citInitial": 1,
   "eqLabelWithNumbers": true,
   "eqNumInitial": 1,
   "hotkeys": {
    "equation": "Ctrl-E",
    "itemize": "Ctrl-I"
   },
   "labels_anchors": false,
   "latex_user_defs": false,
   "report_style_numbering": false,
   "user_envs_cfg": false
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": false,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {
    "height": "calc(100% - 180px)",
    "left": "10px",
    "top": "150px",
    "width": "212px"
   },
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}