{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Import libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Keras documentation can be found on keras.io\n",
    "\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "import seaborn as sns\n",
    "sns.set_style(\"darkgrid\")\n",
    "\n",
    "import keras\n",
    "from keras.datasets import mnist\n",
    "# manual link to data:  http://yann.lecun.com/exdb/mnist/\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Conv2D, MaxPooling2D, Flatten, Dense, Dropout\n",
    "from keras.losses import categorical_crossentropy\n",
    "from keras.optimizers import Adadelta\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "#import os\n",
    "#os.environ[\"CUDA_VISIBLE_DEVICES\"]=\"0\" # for training on gpu\n",
    "\n",
    "import tensorflow as tf\n",
    "tf.logging.set_verbosity(tf.logging.ERROR) # suppress deprecation warnings for demo purposes"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Load and prepare the MNIST dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# the data, split between train and testval sets\n",
    "(train_images, train_labels), (testval_images, testval_labels) = mnist.load_data()\n",
    "\n",
    "# further split testval set into specific test and validation set\n",
    "# the test set is NOT used during any part but inference\n",
    "val_images, test_images, val_labels, test_labels = train_test_split(testval_images, testval_labels, test_size=0.2, random_state=13052020)\n",
    "\n",
    "# explicitly illustrating standardization\n",
    "def standardizeimg(img, mu, sigma):\n",
    "    return (img-mu)/(sigma).astype(np.float32)\n",
    "\n",
    "# save for scaling test data\n",
    "mu_train = np.mean(train_images)\n",
    "sigma_train = np.std(train_images)\n",
    "\n",
    "# Standardize pixel distribution to have zero mean and unit variance\n",
    "train_images = standardizeimg(img=train_images, mu=mu_train, sigma=sigma_train)\n",
    "val_images = standardizeimg(img=val_images, mu=np.mean(val_images), sigma=np.std(val_images))\n",
    "\n",
    "# adapt to format required by tensorflow; Using channels_last --> (n_samples, img_rows, img_cols, n_channels)\n",
    "img_rows, img_cols = 28, 28 # input image dimensions\n",
    "train_images = train_images.reshape(train_images.shape[0], img_rows, img_cols, 1)\n",
    "val_images = val_images.reshape(val_images.shape[0], img_rows, img_cols, 1)\n",
    "\n",
    "# convert class vectors to binary class matrices - one hot encoding\n",
    "num_classes = 10\n",
    "train_labels = keras.utils.to_categorical(train_labels, num_classes)\n",
    "val_labels = keras.utils.to_categorical(val_labels, num_classes)\n",
    "\n",
    "# avoid using statistics intrinsic to test data to ensure unbiased estimate of real model performance\n",
    "test_images = standardizeimg(img=test_images, mu=mu_train, sigma=sigma_train)\n",
    "test_images = test_images.reshape(test_images.shape[0], img_rows, img_cols, 1)\n",
    "test_labels = keras.utils.to_categorical(test_labels, num_classes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Inspect data shapes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set:\n",
      " > images: (60000, 28, 28, 1)\n",
      " > labels: (60000, 10)\n",
      "Validation set:\n",
      " > images: (8000, 28, 28, 1)\n",
      " > labels: (8000, 10)\n",
      "Test set:\n",
      " > images: (2000, 28, 28, 1)\n",
      " > labels: (2000, 10)\n"
     ]
    }
   ],
   "source": [
    "print(\"Training set:\")\n",
    "print(\" > images:\", train_images.shape)\n",
    "print(\" > labels:\", train_labels.shape)\n",
    "print(\"Validation set:\")\n",
    "print(\" > images:\", val_images.shape)\n",
    "print(\" > labels:\", val_labels.shape)\n",
    "print(\"Test set:\")\n",
    "print(\" > images:\", test_images.shape)\n",
    "print(\" > labels:\", test_labels.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Verify first 5 mages in each split dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAArEAAAGoCAYAAABR673jAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjAsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+17YcXAAAgAElEQVR4nO3deYDNdfv/8WtmjJkh+3YzltKNsUtCaSIhxE226ka2hAqVpSJuCcmNsi/ptq/JrhKKKImsjbVNlqxjHWOGMb8/+nZ+53ozZ4yZcz7nPfN8/HO/X+dzzudc5WSu+zPXeX8CEhMTEwUAAACwSKDTBQAAAAApRRMLAAAA69DEAgAAwDo0sQAAALAOTSwAAACsQxMLAAAA69DEAgAAwDo0sQAAALAOTSwAAACsQxMLAAAA69DEAgAAwDo0sQAAALAOTSwAAACsQxMLAAAA69DEAgAAwDo0sQAAALAOTSwAAACsQxMLAAAA69DEAgAAwDo0sQAAALBOJqcLAGz1448/qjx+/HiVZ86c6Vq3a9dOHevevbvKlStXTuPqAABI37gSCwAAAOvQxAIAAMA6NLEAAACwTkBiYmKi00WktYSEBJUvXryYotebs41Xr151rQ8ePKiOTZgwQeXevXurPH/+fJVDQ0NVfvPNN1X+z3/+k6Ja4Tu7du1S+fHHH1f50qVLd3yuHDlyqBwdHX33hSHDWr9+vcqtW7dWeePGjSqXKlXK6zXBeUOGDFF54MCBKps/9jds2KByzZo1vVIXkNa4EgsAAADr0MQCAADAOjSxAAAAsI7f7hP7xx9/qBwfH6/yd99951pv3rxZHbtw4YLKixcvTrO6ihQporK53+fSpUtVzpYtm8oVK1ZUmdkj//XDDz+o3Lx5c5XNWeuAgACVs2fP7lpnzpxZHTt79qzKW7ZsUfnBBx9U2Xw9tG+++Ublc+fOqfz000/7shyf2bZtm8pVqlRxqBI4bcaMGa718OHD1bGgoCCVze+NmH93AbbgSiwAAACsQxMLAAAA69DEAgAAwDp+MxO7c+dOlWvXrq1ySvd6TUvu80Tm/ntZs2ZV2dynsVChQirnypVLZfZtdI77/r8iIjt27FC5TZs2Kp84cSJF5y9RooRr3bdvX3XsmWeeUblGjRoqm5+zfv36pei9Mxpzn8vDhw+rnF5mYm/evKnyb7/9prL5XYJ0uA04knDkyBHXOi4uzsFK4C1bt25Vefbs2Sqb3w346aefPJ5v1KhRrrXZq2zatEnltm3bqlytWjXPxfoIV2IBAABgHZpYAAAAWIcmFgAAANbxm5nYYsWKqZw3b16V03Im1pzlMOdUv/76a5Xd9+g050Jgry5duqg8b968ND3/jz/+6FpfuXJFHTP3BzZnOvfu3ZumtaR3M2fOVPmRRx5xqBLv+vPPP1WeOnWqyubfTxEREV6vCc5Yt26dymPHjk3yuebnYNWqVSoXKFAg7QpDmlm4cKHKPXv2VPnMmTMqmzPwtWrVUtncn7x3795Jvrd5LvO1CxYsSPK1vsSVWAAAAFiHJhYAAADWoYkFAACAdfxmJjZ37twq//e//1V55cqVKj/wwAOudY8ePTyeu1KlSiqbs0TmXq/m3mqeZo1gF/c5VXMuLLk9Nc35okaNGqlszhe577vn/nkVSX4Om/09U8bcPzW9euGFFzwed9+bGOnL5s2bVW7fvr3Kly5dSvK1ffr0Udn8Dgqcc+PGDZW3bdvmWnfu3Fkdi4mJUdn8bsWAAQNUfvTRR1U29w9u1aqVa71mzRqPdVapUsXjcadwJRYAAADWoYkFAACAdWhiAQAAYB2/mYk1NW3aVOXatWurnC1bNtd6z5496ti0adNUNmcVzRlYU7ly5VQ292KEPXbt2qVynTp1XGtzhiwgIEDlhg0bqjx//nyVzb1dhw4dqrL7/GK+fPnUsYoVK3p879WrV6u8Y8cOlStXriwZnft/96dOnXKwEt+5cOGCx+N169b1USXwNXMv5BMnTiT5XHN+//nnn/dGSUgDc+bMUblTp05JPrdevXoqm/vIZs+e3eN7mc/3NAdbpEgRldu1a+fx3E7hSiwAAACsQxMLAAAA6/jtOIHJ02XyHDlyeHytOV7w7LPPqhwYSC+fXhw6dEjlESNGqOx++2LzV/wFCxZU2fz1yT333KOyucWWmVPj6tWrKo8cOVLltL5Fro0+++wz1zo2NtbBSrzHHJP4/fffPT4/PDzci9XAl8zbfH788ccqBwUFqZwzZ07X+u233/ZeYUgV889m2LBhKruPlr388svq2JAhQ1RObnzAZI68eWJuLWr+vPQXdG8AAACwDk0sAAAArEMTCwAAAOtYMxPryaBBg1R2v7WoyK1bIZm3nTW3rYA9zNvomdupmVtVuc8QzZo1Sx0zb6vnT3OWR48edboEv3Pw4MEkj5UtW9aHlXiP+Xk+efKkyqVKlVLZfetB2MWcd27WrFmKXt+9e3fX2tySEs4ZPHiwyuYMbEhIiMpPPvmka/3++++rY2FhYR7f69q1ayp/+eWXKh85ckRl99ubm7esbdKkicf38hdciQUAAIB1aGIBAABgHZpYAAAAWCddzMSat5H96KOPVDZv0dm5c2eVH3/8cZXN2Uj3vdrM24PCWebtWM0ZWNPy5ctd65o1a3qlJjjvoYcecrqEJJm3O/7iiy9Udr8NpTnTZjL3nHTfKxR2MT8He/fu9fj8J554QuWePXumeU1IOfPW0BMnTlTZ7CHcZ2BFRJYtW3bH7/Xzzz+r3Lp1a5W3b9/u8fUtW7Z0rfv27XvH7+tPuBILAAAA69DEAgAAwDo0sQAAALBOupiJNd1///0qz5gxQ+UOHTqobO4XauaYmBjX+vnnn1fHChYseLdlIg28/vrrKrvveyciUqtWLZX9dQ7WrDulx6FFR0en6vW7d+9W+ebNm671+vXr1bFjx46pHB8fr/LcuXOTPJfIrXs/VqtWzbU295C8fv26yub8Puxhzj6++eabHp8fGRmp8syZM1XOkSNH2hSGVDH/+z9z5ozH548dO1bl06dPu9bTp09Xx9y/0yEiEhUVpfLly5dVNudvAwP1dcs2bdq41uZ3i2zBlVgAAABYhyYWAAAA1qGJBQAAgHXS5Uys6emnn1b5n//8p8q9evVSed26dSq/9dZbrrV57+H+/furHB4eftd1InmrVq1SedeuXSqbM0D/+te/vF5TWjDrNnOlSpV8WY4V3GdJzX9fXbp0Udm8X3lyzJlY95nk4OBgdSxLliwqly5dWuWOHTuq/OCDD6pszm0XKFDAtS5cuLA6Fhsbq3JERIRZOvzU77//rnKzZs1S9PrixYur7P45gf/InDmzyvnz51fZfeZVROTee+9VOSV70Zv9Rvbs2VU+ceKEynnz5lW5cePGd/xe/oorsQAAALAOTSwAAACsQxMLAAAA62SImVhT+fLlVV60aJHKK1euVLl9+/au9eTJk9Wxw4cPq7x27do0qBBJMWcCzT35zPmjZ555xus13Ym4uDiVBw0a5PH55n3Rhw8fntYlWc/9nuTFihVTx7777rtUnbto0aIqN2nSxLUuU6aMOla9evVUvZdp6tSprrU5P2fORcIe77//vspBQUEpen1y+8jCP+TMmVNlcz/gRo0aqXzu3DmV3b+z4/73jojuRUREcufOrfKzzz6rsjkTax5PD7gSCwAAAOvQxAIAAMA6NLEAAACwToaciTWZMyxt27ZV+YUXXnCtzXuXf/PNNypv2LBBZXMPSHhXaGioygULFnSoEj0HO2TIEHVsxIgRKhcpUkRlc+/ie+65J42rS1/eeOMNp0tIM+vXr0/yWIsWLXxYCVLLfR/rNWvWpOi15h7XpUqVSpOa4FvVqlVT+cyZM2l2brP/2Lhxo8rmnrPpcaaeK7EAAACwDk0sAAAArJMhxwn27Nmj8uLFi1Xetm2byuYIgTtzu53HHnssldUhNZy8zax5C1z3kYGFCxeqY+bWKUuWLPFeYUg3mjZt6nQJSIF69eq51ufPn/f4XPPXzjNnzvRKTUg/zC0nk7t9OVtsAQAAAH6AJhYAAADWoYkFAACAddLlTOzBgwdVHjdunMrm/OHJkyfv+NyZMul/ZeYWToGB/P8Cb0pMTPSYzVv8jRkzxmu1jB49WuV3331X5YsXL7rWbdq0UcdmzZrltboA+IezZ8+61sndZvbll19WmW31kJwnn3zS6RIcR8cFAAAA69DEAgAAwDo0sQAAALCOtTOx7nOs8+bNU8fGjx+v8u+//56q93rooYdc6/79+6tjTu5LmhEltw+eOd/co0cPlTt27Oha58mTRx37/vvvVZ49e7bKu3fvVvno0aMqFytWTOX69eu71i+99JIAqXX48GGVH374YYcqwe106NBBZfeZ/YSEBI+vfeSRR7xSE9KvlN7KOD3iSiwAAACsQxMLAAAA69DEAgAAwDp+OxN76tQplaOiolR+5ZVXXOsDBw6k6r3Me1b37dtXZff73LMPrH+7ceOGyhMmTFB58eLFrnWOHDnUsUOHDqXovcwZttq1a6s8ePDgFJ0PSM7NmzedLgFudu3apfLatWtVdp/ZDwkJUcfMOfkCBQqkcXVI73755RenS3AcHRkAAACsQxMLAAAA69DEAgAAwDqOzcRGR0er3KVLF5XNWaPUzH7UqFFD5V69eqls3n84LCzsrt8L3mXui1m1alWVf/jhB4+vd99H1py7NuXNm1flZ599VuUxY8Z4fD2Q1rZs2aJy+/btnSkEIiJy4cIFlT39nVKoUCGVR40a5ZWakHFERkaq7L4vcUbBlVgAAABYhyYWAAAA1qGJBQAAgHW8NhO7detWlUeMGKHytm3bVD527Nhdv1eWLFlU7tGjh8r9+/dXOWvWrHf9XnBW4cKFVV6yZInKU6ZMUfndd9+943P37NlT5W7duqlcokSJOz4XAADeVL58eZXNn1Hmd4nMnC9fPu8U5kNciQUAAIB1aGIBAABgHZpYAAAAWMdrM7FLly71mJNTpkwZlRs3bqxyUFCQa927d291LGfOnCl6L9irYMGCKg8aNMhjBvxdgwYNXOtFixY5WAmSExERofIjjzyi8qZNm3xZDjK4fv36qdypUyePx8ePH+9amz2XLbgSCwAAAOvQxAIAAMA6AYkZ8T5lAAAA6cilS5dUbtWqlcpr165VuXnz5q719OnT1TFbtiLlSiwAAACsQxMLAAAA69DEAgAAwDrMxAIAAKQz5oxs//79VZ44caJrvXfvXnXMli23uBILAAAA69DEAgAAwDo0sQAAALAOM7EAAACwDldiAQAAYB2aWAAAAFiHJhYAAADWoYkFAACAdWhiAQAAYB2aWAAAAFiHJhYAAADWoYkFAACAdWhiAQAAYJ1MThfgK6VKlbrt41myZJGdO3f6uBrY4ObNmzJr1ixZsGCBHD9+XHLnzi0NGjSQHj16SJYsWZwuD5aIjY2Vp556So4fPy6tW7eWgQMHOl0S/NCUKVMkKipKoqKi5NixYxIeHi5fffWV02XBz509e1bGjh0rGzdulHPnzknevHmlTp060qNHD8mePbvT5XldhmliRUSqVKkirVq1Uo8FBwc7VA383bBhw2T27NlSt25d6dixo/zyyy8ye/Zs2bdvn8yYMUMCA/lFBpI3duxYOX/+vNNlwM+NHj1acubMKWXKlJHLly87XQ4scO7cOWnVqpWcPn1annnmGSlRooQcPnxYFixYINu3b5f58+dLWFiY02V6VYZqYosUKSJNmjRxugxY4PDhwzJnzhypV6+ejBs3zvV44cKFZciQIbJ69Wpp3LixgxXCBlFRUTJz5kzp06ePDB8+3Oly4MfWrVsnRYoUERGRRo0aydWrVx2uCP5u8uTJcvz4cRk1apQ0atTI9fgDDzwgvXr1kunTp8tLL73kYIXel+EuJcXHx0tMTIzTZcDPrVq1ShITE6Vdu3bq8VatWklYWJisWLHCocpgi4SEBBkwYIBERkZK3bp1nS4Hfu7vBha4U1u3bpXQ0FB56qmn1OMNGzaUkJAQWbJkiUOV+U6GamLXrFkjlSpVksqVK8vDDz8s7777Lr+2wW399NNPEhgYKBUqVFCPh4SESEREhOzdu9ehymCLGTNmyK+//ioDBgxwuhQA6VB8fLyEhIRIQECAejwwMFBCQ0Pl6NGjEh0d7VB1vpFhmtgKFSrIK6+8ImPHjpX3339fqlevLnPmzJF///vfXJnFLU6fPi25cuWSzJkz33KsQIECcv78eYmPj3egMtjg6NGjMm7cOHnppZekcOHCTpcDIB0qUaKEXLx4Ufbv368e379/v1y8eFFERP78808nSvOZDDMT+8knn6jctGlTKVWqlHzwwQcya9Ys6datm0OVwR/FxsbetoEV+etqrIjItWvXknwOMrZBgwZJ4cKFpUOHDk6XAiCdateunaxbt05effVV6devn+uLXcOGDZPg4GC5fv26xMbGOl2mV2WYK7G306lTJwkODpaNGzc6XQr8TFhYWJJXWuPi4kREJDQ01JclwRLLly+Xb7/9VgYNGsTuJwC8pkqVKjJ69GiJiYmRF198UR5//HHp1q2bVKtWTWrVqiUiIvfcc4+zRXpZhrkSezvBwcGSP39+tr/BLfLnzy8///yzxMfH33K19dSpU0mOGiBji4+Pl+HDh0vNmjUlX758cuTIERH56zMjInL58mU5cuSI5MqVK0Ps4QjAuxo0aCD16tWTQ4cOSUxMjNx3332SJ08eadGihWTKlEmKFSvmdIlelaGb2Li4ODl16pRUrFjR6VLgZ8qVKyebN2+WPXv2SJUqVVyPx8XFyYEDB9RjwN+uXbsm0dHRsmHDBtmwYcMtx1esWCErVqyQvn37SqdOnXxfIIB0JygoSEqXLu3KZ86ckf3798tDDz3EPrHpwfnz5yVXrly3PP7hhx/KjRs35PHHH3egKvizhg0bypQpU2TmzJmqYV20aJHExsayRyxuKywsTMaMGXPL49HR0fLOO+9IZGSktGjRIsk7CAJAaty8eVOGDBkiCQkJ0rVrV6fL8boM0cROmjRJdu/eLdWqVZOCBQvK1atXZePGjbJ161apWLGitG3b1ukS4WdKlSolrVu3ljlz5sgrr7wiNWvWdN2xq2rVqjSxuK3g4GCpX7/+LY8fO3ZMRESKFi162+PAsmXL5MSJEyLy1//puX79ukycOFFERAoVKiRNmzZ1sjz4oZiYGGnZsqXUrVtXChcuLJcvX5ZVq1ZJVFSUvPbaa1K9enWnS/S6DNHEVq1aVX755RdZunSpXLhwQYKCgqRYsWLy2muvSYcOHVzfNgfc9evXT8LDw2XhwoWyYcMGyZUrl7Rp00Z69OjBLWcBpKlPP/1UfvjhB/XY31f1q1atShOLWwQHB0upUqVk5cqVcubMGQkLC5Py5cvLtGnTJDIy0unyfCIgMTEx0ekiAAAAgJTgchIAAACsQxMLAAAA69DEAgAAwDo0sQAAALAOTSwAAACsQxMLAAAA69DEAgAAwDo0sQAAALAOTSwAAACsQxMLAAAA69DEAgAAwDo0sQAAALAOTSwAAACsQxMLAAAA69DEAgAAwDo0sQAAALAOTSwAAACsQxMLAAAA69DEAgAAwDo0sQAAALAOTSwAAACsQxMLAAAA69DEAgAAwDo0sQAAALAOTSwAAACsQxMLAAAA69DEAgAAwDqZnC7AG65fv67yjh07VP700089vv6xxx5T+eGHH3at8+TJk8rqAABImZiYGJWXLl2q8ubNm+/63PXq1VO5WbNmd32ujO706dMqFyhQwGvv9Y9//EPl/v37q9ytWzeVg4KCvFaLU7gSCwAAAOvQxAIAAMA6NLEAAACwTkBiYmKi00Wk1pYtW1R+7733VF69enWKzmf+KylVqlSS5ypevHiKzo30Y+vWrSpv27ZN5dGjR7vWv/32m8dz5cyZU+UBAwao/Prrr3t8/XfffedaZ86cWR2rUqWKx9cC8L39+/er/Pzzz3t8vjkTe/DgQZXdf24FBAQkeex2x7NkyaKy+XdZRESEx9rw/9WqVUvlb775xplCROQ///mPx5wecCUWAAAA1qGJBQAAgHVoYgEAAGAda2diN2zY4FrXrl1bHTPnffLmzaty5cqVPZ57zZo1SZ7vwQcfVMfWr1+vcrZs2TyeG/Yw9xt+++23VXafeRURuXHjxl2/V8GCBVU+efKkyv/6179U3rlzp8onTpxwratVq6aOpWb/SKTer7/+qvLkyZNV/uSTT1Q+cuRIkudatWqVyg0bNkxldXCKORNbpkwZlVMz11q6dGl17MqVKyp7mqe93ev37dsnuDMtW7ZU2fx3X65cOZXbtWunsvt3LQ4cOKCOTZs2zeO5ExISVM6fP7/K7j8nREQCA+2/jmn/PwEAAAAyHJpYAAAAWIcmFgAAANaxZibWnAVr0aKFax0fH6+ORUZGqvzOO++obO7jZipfvrzKnuaBdu3a5fG1sMf8+fNVHjRokMqHDh1S2ZxtMv/se/fu7Vrfd9996pg5J/nhhx+qPGfOHI+1mnvBut8zu1evXupY1qxZPZ4LqXP+/HmVlyxZonLnzp1VNmcZUyIkJETlefPmqdy0adO7Pjd8y5yJHTt2bIpe//TTT6tctGhR19rc17VZs2YqL1++XGWzDTDn/wcPHpyi2uAb5t7Cyf3cWLBggcqtWrVK85p8jSuxAAAAsA5NLAAAAKxDEwsAAADrZHK6gKS47wMrItK2bVuV3ffwNPdKXLx4scqhoaEpeu8GDRqozB556cfVq1dd63Xr1qljHTp0UPnmzZsqP/rooyqvXbtWZfNz9ttvv7nW7vOxIrfuDXr58mWP56pQoYLK7733nsrmXsnwnnPnzqlszpWZf3eZzNlpc49gT+e6du2ayua8PzOx9jD3Yp00aVKqzhcTE+Nam7ORS5cuVdnTHrMiIg888ECqagF8hSuxAAAAsA5NLAAAAKxDEwsAAADr+O1MrMmcEXTP5h6yqWXumedpK133OST4v2XLlrnWrVu3VsfMvVfNOdZhw4Z5PLe5t+KECRNc6wsXLnh8bb169VTu3r27yo0aNfL4eniX+0yh+bn4/fffPb62a9euKg8dOlTlXLlyJflac1721KlTHt8LGZf7Z7Rdu3bqmDkDa2ZzH1lzD1r4B3NP6p9++snj8809ws39g9MDrsQCAADAOjSxAAAAsA5NLAAAAKzjtzOxtWrVUvnPP//02XsnNz/kbvLkySpXr17dKzXh7nz//fcqv/baa661OQP7+uuvq2zOwB49elTlF198UeUvvvhCZfd5pIcfflgdM++TXqxYMZXz5csncM5nn32msvs+1bGxsepY2bJlVV6xYoXK9957b9oW56Zbt25eOzf8y48//qiy+ffVN99841qbP7Py5s2r8vbt21UuWrRoWpSI2zD/vjhw4IDK06dPV/nKlStJnmv+/Pkqx8XFeXxv82dcQkKCyu77ppt7BduCK7EAAACwDk0sAAAArOO34wS2aN68udMlwI3565WOHTuqfPr0adfa3CrJ/PXcSy+9pLL5qxxz2yxzqyT3Wz+at0aGfzG3yhs0aJDK7r92q1Spkjo2ZMgQlVM7PuD+a2PzV4vmdn/mrUrNERekzP79+1U2fy3vzS2Kpk6dqrJ5q9gdO3aofPbsWZXdazXrNp/rPlYlcut2g+aWW7hz169fV7lGjRoq79q1y2e1mFtyPfjggypXrFjRtZ41a5Y6Zt4e219xJRYAAADWoYkFAACAdWhiAQAAYB1mYm9j3759d/xc921NREQaN26c1uUgBbZu3aqyOePmvtVM5cqV1TFze7Rff/1VZfPWxz169FDZvO0s22TZ491331XZ3ILIfcbw/fffV8fq1q2bprWMGjXKtXafxTXrQOqdOXNG5TFjxqjcr1+/NDv/mjVr1LHnn39eZXPe2fyzTu64+xZJpUuXVsc6d+6ssrnNk/nPbW75Z85SImnmn1NazsCaf+Y5cuTw+Pxr1655zLt373atzdndnj17qmx+TyAoKMjje/sKV2IBAABgHZpYAAAAWIcmFgAAANZhJvY2zFtOus+hmLNCXbt29UlNSBvu+8j+5z//UcfMGVhzZnbmzJkqlytXLo2rg6+sW7dOZXMm0FS7dm3X2vxcpNa5c+dUNue6PbnvvvvStJaM5r333lPZnJE1b9eaHPP1vXr1cq3d940WSX6+2TzepUsXj8+vV6+ea/300097fK6pTJkyKm/evFllZmLvXKZMuq3q0KGDyuZ3bnbu3KlyeHi4a50/f351zOw32rVr57GWLVu2qGzuhf7TTz+51uae1EOHDlXZvGWtedypeX2uxAIAAMA6NLEAAACwDk0sAAAArMNMrNw6A2vu8xYcHOxaT58+XR27//77vVcYUqxs2bIqm/e5d9+z7/Lly+pY06ZNVZ42bZrKefLkSYsS4QfMvV7j4+NVNvcEfuedd1zr1H4Oli5dqnLv3r1V/v3335N87WOPPabyxx9/nKpaMrqIiAiVzbnVs2fPqly0aFGP5zP3hnafqTV/rpiefPJJlc09aiMjIz2+PiXM2V1zP21zH1ncucBAfW0wuf9GzX1k3WdiU7vX+MMPP6yyOSP7ySefuNbmXtnu87IiIsOHD1fZ/e9EEd0n+RJXYgEAAGAdmlgAAABYhyYWAAAA1kkXM7Hu9/+9neLFi6ucLVs2lTt16qSyud9ZrVq1XOuaNWveRYXwleTmzjwx58BWrVqlcnJ78sEe5qyjybwn+SOPPJLkc839g815w1GjRqlszmLHxsZ6rMXd6tWrVc6aNesdvxZ/cf/z+eijj9SxtN7r0n2utU2bNh6fm9b7D7szP5MNGzZU2fznTuk+s7h75vc2fKlly5au9Y4dO9QxcybWNGXKFJVfeeWVtCssBbgSCwAAAOvQxAIAAMA6NLEAAACwjjUzsSdOnFDZff5i2bJl6pg532Pe9zl37twqX7hwweN7u88qmfc5Z+9QZ5lzO23btlXZ3IOvcePGrrV5z2pzJta85/XChQtVNmeAzDkz+C9zH1hzlvrkyZMqm3s/poR57pTMXbrP44swA5sW/vjjj9uuRUTCwsJUvnr1ahpLfwgAACAASURBVKrey32fWPe1t5n7vjZr1kzlgwcPqmzuUevN+VykD+Z/O07hSiwAAACsQxMLAAAA69DEAgAAwDp+OxPbrVs3lT///HOVjx49esfn+vHHH1NVy4gRI1zrxYsXq2P33HOPyl27dlXZnL9135cNqWf+eZgzsC1atFB52rRprvWVK1fUse7du6v89ddfq2x+Br/66iuVS5YsqfKYMWNc60cffVQdc+o+0/jL3LlzVX7++edV/u6771ROyRxr2bJlVX7sscdU3rhxo8r79u1L8lwDBw684/fFndm0aZNrbe4XXK9ePZUjIiJ8UlNacN8L1pzPP3LkiMrm57lz584q+3J+N707f/68yteuXVO5YMGCviwnzWTK5B/tI1diAQAAYB2aWAAAAFgnIDE19+lMhZiYGJXdtz4SEdmyZYvK8fHxKj/00EOutfmPsH379lTVlpotcUzm1jwp+TWN+z+jiMiKFSvuuo70wvxV7L/+9S+VW7durbL7KIjIreMfnnz//fcqz5s3T2VzvODnn39W2f3XRJMnT1bHzLrhX9auXXvXrzVvTR0dHa1yeHi4x9fXqVPHtV65cqU6ljlz5ruuC39x//NxHy0QufXvj9mzZ/ukprsxdepUlT/44APX2txCq2jRoiqb/1yRkZFpXB3+Zo6pbdiwQeU1a9aoXKhQIW+XdFtmnRMmTPD4fPNn2osvvpjmNd0JrsQCAADAOjSxAAAAsA5NLAAAAKzj2B4Jp06dUtmcEzGZW1dNnDjRtZ4+fbo61qlTJ5VDQkJUzpkzp8rmdl5ZsmRR+bPPPnOtzdv5mczbFF66dEllcz7O3IKrTJkyrnWvXr08vldGZM6hmv9+//3vf6uckhlYU/Xq1T1mc4su83M0Z84c1/rZZ59Vx8aNG6ey+ZmFs+rWrZtm53rzzTdT9Pw33njDtWYGNu09/fTTrvXmzZvVMTObM7NOzo6at9Q2a3XfRsv8Hof7vKwIM7C+lD9/fpWjoqJUHj9+vMrDhg3zek1/c+8xpkyZkqLXmrdldwpXYgEAAGAdmlgAAABYhyYWAAAA1nFsn1hznmLWrFken9+3b1+VL1y44FqbsxzmTOvYsWNV7tix4x3XmVLmLSR37NihsjmP26hRI6/Vkh4999xzKi9YsEDlnTt3qlypUiWv1/Q3cz7XfZ/fQ4cOqWPuc3kit37+UzPLC2dt27ZN5WrVqqlsziua+8p++eWXrrW/3NoxPfnmm29ca/N7B+Ye46VLl1bZnCU1j5u3l3b/szZ/1B44cEDlJUuWqLxs2TKVk9u//Mknn3Stze8OwDknTpxQ2fz7ICEhQWX3fsW8bXpqmT9n3L9rZN4O1+T+fR0Rkd27d6scFBSUyuruDldiAQAAYB2aWAAAAFiHJhYAAADWcWwmNjBQ98/mfE9KmHs6tmrVSmVvzsDCtx544AGVzf2GDx8+rHLWrFnT7L1jYmJU3rNnj8ojR45U2X2mzdzv8+OPP1bZ3N8W9oiNjVW5efPmKpvziebfdUOGDFG5X79+aVgdPDl79qzKjz32mMoHDx5UObm5VE/HU/NaEZG8efOqbH5OWrduneRz4T/M70MsX748yecGBwerbO4nHhoa6vG9zDlX8/tDntq/Bx98UGX3/fJFRPLly+fxvX2FK7EAAACwDk0sAAAArEMTCwAAAOs4NhNrzhPOnz9f5REjRnh8fcuWLV1rc69QpF9PPPGEyl999ZXKPXr0UPnZZ59VuVSpUkme25xtHDVqlMrmTJA5L2cqUaKEa/3ee++pY+bcJOw1evRolfv06aOy+VdslSpVVDZn4goWLJiG1SElzpw5o7L53+2mTZs8vv7IkSMqu8/cmvuXm3vMdu7cWeVmzZqpzJxr+nDy5EmV33rrLZVnzpzpy3JczO+brF69WuV//OMfviznjnElFgAAANahiQUAAIB1aGIBAABgHcdmYk3Xr19X+fz58x6f7z4fZO45i/TLnEvt27evyvv27VM5LT/e5p59999/v8ruc9oiIl26dHGtw8PD06wOOMvcn3PixIkqX758WWVzj+BVq1apbM55w15//PGHyp5mYiMiInxSE/xbdHS0ygcOHHCtzT2kv/jii1S9V6FChVR+++23XWtzf/3cuXOn6r18he4PAAAA1qGJBQAAgHVoYgEAAGAdv5mJBdKCOTNrzlq773135coVdcy8D3WTJk1ULlq0qMrmvnrIGOrXr6/y2rVrPT6/Vq1aKq9fvz6tSwKADIkrsQAAALAOTSwAAACswzgBAKTARx99pPKAAQNUvvfee1U2xweyZs3qlboAIKPhSiwAAACsQxMLAAAA69DEAgAAwDrMxAIAAMA6XIkFAACAdWhiAQAAYB2aWAAAAFiHJhYAAADWoYkFAACAdWhiAQAAYB2aWAAAAFiHJhYAAADWoYkFAACAdWhiAQAAYB2aWAAAAFiHJhYAAADWoYkFAACAdTI5XYAv/Pbbb7JixQr59ttv5Y8//pC4uDgpWrSo1K9fX9q1aydZsmRxukT4oZiYGJk9e7asXr1ajh07JpkzZ5b77rtPWrVqJU8//bQEBAQ4XSL80NmzZ2Xs2LGyceNGOXfunOTNm1fq1KkjPXr0kOzZsztdHvzQr7/+KhMmTJB9+/bJ6dOn5caNG1KwYEGpWbOmdOrUSfLnz+90ifBDfG5EAhITExOdLsLbRo4cKXPnzpXatWtLpUqVJFOmTLJ161b5/PPPpVSpUrJo0SIJDQ11ukz4kZs3b0qbNm1k586d0rRpU6lUqZLExsbK6tWrZc+ePfLCCy9Inz59nC4TfubcuXPSsmVLOX36tDzzzDNSokQJOXz4sCxcuFD++c9/yvz58yUsLMzpMuFntmzZIpMmTZJKlSpJgQIFJFOmTHLo0CFZsmSJZM2aVZYvXy558uRxukz4GT43IpKYAezZsyfx0qVLtzw+evToxJIlSybOnj3bgargz3bs2JFYsmTJxKFDh6rH4+LiEmvXrp344IMPOlQZ/NmQIUMSS5Ysmbhy5Ur1+MqVKxNLliyZOGHCBIcqg40+++yzxJIlSyZOnTrV6VJgkYz0uckQM7Hly5eXbNmy3fJ4w4YNRUTk0KFDvi4Jfu7KlSsiIrf8OiZz5sySK1curqbhtrZu3SqhoaHy1FNPqccbNmwoISEhsmTJEocqg43Cw8NFROTSpUsOVwKbZKTPTYaYiU3KyZMnRUQkb968DlcCf1OhQgXJnj27TJs2TcLDw6VixYpy7do1Wbp0qURFRck777zjdInwQ/Hx8RISEnLLvHRgYKCEhobK0aNHJTo6WnLnzu1QhfBncXFxEhMTI/Hx8fLzzz/LyJEjRUSkZs2aDlcGf5aRPzcZtolNSEiQiRMnSqZMmaRRo0ZOlwM/kyNHDpk0aZL0799fXn31VdfjWbNmlXHjxkmdOnUcrA7+qkSJEvLll1/K/v37pXTp0q7H9+/fLxcvXhQRkT///JMmFrf1ySefyLvvvuvK4eHh8t///leqVKniYFXwdxn5c5Nhm9hhw4bJrl275PXXX5fixYs7XQ78UJYsWaRkyZJSu3ZtqVy5sly4cEHmzZsnvXr1kokTJ0qNGjWcLhF+pl27drJu3Tp59dVXpV+/fq4vdg0bNkyCg4Pl+vXrEhsb63SZ8FN16tSR4sWLy9WrV2Xfvn3y1VdfSXR0tNNlwc9l5M9NhtidwPThhx/KpEmT5JlnnpHBgwc7XQ780MGDB6Vly5by1ltvyXPPPed6PDY2Vho1aiSJiYmydu1aCQoKcrBK+KPPP/9chg4dKmfOnBERkaCgIGnRooVER0fL2rVrZfny5RIREeFwlbDBgQMHpEWLFtK9e3fp0qWL0+XAEhnpc5Mhvtjlbty4cTJp0iRp1qwZc41I0owZMyQuLk7q16+vHg8LC5NatWrJ8ePH5fjx4w5VB3/WoEED2bhxoyxbtkzmzp0rmzZtksGDB8vJkyclU6ZMUqxYMadLhCUiIiKkTJkyMm/ePKdLgUUy0ucmQ40TjB8/XsaPHy9NmzaVoUOHslk9knT69GkR+Wu/WNONGzfU/wKmoKAgNRN75swZ2b9/vzz00EPsbIEUuXbtmmueGrhTGeVzk2GuxI4fP17GjRsnTZo0kffee08CAzPMPzruwv333y8icsuWSJcuXZL169dLjhw5pGjRok6UBsvcvHlThgwZIgkJCdK1a1eny4Ef+nv0xPT999/L4cOHpWLFij6uCDbgc5NBZmLnzp0rgwcPlkKFCknPnj1vuQKbN29evqQD5fjx49KsWTO5ePGiNG7cWCpXriwXL16URYsWyfHjx2XgwIHSunVrp8uEn4mJiZGWLVtK3bp1pXDhwnL58mVZtWqVREVFyWuvvUYTi9t6+eWX5cyZM1K9enUpVKiQxMXFSVRUlHz22WcSGhoqs2fPVlf2ARE+NyIZpIl98803ZenSpUker1q1qsyePduHFcEGf/zxh0yYMEG2bNki586dk5CQECldurS0a9dO6tWr53R58EPx8fHyxhtvyO7du+XMmTMSFhYm5cuXl/bt20tkZKTT5cFPffbZZ7Js2TI5ePCgREdHS0BAgBQqVEhq1KghnTp1kkKFCjldIvwQn5sM0sQCAAAgfWEwFAAAANahiQUAAIB1aGIBAABgHZpYAAAAWIcmFgAAANahiQUAAIB1aGIBAABgHZpYAAAAWIcmFgAAANahiQUAAIB1aGIBAABgHZpYAAAAWIcmFgAAANahiQUAAIB1aGIBAABgHZpYAAAAWIcmFgAAANahiQUAAIB1aGIBAABgHZpYAAAAWIcmFgAAANahiQUAAIB1aGIBAABgHZpYAAAAWIcmFgAAANahiQUAAIB1aGIBAABgHZpYAAAAWCeT0wUAAABA++KLL1Q+efKkyocOHVJ5+PDhKicmJqocEBCgctOmTV3rF198UR2rX79+yop1CFdiAQAAYB2aWAAAAFiHJhYAAADWCUg0hyYAAIC1EhISVN6+fbvKderUUfnKlSsqr1y5UuVGjRqlYXVwd/bsWZXd52B79uypjl26dClF5zY/B0FBQUk+N3v27CpXrFhR5UWLFqmcN2/eFNXiLVyJBQAAgHVoYgEAAGAdmlgAAABYh31i78Dx48dd66NHj6pjlSpVUjk0NNQnNQEAcDvTp09X2dwD1GTuH7pt2zaVmYn1nr59+6o8Y8YM19rTDGtaM+dtN23apHLdunVVbt++vcrm/K6vcCUWAAAA1qGJBQAAgHVoYgEAAGAdr83EfvzxxyqvXr1a5UGDBqlcoUIFb5WSav/73/9c64EDB6pjffr0UXnw4MEqMyPrrF9++UXlkSNHJvncL7/8UuUsWbKo/Oijj6ocERGhco8ePVQ258xgr88///yuX1uoUCGVT5w44fH5e/bsca1/+ukndez3339XefPmzSo/99xzKs+bN+9Oy4Tl3D8L5s+l5NSqVUvlt99+Oy1KQjqye/dulf/73/+q3KRJE5Xvvfdeb5ckIlyJBQAAgIVoYgEAAGAdmlgAAABYx2szsS+88ILH4zt27FB51apVKpcrVy7Na/IGcy7EnIN8//33fVlOhjd58mSVhw4dqvKxY8fu+tzmfKJpzZo1Krvv1VigQIG7fl/4nnk/84YNG7rWKZ11DgkJUfnatWsqp+Xs9G+//ZZm54J/Mz9HnTt3dq0vXrzo8bXm30f9+/dXOTg4OJXVISkHDhxQ+dtvv3WoktQ5deqUyrNmzVLZ/P6Qt3AlFgAAANahiQUAAIB1aGIBAABgHa/NxL777rsqmzOwS5cuVdl9nkdEZMuWLd4p7C6476FnzrfFxcWpvHXrVl+UhP9jzuG8/PLLKt+8eTPJ15r72JlzYImJiSr//PPPHmsx9xL94YcfXOvGjRurY1euXFHZnAl3f62ISL9+/VTOmzevx1qQOl988cUdP9f8O6Fw4cIqm3sVlyxZUuU8efKo/OSTT7rWBw8eVMfMv0fN40i/rl69qnKXLl1U9vRZyJYtm8ruexGLiOTLly+V1eFOLV68WOXDhw+r7P5zJyEhweO5WrZsqXL37t1VNvc2N/eZLl68uMfzJ1XXndTmK1yJBQAAgHVoYgEAAGAdmlgAAABYx2szsea9l933zBS5dSbWnCXt1auXyqNGjUrD6lImMjLStR45cqQ6Zs6gfPPNNyqbs4zDhg1L4+oytiFDhqhszsCas6MLFixwrWvUqKGOhYaGejyXuQetOX9rGjFihGvt/hkSERk+fLjKye0nXL9+fZXr1avn8flIGfPPevXq1Uk+t3fv3iq3adNG5TJlyqhszrFWrVr1bkoUkVtndRs0aKCy+zwt0pdPP/1U5blz5yb53OzZs6u8cOFClZmB9Z0ZM2aobP7dHxQUpLL7rKl5zOT+M0ZEpFixYh6f/49//EPl9u3bu9azZ8/2+FpzBja52nyFK7EAAACwDk0sAAAArOO1cYKUMrdvMH9V0rVrV5VLlCjh9Zpup3Tp0iqbW5dcvnxZZXP8oFKlSiq3atUqDavLeJK7jWymTPoj7j5CYI4PmAID9f/Ha9q0qcrmaMjx48dV7tixo2v93HPPqWPmr4XN96pSpYrKaXlrUtzq/PnzKpu/fnVXtGhRlStUqODx3KkZHzCZv1IODw9XuXXr1mn2XnDW+vXrVX799dfv+LXmlpWMmTjnyJEjKpvbcvqS+TMvufGDlNi4caPKFy5cUDlnzpxp9l7uuBILAAAA69DEAgAAwDo0sQAAALCOz2ZimzRporJ5y0/zdminTp1S2dzeyKktt5544gmVzS2czJnY69evq2zeOhCpY870HDhwQOWTJ0+q7L69yRtvvKGOhYWFeXyvQoUKqWzOOprbxrnPxCbHvPUoty92ljmj7+7f//63z+p46623VJ42bZrK5hZbTn1XAKln/uzo27evyufOnfP4+ly5crnWPXr0SLvC4DfMvim573X4kvl9n3vuuccn78uVWAAAAFiHJhYAAADWoYkFAACAdXw2E5s7d26VixcvrrI5E2uaOnWqyn369HGtzVupIeNo3ry5yuZt+MyZ5Hfeece1Nuce3T9TIrfO9Ny4cUPlbt26qbxs2TKVPc1VBgcHq/zCCy8k+Vx4n7mfsDlb6n67Ym/td/g391ntDz/8UB3Lnz+/ysndrhj2WLlypco7d+5M0ev/97//udZFihRJk5rgX9q2batygQIFHKrkVjly5FDZ/DvVW7gSCwAAAOvQxAIAAMA6NLEAAACwjs9mYk3m/offfvutyub9ha9cuaKy+72gV6xYoY4VLlxY5aCgoLuuU0Tk0qVLt12LiCQkJKTq3EidIUOGqGzOsZqfM3eDBw9W2dyLuHbt2ipHR0er/OWXX95xnSZzBnbYsGF3fS6knjnP9eOPP6qcLVs2r733xIkTVR4zZoxrbc509+7dW+Xy5ct7rS54l/mz5IMPPvD4fHMv6RYtWqhs7mEO/2B+NyK5nsH9+WndX5jfPXL/jkhyUvrP4StciQUAAIB1aGIBAABgHZpYAAAAWMexmdg6deqoPHToUJXN2S/Tnj17XGvzfsLmPetTew/fH374wbX+/vvvU3Uuc363ffv2qToftPr166v8yy+/qPzxxx+71uaMz+nTp1V23xtURCQgIMBj9rQvbHh4uMrmfrbwL96cgZ0+fbrKr732msruc7CRkZHqWOfOnb1WF3xr1apVKptz2Kbq1aurPGnSpDSvCal38uRJlWfNmqVyct/RcZ81Te33eZKTkvObM7Deru1OcSUWAAAA1qGJBQAAgHVoYgEAAGAdx2ZiTa1atVLZ3M9s/Pjxd3wu93tI+xtzb9EtW7ao/PDDD/uynHSnUqVKKn/00Ucqu3/O/vjjjxSdO2/evCqbe9Ru3749ydfWrVtX5dTOacNeAwYMUNncC7ZNmzau9dixY9WxnDlzeq8weNWbb76p8tSpUz0+P1++fCqbs9PwT9euXVP56NGjDlVyq2XLljldQprjSiwAAACsQxMLAAAA69DEAgAAwDp+MxNbpEgRld3vHy4iUqVKFZXdZ8V27NjhvcIMZcuWVdncx3Hp0qUqnzp1SuWYmBiVR44cqfKnn36a2hLhgTmbmhL79u1Tef/+/Xf82qZNm971+8Iu5jy/Oe9//Phxlc051+HDhyd5DP7typUrKs+dO9e1njJlijp28eJFj+cyvwdSu3btVFaHjCa5PWtTo3Dhwir37Nkzzc6dElyJBQAAgHVoYgEAAGAdvxknMAUG6v66Xbt2Kjdu3Ni1vnDhgjpmbn10+fJllTdt2qSyedval19+WeUaNWq41tmzZ1fHzG2X6tWrp3KzZs3EkyVLlng8Dv9hjoKY2VS5cmXX2rwdLtKPM2fOqFyrVi2Vza3cSpQoobK5JWChQoXSrjj41Lp161Tu1q1bks81f3aYt0tv2LBh2hUGnzH/XN17FRGRlStX+qwWc5Rp7969aXbuF198UeUcOXKk2blTgiuxAAAAsA5NLAAAAKxDEwsAAADr+O1MbHJy585927VI8redPXDggMrm7f3y5Mlz13U98cQTKm/duvWuzwX/smLFihQ9v1GjRq51SEhIWpcDh5i3lTRvB2rOwD766KMqr169WuVs2bKlYXXwpc8//1zl9u3b3/Frze9LuG+tBnuZtxSvUKGCysnd+jUxMdG1TkhIUMcGDRrkMXs6l4hIQECAx+d7Yn4v6a233rrrc6UlrsQCAADAOjSxAAAAsA5NLAAAAKxj7UxsakRERHjt3OY+slWrVvXae8G7zP2ER4wYkaLXm/PRsJf7rWSbN2+uju3cuVPl/Pnzq2zebpQZWHtdvXpV5ffee0/lS5cuJfnatm3bqvzBBx+kXWHwW+YcalBQkMfnu8/BJvfc5JgztSk5n7nf7ZtvvpmqWryFK7EAAACwDk0sAAAArEMTCwAAAOtkyJlY4E6Ys47x8fEen2/OOpr7F8Ne7vcgNz8XFStWVHnx4sUq33///d4rDF5l7gncokULlTdv3uzx9e57jr/++uvqWFhYWCqrA9LOvffeq7L591ipUqV8WM2d40osAAAArEMTCwAAAOvQxAIAAMA6zMQCSZg9e3aKnv/cc8+pXK5cubQsBz5k7gm8bds217ps2bLq2OrVq1UuVKiQ9wqDV61fv17lli1bqnzhwgWPrzfnn7dv3+5a58iRI5XVwUY9e/ZU+bvvvlP5q6++8mU5ivsc7PLly9WxMmXK+Liau8OVWAAAAFiHJhYAAADWoYkFAACAdQISExMTnS4C8EelS5dW+cCBAx6fb+4jGxwcnOY1wTvOnz+vcsGCBVV239Pz8OHD6ph5j3HYw/xvNnv27B6Pm5599lmVp02bpnKWLFlSUR3So5MnT6r85Zdfqvztt9+61v/73/9S9V4JCQkqFyhQQOUvvvjCtTb3u7YFV2IBAABgHZpYAAAAWIcmFgAAANZhJhb4P1OnTlW5e/fuKic38xoTE+PxOPzX9OnTVe7UqZPKs2bNcq3btGnjk5rgfXFxcSq7zz6L3DrTOnDgQJV79+6tcmAg14UAX+K/OAAAAFiHJhYAAADW4bazwP/5+uuvVU5ue52+ffuqzPiAvfbu3evxOCME6VNISIjKN2/edKgSAHeDK7EAAACwDk0sAAAArEMTCwAAAOswEwv8nzp16qh84sQJlcuWLaty586dvV4TfOPxxx9X+cMPP1T5t99+c63vu+8+n9QEAPCMK7EAAACwDk0sAAAArEMTCwAAAOtw21kAAABYhyuxAAAAsA5NLAAAAKxDEwsAAADr0MQCAADAOjSxAAAAsA5NLAAAAKxDEwsAAADr0MQCAADAOjSxAAAAsE4mpwtwQmxsrDz11FNy/Phxad26tQwcONDpkuCHzp49K2PHjpWNGzfKuXPnJG/evFKnTh3p0aOHZM+e3eny4IfGjRsn48ePT/J4pkyZJCoqyocVwRZTpkyRqKgoiYqKkmPHjkl4eLh89dVXTpcFP3fz5k2ZNWuWLFiwQI4fPy65c+eWBg0aSI8ePSRLlixOl+d1GbKJHTt2rJw/f97pMuDHzp07J61atZLTp0/LM888IyVKlJDDhw/LggULZPv27TJ//nwJCwtzukz4mbp160rRokVvefzgwYPy8ccfy+OPP+5AVbDB6NGjJWfOnFKmTBm5fPmy0+XAEsOGDZPZs2dL3bp1pWPHjvLLL7/I7NmzZd++fTJjxgwJDEzfv3DPcE1sVFSUzJw5U/r06SPDhw93uhz4qcmTJ8vx48dl1KhR0qhRI9fjDzzwgPTq1UumT58uL730koMVwh9FRERIRETELY///dueFi1a+LokWGLdunVSpEgRERFp1KiRXL161eGK4O8OHz4sc+bMkXr16sm4ceNcjxcuXFiGDBkiq1evlsaNGztYofel7xbdkJCQIAMGDJDIyEipW7eu0+XAj23dulVCQ0PlqaeeUo83bNhQQkJCZMmSJQ5VBtvExsbK6tWrpUCBAhIZGel0OfBTfzewwJ1atWqVJCYmSrt27dTjrVq1krCwMFmxYoVDlflOhmpiZ8yYIb/++qsMGDDA6VLg5+Lj4yUkJEQCAgLU44GBgRIaGipHjx6V6Ohoh6qDTT7//HO5cuWKNGvWTIKCgpwuB0A68dNPP0lgYKBUqFBBPR4SEiIRERGyd+9ehyrznQzTxB49elTGjRsnL730khQuXNjpcuDnSpQoIRcvXpT9+/erx/fv3y8XL14UEZE///zTidJgmcWLF0tAQIA0b97c6VIApCOnT5+WXLlySebMmW85VqBAATl//rzEx8c7UJnvZJgmdtCgQVK4cGHp0KGD06XAAu3atZPAwEB59dVXZePGjXLixAnZuHGjvPrqqxIcHCwif/2aGPDk119/lR9//FGqV6/Or4sBpKnY2NjbNrAif12NFRG5du2aL0vyuQzxxa7ly5fLt99+pbN60wAAAtlJREFUK3PmzHE1IIAnVapUkdGjR8vQoUPlxRdfFBGRoKAgadGihURHR8vatWvlnnvucbhK+LvFixeLiEjLli0drgRAehMWFibnzp277bG4uDgREQkNDfVlST6X7pvY+Ph4GT58uNSsWVPy5csnR44cERGRU6dOiYjI5cuX5ciRI5IrVy72/oTSoEEDqVevnhw6dEhiYmLkvvvukzx58kiLFi0kU6ZMUqxYMadLhB+7ceOGLF++XHLmzMkXSQGkufz588vPP/8s8fHxt1yRPXXqVJKjBulJum9ir127JtHR0bJhwwbZsGHDLcdXrFghK1askL59+0qnTp18XyD8WlBQkJQuXdqVz5w5I/v375eHHnqIfWLh0ddffy1nz56V559/Pt3/IAHge+XKlZPNmzfLnj17pEqVKq7H4+Li5MCBA+qx9CrdN7FhYWEyZsyYWx6Pjo6Wd955RyIjI6VFixZSqlQpB6qDTW7evClDhgyRhIQE6dq1q9PlwM/9PUrA3rAAvKFhw4YyZcoUmTlzpmpYFy1aJLGxsel+j1iRDNDEBgcHS/369W95/NixYyIiUrRo0dseR8YWExMjLVu2lLp160rhwoXl8uXLsmrVKomKipLXXntNqlev7nSJ8GOnTp2STZs2SYUKFfg/yLgjy5YtkxMnTojIXxdZrl+/LhMnThQRkUKFCknTpk2dLA9+qFSpUtK6dWuZM2eOvPLKK1KzZk3XHbuqVq1KEwtkVMHBwVKqVClZuXKlnDlzRsLCwqR8+fIybdo0NqxHspYuXSoJCQl8oQt37NNPP5UffvhBPfb3bxGrVq1KE4vb6tevn4SHh8vChQtlw4YNkitXLmnTpo306NEj3d9yVkQkIDExMdHpIgAAAICUSP9tOgAAANIdmlgAAABYhyYWAAAA1qGJBQAAgHVoYgEAAGAdmlgAAABYhyYWAAAA1qGJBQAAgHVoYgEAAGCd/wcG7KD3iARQFgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 720x432 with 15 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(nrows=3, ncols=5, figsize=(10,6))\n",
    "for i in range(5):\n",
    "\n",
    "    # train\n",
    "    ax[0,i].imshow(train_images[i].reshape(28,28), cmap=plt.cm.binary)\n",
    "    ax[0,i].set_xlabel(np.argmax(train_labels[i]), fontsize=18)\n",
    "    ax[0,i].set_xticks([]); ax[0,i].set_yticks([]); ax[0,i].grid(False)\n",
    "    # val\n",
    "    ax[1,i].imshow(val_images[i].reshape(28,28), cmap=plt.cm.binary)\n",
    "    ax[1,i].set_xlabel(np.argmax(val_labels[i]), fontsize=18)\n",
    "    ax[1,i].set_xticks([]); ax[1,i].set_yticks([]); ax[1,i].grid(False)\n",
    "    # test\n",
    "    ax[2,i].imshow(test_images[i].reshape(28,28), cmap=plt.cm.binary)\n",
    "    ax[2,i].set_xlabel(np.argmax(test_labels[i]), fontsize=18)\n",
    "    ax[2,i].set_xticks([]); ax[2,i].set_yticks([]); ax[2,i].grid(False)\n",
    "    \n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Check for (un)balanced data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x288 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(nrows=1, ncols=3, figsize=(16,4))\n",
    "ax[0].hist(np.argmax(train_labels, axis=1).flatten()); ax[0].set_title('train', fontsize=18); ax[0].set_xticks(np.arange(10)); ax[0].tick_params(axis='both', which='major', labelsize=16);\n",
    "ax[1].hist(np.argmax(val_labels, axis=1).flatten()); ax[1].set_title('validation', fontsize=18); ax[1].set_xticks(np.arange(10)); ax[1].tick_params(axis='both', which='major', labelsize=16);\n",
    "ax[2].hist(np.argmax(test_labels, axis=1).flatten()); ax[2].set_title('test', fontsize=18); ax[2].set_xticks(np.arange(10)); ax[2].tick_params(axis='both', which='major', labelsize=16);\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Create architecture"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_1 (Conv2D)            (None, 28, 28, 32)        320       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 14, 14, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 14, 14, 64)        18496     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 7, 7, 64)          0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 7, 7, 64)          0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 3136)              0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 128)               401536    \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                1290      \n",
      "=================================================================\n",
      "Total params: 421,642\n",
      "Trainable params: 421,642\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "\n",
    "model.add(Conv2D(filters=32,\n",
    "                 kernel_size=3,\n",
    "                 strides=1,\n",
    "                 padding='same',\n",
    "                 activation='relu',\n",
    "                 input_shape=(img_rows, img_cols, 1)))\n",
    "\n",
    "model.add(MaxPooling2D(pool_size=2, strides=None))\n",
    "\n",
    "model.add(Conv2D(filters=64,\n",
    "                 kernel_size=3,\n",
    "                 strides=1,\n",
    "                 padding='same',\n",
    "                 activation='relu'))\n",
    "\n",
    "model.add(MaxPooling2D(pool_size=2, strides=None))\n",
    "\n",
    "model.add(Dropout(rate=0.40))\n",
    "\n",
    "model.add(Flatten())\n",
    "\n",
    "model.add(Dense(units=128, activation='relu'))\n",
    "\n",
    "model.add(Dense(units=num_classes, activation='softmax'))\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Compile and train model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 60000 samples, validate on 8000 samples\n",
      "Epoch 1/5\n",
      "60000/60000 [==============================] - 24s 403us/step - loss: 0.1903 - categorical_accuracy: 0.9410 - val_loss: 0.0438 - val_categorical_accuracy: 0.9844\n",
      "Epoch 2/5\n",
      "60000/60000 [==============================] - 20s 340us/step - loss: 0.0617 - categorical_accuracy: 0.9809 - val_loss: 0.0319 - val_categorical_accuracy: 0.9898\n",
      "Epoch 3/5\n",
      "60000/60000 [==============================] - 20s 338us/step - loss: 0.0446 - categorical_accuracy: 0.9859 - val_loss: 0.0270 - val_categorical_accuracy: 0.9906\n",
      "Epoch 4/5\n",
      "60000/60000 [==============================] - 21s 356us/step - loss: 0.0373 - categorical_accuracy: 0.9877 - val_loss: 0.0271 - val_categorical_accuracy: 0.9899\n",
      "Epoch 5/5\n",
      "60000/60000 [==============================] - 21s 348us/step - loss: 0.0304 - categorical_accuracy: 0.9903 - val_loss: 0.0193 - val_categorical_accuracy: 0.9931\n"
     ]
    }
   ],
   "source": [
    "# specify optimization strategy and metric used for monitoring during training\n",
    "model.compile(loss=categorical_crossentropy,\n",
    "              optimizer=Adadelta(),\n",
    "              metrics=['categorical_accuracy'])\n",
    "\n",
    "# the history object will contain a record of loss and metric values during training\n",
    "history = model.fit(train_images, train_labels,\n",
    "                    batch_size=128,\n",
    "                    epochs=5,\n",
    "                    verbose=1,\n",
    "                    validation_data=(val_images, val_labels))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Inspect learned kernels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1st convolution layer:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 10 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2nd convolution layer:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 10 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# model.layers will print a list of layer parameters/values\n",
    "filters1, biases1 = model.layers[0].get_weights()\n",
    "filters2, biases2 = model.layers[2].get_weights()\n",
    "\n",
    "# normalize filter values to range 0-1 for better colormapping during plotting\n",
    "def norm_filter(kernel):\n",
    "    return (kernel - np.min(kernel)) / (np.max(kernel) - np.min(kernel))\n",
    "\n",
    "print('1st convolution layer:')\n",
    "fig, axs = plt.subplots(2,5, figsize=(10, 6))\n",
    "axs = axs.ravel()\n",
    "for i in range(10):\n",
    "    axs[i].imshow(norm_filter(filters1[:,:,0,i]), cmap=plt.cm.binary)\n",
    "    axs[i].set_xticks([]); axs[i].set_yticks([]); axs[i].grid(False)\n",
    "plt.show()\n",
    "\n",
    "print('2nd convolution layer:')\n",
    "fig, axs = plt.subplots(2,5, figsize=(10, 6))\n",
    "axs = axs.ravel()\n",
    "for i in range(10):\n",
    "    axs[i].imshow(norm_filter(filters2[:,:,0,i]), cmap=plt.cm.binary)\n",
    "    axs[i].set_xticks([]); axs[i].set_yticks([]); axs[i].grid(False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Evaluate training process"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x864 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# evaluating model using all data (not in batches)\n",
    "val_loss, val_acc = model.evaluate(val_images, val_labels, verbose=2)\n",
    "\n",
    "fig,ax = plt.subplots(nrows=2,ncols=1,figsize=(12,12))\n",
    "fs_L, fs_M, fs_S = 18, 16, 14\n",
    "ax[0].plot(history.history['categorical_accuracy'], label='train')\n",
    "ax[0].plot(history.history['val_categorical_accuracy'], label='validation')\n",
    "ax[0].set_xlabel('Epoch', fontsize=fs_M)\n",
    "ax[0].set_ylabel('Accuracy', fontsize=fs_M)\n",
    "ax[0].tick_params(axis='both', which='major', labelsize=fs_S)\n",
    "ax[0].set_title('Final mean validation accuracy: {}'.format(val_acc), fontsize=fs_M)\n",
    "ax[0].set_xticks(range(0,5))\n",
    "ax[0].legend(loc='lower right', fontsize=fs_M)\n",
    "\n",
    "ax[1].plot(history.history['loss'], label='train')\n",
    "ax[1].plot(history.history['val_loss'], label='validation')\n",
    "ax[1].set_xlabel('Epoch', fontsize=fs_M)\n",
    "ax[1].set_ylabel('Loss', fontsize=fs_M)\n",
    "ax[1].tick_params(axis='both', which='major', labelsize=fs_S)\n",
    "ax[1].set_xticks(range(0,5))\n",
    "ax[1].legend(loc='upper right', fontsize=fs_M)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Inference"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# using until now unseen data\n",
    "predicted_prob = model.predict(test_images)\n",
    "predictions = np.argmax(predicted_prob, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "confusion = tf.confusion_matrix(labels=np.argmax(test_labels, axis=1), predictions=predictions, num_classes=num_classes)\n",
    "sess = tf.InteractiveSession()\n",
    "conf_matrix = confusion.eval(session=sess)\n",
    "sess.close()\n",
    "\n",
    "# accuracy score for inference\n",
    "error_rate = (np.sum(conf_matrix)-np.sum(np.diag(conf_matrix))) / np.sum(np.diag(conf_matrix))\n",
    "inf_acc = 1-error_rate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Correct: 1985/2000\n",
      "Wrong: 15/2000\n"
     ]
    }
   ],
   "source": [
    "plt.figure(figsize=(10,10))\n",
    "ax = sns.heatmap(conf_matrix, annot=True, annot_kws={\"size\": 16}, fmt=\"d\", linewidths=.5, square=True, cbar=False, cmap='Greens')\n",
    "ax.invert_yaxis()\n",
    "plt.ylabel('Actual label', fontsize=16)\n",
    "plt.xlabel('Predicted label', fontsize=16)\n",
    "plt.xticks(fontsize=14); plt.yticks(fontsize=14)\n",
    "plt.title('Accuracy Score: {:.4f}'.format(inf_acc), fontsize=16)\n",
    "plt.show()\n",
    "\n",
    "print('Correct: {0}/{1}'.format(np.sum(np.diag(conf_matrix)),np.sum(conf_matrix)))\n",
    "print('Wrong: {0}/{1}'.format((np.sum(conf_matrix)-np.sum(np.diag(conf_matrix))),np.sum(conf_matrix)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Inspection of predictions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Correctly predicted images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x504 with 10 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# show first 10 images that were correctly predicted\n",
    "correct_idx = np.where(predictions == np.argmax(test_labels, axis=1))[0]\n",
    "\n",
    "fig, axs = plt.subplots(2,5, figsize=(15, 7))\n",
    "axs = axs.ravel()\n",
    "for i in range(10):\n",
    "    axs[i].imshow(test_images[correct_idx[i],:,:,0], cmap=plt.cm.binary)\n",
    "    axs[i].set_title('predicted: {} \\n actual: {}'.format(predictions[correct_idx[i]], np.argmax(test_labels, axis=1)[correct_idx[i]]), fontsize=18)\n",
    "    axs[i].set_xticks([]); axs[i].set_yticks([]); axs[i].grid(False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Wrongly predicted images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x504 with 10 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# show first 10 images that were wrongly predicted\n",
    "wrong_idx = np.where(predictions != np.argmax(test_labels, axis=1))[0]\n",
    "\n",
    "fig, axs = plt.subplots(2,5, figsize=(15, 7))\n",
    "axs = axs.ravel()\n",
    "for i in range(10):\n",
    "    axs[i].imshow(test_images[wrong_idx[i],:,:,0], cmap=plt.cm.binary)\n",
    "    axs[i].set_title('predicted: {} \\n actual: {}'.format(predictions[wrong_idx[i]], np.argmax(test_labels, axis=1)[wrong_idx[i]]), fontsize=18)\n",
    "    axs[i].set_xticks([]); axs[i].set_yticks([]); axs[i].grid(False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Final note: This reflects a very simple and crude model. Since MNIST is a relatively straightforward dataset, using a \"proper\" CNN, you should expect even better results!"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.7.3 64-bit ('base': conda)",
   "language": "python",
   "name": "python37364bitbaseconda6bcb047ec2594927be6c6fcc220abba4"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}