Digit-Recogniser.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Project Overview\n",
    "I built a Convolutional Neural Network with keras to build identify handwritten digits.\n",
    "\n",
    "References:\n",
    "\n",
    "https://www.kaggle.com/kanncaa1/convolutional-neural-network-cnn-tutorial\n",
    "\n",
    "https://www.kaggle.com/poonaml/deep-neural-network-keras-way\n",
    "\n",
    "https://www.kaggle.com/yassineghouzam/introduction-to-cnn-keras-0-997-top-6/notebook"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np # linear algebra\n",
    "import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.image as mpimg\n",
    "import seaborn as sns\n",
    "%matplotlib inline\n",
    "\n",
    "np.random.seed(2)\n",
    "\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import confusion_matrix\n",
    "import itertools\n",
    "\n",
    "from keras.utils.np_utils import to_categorical # convert to one-hot-encoding\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Flatten, Conv2D, MaxPool2D\n",
    "from keras.optimizers import RMSprop\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from keras.callbacks import ReduceLROnPlateau\n",
    "\n",
    "sns.set(style='white', context='notebook', palette='deep')\n",
    "\n",
    "import os"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data Preparation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "train = pd.read_csv(r\"C:\\Users\\Weston\\workspace\\Digit-Recognition\\train.csv\")\n",
    "test = pd.read_csv(r\"C:\\Users\\Weston\\workspace\\Digit-Recognition\\test.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "#from sklearn.datasets import fetch_openml\n",
    "#mnist = fetch_openml(\"mnist_784\", version = 1)\n",
    "#mnist.keys()\n",
    "#dictkeys(['data', 'target', 'feature_names', 'DESCR', 'details', 'categories', 'url'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1    4684\n",
       "7    4401\n",
       "3    4351\n",
       "9    4188\n",
       "2    4177\n",
       "6    4137\n",
       "0    4132\n",
       "4    4072\n",
       "8    4063\n",
       "5    3795\n",
       "Name: label, dtype: int64"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "Y_train = train[\"label\"]\n",
    "\n",
    "# Drop 'label' column\n",
    "X_train = train.drop(labels = [\"label\"],axis = 1) \n",
    "\n",
    "# free some space\n",
    "#del train \n",
    "\n",
    "g = sns.countplot(Y_train)\n",
    "\n",
    "Y_train.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count       784\n",
       "unique        1\n",
       "top       False\n",
       "freq        784\n",
       "dtype: object"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.isnull().any().describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count       784\n",
       "unique        1\n",
       "top       False\n",
       "freq        784\n",
       "dtype: object"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test.isnull().any().describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Grayscale normalization to help clean the images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Normalize the data\n",
    "X_train = X_train / 255.0\n",
    "test = test / 255.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Reshape image in 3 dimensions (height = 28px, width = 28px , canal = 1)\n",
    "# the extra dimension is required by keras and corresponds to channels\n",
    "\n",
    "X_train = X_train.values.reshape(-1,28,28,1)\n",
    "test = test.values.reshape(-1,28,28,1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "#one hot encoding the labels\n",
    "Y_train = to_categorical(Y_train, num_classes = 10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "random_seed = 2\n",
    "\n",
    "# Split the train and the validation set for the fitting\n",
    "X_train, X_val, Y_train, Y_val = train_test_split(X_train, Y_train, test_size = 0.1, random_state=random_seed)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "g = plt.imshow(X_train[0][:,:,0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Define the model\n",
    "I used the Keras Sequential API, where you have just to add one layer at a time, starting from the input.\n",
    "\n",
    "The first is the convolutional (Conv2D) layer.\n",
    "\n",
    "It is like a set of learnable filters. I choosed to set 32 filters for the two firsts conv2D layers and 64 filters for the two last ones. Each filter transforms a part of the image (defined by the kernel size) using the kernel filter. The kernel filter matrix is applied on the whole image. Filters can be seen as a transformation of the image. The CNN can isolate features that are useful everywhere from these transformed images (feature maps).\n",
    "\n",
    "The second important layer in CNN is the pooling (MaxPool2D) layer.\n",
    "\n",
    "This layer simply acts as a downsampling filter. It looks at the 2 neighboring pixels and picks the maximal value. These are used to reduce computational cost, and to some extent also reduce overfitting. We have to choose the pooling size (i.e the area size pooled each time) more the pooling dimension is high, more the downsampling is important.\n",
    "\n",
    "Combining convolutional and pooling layers, CNNs are able to combine local features and learn more global features of the image.\n",
    "\n",
    "Dropout is a regularization method, where a proportion of nodes in the layer are randomly ignored (setting their wieghts to zero) for each training sample. This drops randomly a proportion of the network and forces the network to learn features in a distributed way. This technique also improves generalization and reduces the overfitting.\n",
    "\n",
    "'relu' is the rectifier (activation function max(0,x). The rectifier activation function is used to add non linearity to the network.\n",
    "\n",
    "The Flatten layer is use to convert the final feature maps into a one single 1D vector. \n",
    "\n",
    "This flattening step is needed so that you can make use of fully connected layers after some convolutional/maxpool layers. It combines all the found local features of the previous convolutional layers.\n",
    "\n",
    "In the end I used the features in two fully-connected (Dense) layers which is just artificial an neural networks (ANN) classifier. In the last layer(Dense(10,activation=\"softmax\")) the net outputs distribution of probability of each class."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set the CNN model \n",
    "# my CNN architechture is In -> [[Conv2D->relu]*2 -> MaxPool2D -> Dropout]*2 -> Flatten -> Dense -> Dropout -> Out\n",
    "\n",
    "model = Sequential()\n",
    "\n",
    "# stage 1\n",
    "model.add(Conv2D(filters = 32, kernel_size = (5,5),padding = 'Same', \n",
    "                 activation ='relu', input_shape = (28,28,1)))\n",
    "model.add(Conv2D(filters = 32, kernel_size = (5,5),padding = 'Same', \n",
    "                 activation ='relu'))\n",
    "model.add(MaxPool2D(pool_size=(2,2)))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "# stage 2\n",
    "model.add(Conv2D(filters = 64, kernel_size = (3,3),padding = 'Same', \n",
    "                 activation ='relu'))\n",
    "model.add(Conv2D(filters = 64, kernel_size = (3,3),padding = 'Same', \n",
    "                 activation ='relu'))\n",
    "model.add(MaxPool2D(pool_size=(2,2), strides=(2,2)))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "# stage 3\n",
    "model.add(Flatten())\n",
    "model.add(Dense(256, activation = \"relu\"))\n",
    "model.add(Dropout(0.5))\n",
    "model.add(Dense(10, activation = \"softmax\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define the optimizer\n",
    "optimizer = RMSprop(lr=0.001, rho=0.9, epsilon=1e-08, decay=0.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Compile the model\n",
    "model.compile(optimizer = optimizer , loss = \"categorical_crossentropy\", metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A high learning rate uses bigger step sizes, but it will reduce the accuracy of our model and the optimizer could fall into a local minima. However step sizes that are too small will make our model train too  slowly. \n",
    "\n",
    "To solve this I reduced the learning rate, if the model's accuracy is not improved after 3 epochs."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set a learning rate annealer\n",
    "learning_rate_reduction = ReduceLROnPlateau(monitor='val_accuracy', \n",
    "                                            patience=3, \n",
    "                                            verbose=1, \n",
    "                                            factor=0.5, \n",
    "                                            min_lr=0.00001)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "epochs = 2 # Turn epochs to 30 to get 0.9967 accuracy\n",
    "batch_size = 86"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data Augmentation\n",
    "Data augmentation expands our training data set size and reduces the chance of overfitting.\n",
    "\n",
    "To achieve this I:\n",
    "* randomly rotated some images by 10 degrees\n",
    "* randomly zoomed some images by 10%\n",
    "* randomly shifted images horizontally or vertically by 10% \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "datagen = ImageDataGenerator(\n",
    "        featurewise_center=False,  # set input mean to 0 over the dataset\n",
    "        samplewise_center=False,  # set each sample mean to 0\n",
    "        featurewise_std_normalization=False,  # divide inputs by std of the dataset\n",
    "        samplewise_std_normalization=False,  # divide each input by its std\n",
    "        zca_whitening=False,  # apply ZCA whitening\n",
    "        rotation_range=10,  # randomly rotate images in the range (degrees, 0 to 180)\n",
    "        zoom_range = 0.1, # Randomly zoom image \n",
    "        width_shift_range=0.1,  # randomly shift images horizontally (fraction of total width)\n",
    "        height_shift_range=0.1,  # randomly shift images vertically (fraction of total height)\n",
    "        horizontal_flip=False,  # randomly flip images\n",
    "        vertical_flip=False)  # randomly flip images\n",
    "\n",
    "datagen.fit(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/2\n",
      "439/439 - 79s - loss: 0.4132 - accuracy: 0.8673 - val_loss: 0.0565 - val_accuracy: 0.9824\n",
      "Epoch 2/2\n",
      "439/439 - 80s - loss: 0.1245 - accuracy: 0.9640 - val_loss: 0.0660 - val_accuracy: 0.9805\n"
     ]
    }
   ],
   "source": [
    "# Fit the model\n",
    "\n",
    "history = model.fit(datagen.flow(X_train,Y_train, batch_size=batch_size),\n",
    "                              epochs = epochs, validation_data = (X_val,Y_val),\n",
    "                              verbose = 2, steps_per_epoch=X_train.shape[0] // batch_size\n",
    "                              , callbacks=[learning_rate_reduction])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model Evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the loss and accuracy curves for training and validation \n",
    "fig, ax = plt.subplots(2,1)\n",
    "ax[0].plot(history.history['loss'], color='b', label=\"Training loss\")\n",
    "ax[0].plot(history.history['val_loss'], color='r', label=\"validation loss\",axes =ax[0])\n",
    "legend = ax[0].legend(loc='best', shadow=True)\n",
    "\n",
    "ax[1].plot(history.history['accuracy'], color='b', label=\"Training accuracy\")\n",
    "ax[1].plot(history.history['val_accuracy'], color='r',label=\"Validation accuracy\")\n",
    "legend = ax[1].legend(loc='best', shadow=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Below is a confusion matrix of our predictions vs the real identity of the number"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Predict the values from the validation dataset\n",
    "Y_pred = model.predict(X_val)\n",
    "# Convert predictions classes to one hot vectors \n",
    "Y_pred_classes = np.argmax(Y_pred,axis = 1) \n",
    "# Convert validation observations to one hot vectors\n",
    "Y_true = np.argmax(Y_val,axis = 1) \n",
    "# compute the confusion matrix\n",
    "confusion_mtx = confusion_matrix(Y_true, Y_pred_classes) \n",
    "# plot the confusion matrix\n",
    "f,ax = plt.subplots(figsize=(8, 8))\n",
    "sns.heatmap(confusion_mtx, annot=True, linewidths=0.01,linecolor=\"gray\", fmt= '.1f',ax=ax)\n",
    "\n",
    "plt.xlabel(\"Predicted Label\")\n",
    "plt.ylabel(\"True Label\")\n",
    "plt.title(\"Confusion Matrix\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Some of our model's mistakes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXIAAAEUCAYAAAA2ib1OAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjMsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+AADFEAAAgAElEQVR4nO3deVyU1f4H8M9s7KKiLC6IXU3NUCxTwxRSyw1xSftlt9SyxMrUtMUlbhSKkllkmubWtUz75XXXq9w00xLM/QpoppkDw+IwghvbrOf3Bz+fZ87owADDwAPf9+vl63UO51nOzBcPz5w5i4wxxkAIIUSy5HVdAUIIITVDDTkhhEgcNeSEECJx1JATQojEUUNOCCESRw05IYRIXJ035NnZ2XjooYcwatQo4d/IkSOxdevWGl976tSp2L59OwBg1KhRuH37tt1j79y5g4kTJ1b5HsnJyZgwYcI9Pz9+/DhGjBhR6fmdO3dGYWFhle45d+5crF+/vsJjdu7cyb2nAwcOxMMPP4zr169X6V7VRXGtnbgCwB9//IEJEyZg9OjReOaZZ5CRkVGl+9QExbX24pqYmIgnn3xSeF/feusth++hrFKNaomHhwd27dol5LVaLUaMGIHQ0FB06dLFKfewvv793Lp1C+np6U65V30wevRojB49GgBgNBrx4osvIiYmBi1btnRZHSiuzldaWopXXnkFCQkJiIyMxMGDB/HOO+8gOTnZZXWguNaOs2fP4rPPPsOjjz5a5XPrRUNuKzAwECEhIVCr1bhw4QK2bt2K0tJS+Pj4YOPGjfjXv/6F77//HhaLBc2aNcM//vEPdOjQAVqtFnPnzkV+fj5at26NgoIC4ZqdO3fGsWPH4Ofnh9WrV2PHjh1QKpUICQlBYmIi5s2bh7KyMowaNQrbt2+HWq1GQkICbt68CbPZjAkTJmDcuHEAgGXLlmHPnj1o1qwZQkJCKn09V69eRXx8PIqLi6HT6dClSxd8/vnncHd3BwB8/vnnSE9Ph8ViwVtvvYUBAwYAgN3XaU96ejpiY2Pv+U+wdu1a+Pn5Yfz48VWOhTNRXGse15SUFAQHByMyMhIAMGjQILRt27baMXEGimvN42owGHDhwgWsW7cOGo0G7du3x7x589C6dWvHgsDqmEajYT169OB+dubMGdarVy+Wm5vLtm3bxnr16sXu3LnDGGPs+PHj7O9//zsrKSlhjDH266+/sqFDhzLGGHvjjTdYUlISY4wxtVrNevTowbZt28YYY6xTp06soKCAHTx4kA0ePJjdvHmTMcbYokWL2MqVK7l6GI1GNnz4cJaRkcEYY+z27dts2LBh7OzZs+zAgQNs+PDh7M6dO8xoNLKYmBj24osv3vO6fvvtNxYVFcUYYywxMZHt3LmTMcaYwWBgI0aMYMnJyUK9Vq9ezRhj7I8//mC9e/dmBQUFFb7OOXPmsHXr1jn0/hYUFLDHHnuMZWVlOXS8s1Bcayeua9asYdOnT2fz5s1jY8aMYZMmTRJejytQXGsnrllZWezVV19lf/zxB7NYLGzt2rVs1KhRzGKxOBAVxurFE/ndv6wAYDab0bx5c3zyySdo1aoVgPK/zj4+PgCAw4cPIzMzk3u6vH37Nm7evInU1FTMmTMHABASEoI+ffrcc69jx45h6NChaNq0KQBg3rx5AMr7/u5Sq9XIysrC/PnzuTpeuHABV65cwdNPPy3UZ+zYsdi4cWOFr+/dd99FSkoK1q5dC7Vajfz8fJSUlAjlzz//PACgU6dO6NChA86ePYvTp0/bfZ1VsWXLFgwaNAjBwcFVOs8ZKK7Oj6vJZMKRI0fw7bffIiwsDAcPHkRMTAx+/vlnuLm5OXSNmqK4Oj+uwcHBWLt2rZB/5ZVXsHLlSmRnZzv0f7deNOS2fW62vLy8hLTFYsGoUaPw7rvvCvn8/Hw0bdoUMpkMzGrpGKXy3penUCggk8mE/O3bt+/5UsVsNqNJkyZcna5fv44mTZpgyZIl3D0UCkWlr2/27Nkwm80YNmwYnnzySeTl5XHXkMvF75wtFguUSmWFr7Mq9u3bh9jY2Cqd4ywUV+fHNSAgAB06dEBYWBgA4KmnnkJsbCw0Gk2FH+OdieLq/LhevHgRFy9eFL7XAgDGGFQqlUPn1/molarq168f/v3vfyM/Px8A8P3332PSpEkAgP79++OHH34AAOTm5uL48eP3nN+3b18cOHAARUVFAIDly5djw4YNUCqVMJvNYIzhgQce4H5Z8/LyMGLECGRkZCAiIgLJycm4ffs2LBZLpV/KAMDRo0cxbdo0DB8+HABw7tw5mM1moXzHjh0AgPPnzyMrKwthYWEVvk5H3bp1C1lZWXjkkUeqdF5doLg6JiIiAtnZ2cJIlZMnT0Imk9V5P7k9FFfHyOVyJCQkQKPRAAA2b96Mzp07IygoyKHz68UTeVX069cPU6ZMweTJkyGTyeDj44MVK1ZAJpMhLi4O8+bNw7BhwxAUFHTfb9AjIyPx559/Ch+POnbsiAULFsDT0xPdu3dHVFQUNm3ahJUrVyIhIQHr1q2DyWTCzJkz0bNnTwDlw7/Gjh0LX19fdOnSBTdu3KiwzrNmzcK0adPg5eUFHx8f9OrVC1lZWUK5RqPB6NGjIZPJ8Nlnn6FZs2YVvk57bL/szMzMhL+/v8N/1esSxdWxuPr7++PLL7/ERx99hNLSUri5uWH58uXCF3H1DcXVsbh26tQJsbGxeP3112E2mxEUFITPPvvM4fdZxhgtY0sIIVImua4VQgghPGrICSFE4qghJ4QQiZNkQ75w4UJhPYLQ0FAMGTJEyJeVlTn9ftnZ2Q6N/Bg4cGCVpw0vX74c8fHxlR63b98+REVFYeTIkYiJiUFOTk6V7iMFjTGud6WlpSE0NLTK63hIQWOM6/79+zFy5EhER0dj4sSJUKvVVbpPVUlu1AoAblz0wIEDsXTpUnTr1q0Oa1S71Go14uLi8N1336Fz5844efIkZsyYgW3bttV11ZyqscX1rsLCQnz44YcwGo11XZVa0djiqtPpEBcXh927dyMoKAjfffcdFixY4NDCWdUlySfyyoSGhmLmzJkYMmQI0tPT71mxzDp/6NAhPPvssxg9ejTGjx+Ps2fPVnjt69ev44033sBzzz2HgQMHYsKECdwaEZs3b8aYMWMQFRXFrQhX1ftotVqMGjUKWq0WFy9eRJcuXdC5c2cAQK9evZCTk8PNbmsMGlpcgfKJI++++y5mzZpV5fejoWhocfX390dKSgqCgoJgMpmQk5ODZs2aVeetcZgkn8grYzQaMWDAACxbtqzC49RqNZKSkvDtt9+iefPmuHz5Ml5++WX8+OOP3Ow0a//+97/Ro0cPxMTEgDGGmJgY7Nq1C5MnTwYAuLu7Y8eOHdBqtRgzZgzCwsKgUqns3seewMBAYTx4165dcenSJfz+++946KGHcOjQIdy8eRM6na7eTgSpDQ0trkD5gk7du3dH//79q/GONAwNMa4qlQrp6el47bXXUFZWhnXr1lXjnXFcg2zIAeCxxx6r9JiUlBTk5+fjpZdeEn4mk8mQlZVldznOSZMm4dSpU/jnP/8JtVqNy5cvC9OlAQhrLQQGBuKJJ57AsWPHoFAo7N7HEe3atcOiRYsQFxcHg8GAQYMGoUuXLpKY6ONsDSmuhw8fRlpaWq1+5JaKhhTXu7p164aUlBT88ssvmDp1Kg4ePAhfX98qXcNRDbYht/cX2mAwCGmLxYLw8HB8/vnnws/y8vIQEBBg97qffPIJ0tLSMHbsWPTp0wcmk6nSdRjMZrPd+xw4cKDS12IwGBASEoItW7YI+W+++aZRPY3f1ZDium3bNly7dg1jxowRfjZp0iQsWrSoQfch309DiqtWq8WlS5eET1kRERHw8fFBVlYWQkNDKz2/OhpkH7ktPz8/4dvpvXv3Cj8PDw9HSkoKrly5AgA4cuQIRo4cWeE36UePHsWkSZMwevRotGjRAqmpqfddhyE3NxfHjh1DeHh4te5jzWAw4Pnnn0deXh4AYMOGDejZs2et97vVd1KP6/Lly7F//37s2rVL+Fj+zTffNLpG3JbU42owGDB79mxkZmYCAH777TeYTKZaXdSswT6RW4uNjUV8fDx8fX3Rt29f+Pv7AyhftyE+Ph6zZ88GYwxKpRKrVq2Ct7e33WtNmzYNS5YswbJly6BSqfDoo49yH7n0ej3GjBkDo9GI2NhYPPDAAwBQ5ftotVrExMRgzZo1CAwMxIIFCzBlyhSYzWZ06NABixcvdtK7I10NIa7kXlKPa3BwMBYuXIjp06dDJpPB19cXX331FTw9PZ30Dt2L1lohhBCJaxRdK4QQ0pBRQ04IIRJHDTkhhEhcjb7s3LNnD1atWgWTyYRJkybhhRdeqPScsrIyZGRkwN/f36Ftl0jtMJvN0Ol0CA0NhYeHR42vR3GtP5wZW4pr/VFRXKvdkGu1WiQlJWH79u1wc3PD+PHj0adPH3Ts2LHC8zIyMhxq8IlrbNq0yaHJGJWhuNY/zogtxbX+uV9cq92Qp6am4vHHHxfGMg8ZMgTJycl48803Kzzv7lCi7JximMw0YKauKBUytG3jLcSjpiiu9YczY0txrT8qimu1G/L8/HzuggEBAUhLS6v0vLsfz0xmBpOJfjHqmrM+LlNc6x9nxJbiWv/cL67V/rLTYrFwG4syxircaJQQQkjtqHZDHhQUBJ1OJ+R1Ol2Fax4QQgipHdVuyPv27Ytjx46hsLAQpaWl+PHHHxEREeHMuhFCCHFAtfvIAwMDMWvWLEycOBFGoxHjxo1D9+7dnVk3QgghDqjROPLo6GhER0c7qy6EuERWz85cPmCPuOh/Us8PuLJ5eT+7pE6E1ATN7CSEEImjhpwQQiSOGnJCCJG4RrGxBCHWmr/Wm8szs0lI/4+fliubl+eSKhFSI/RETgghEkcNOSGESBx1rZBGoamH1X6L3k3sHrfkpp8LakOIc9ETOSGESBw15IQQInHUkBNCiMRRH3kd6NlS3EWpvao5V/awzEdIx+fS9HBnmdiip5BW9n/W7nE/Fv3piuoQF3iudR+7ZW8Z+bXVw47OF9KGZfFc2eZtzYT0KtNfXFl6gboGNXQeeiInhBCJo4acEEIkjrpWqqmPP7+CXhcVP2zt014FQjr3rDdX9rfNL4mZ69lcGTPohXT8/1DXiisYvnhfSF8ruVGHNSEA4KZQcfkAr6ZCushYxpXta/KwkO62vBdXprLqQmMWi8P3d5/3KZef/L74vNsv/C2urDt1rRBCCHEGasgJIUTiqCEnhBCJoz7yCoT6hXD5Q73Fvjvv+Le5MnlQBy7PigqFdIcivt9V0UbsX7+9aAVXVpqnqF5lSbWVnMwX0qVGfQVHElfI7M//v2u2ca2QNqfz3xspug1wyj2NGxYLadVL85xyTVeiJ3JCCJE4asgJIUTiqGvFRkuroU6ps/ghhm6T37c9XGD66wyXtxzYI6TXrOGPXXTrpJC+YyjlyoxWmxwQ0lj8+ZA4jLD5ppVcmfXQQduuFPNvu4T0gpn/5cpSTGKXGQM/k/ObFvwzbMCLwXbrZkz+Wkh/YvKwe1xdoidyQgiROGrICSFE4qghJ4QQiZNUH3n7poFcXn1La+dIx9mukLb2dbGP3LZP3HLjmpAuS/iQK3s8+TaXv3pbrFt1+73ntH6Syz9jKhHSffJPVOuahNRHmVpxhcG2FRxnPUwQAB777LyQvnQjx+55xwP4DbdbbXiZy8tbdxHSpmM7uLJus5OFtDPanNpAT+SEECJxDjXkRUVFGDFiBLKzyxd4Sk1NRXR0NAYPHoykpKRarSAhhJCKVdq1cu7cOcTGxkKtVgMAysrKMH/+fGzcuBGtWrXC1KlTceTIEURGRtZ2XdHSzZfLq1G9jznWGzt8veMlrkzRtquQPvfIu1xZklImpHdfv8qVlRj4Vdkc1aFZKy6f+pi4UqLv6ve4MsNXC8XMJ9W6XaO18HljXVeBVGBQYYqQLpPxz5fW2Rvbs7iyI2GeQtp7yDCuzG2SuFkEGL/6ofHID1w+I/qfQlqK3ZaVPpFv2bIFcXFxCAgIAACkpaUhJCQEwcHBUCqViI6ORnJyciVXIYQQUlsqfSJPSEjg8vn5+fD39xfyAQEB0Grr5xcAhBDSGFT5y06LxQKZTOxiYIxxeUIIIa5V5eGHQUFB0Ol0Ql6n0wndLrXtlO5ytc57zP9BLv+1p7jBMSsr5sr2dftASD93i592X91hhLb94D4KcZpvC6UPV+Y9dZCQlnnwOwu5vyUOvZq5+R9c2bKcX6pVt8ZC1qKZ3bJxF1R2y4jrTX6M/25q7Y5JQjpg9yqHr3M5fLqQnm/gVxX99cYfXP5mWVFVqljvVPmJPCwsDFevXkVmZibMZjP27t2LiIiI2qgbIYQQB1T5idzd3R2JiYmYPn069Ho9IiMjMXTo0NqoGyGEEAc43JAfOnRISIeHh2P37t21UiFn6daivZA+NOMBrsySLa6K1iGK/zK3sPSOkK7JSoTvtBaHY54wX+fK3GTix7w+cn7TZvnDTwhp87UrXFn2858J6a+v/1XtujUGCjn/UVrm7i6m5fwH0TILDU2sTzbn/sblV20V4+U2I8H2cLv82ojdpu0z+e7NPRLvSrFFMzsJIUTiqCEnhBCJo4acEEIkTlKrH1akhRc/fT816Skhzf7ip9OnbxWH/2ltNkZ2VKBPcy7f3acdl29i9TfyF+15rmy7n9h/PvTYW1yZ9XDI8SO+5Mr25GVUq66N0XNBvbi88tmZQtq4ZzVXllmiQ3XY9sO39xWH4erNfL979h3+exLiuDbLzgrped/Fc2WvPymuSOoWw69o2OJf64T0EptNtRclb+Dyn8bnCeml1/k++iKbXbzqI3oiJ4QQiaOGnBBCJK7BdK1s9ujO5VWDJoiZQfyx3UvFDSM2rec3c92oFLtaTt3mu2Sul9wS0s82DeXK1mqPc/mLVps492j5N67s6QRx6XxLPn+PY6O2Cuk9hadBnM/wMz9j1zquVWHdlQIAGembhbT5Mv/70P9/1grps9f5YaWkYtZdG+/n/cyVvf+9mG6/71Ou7AXvh4T0/D0TuTJl1BQuPydKTM+YO40rW/izuLbUZ7lHHKu0i9ETOSGESBw15IQQInHUkBNCiMQ1mD7y5e4lXL5f9gUhbdr9v1xZy6Xi8KLwlp25sl+uiUMFIwP5fvAjVn2pK3J+rbA+rd3Fqfc//7KIK5N5NhHSab1jubKnC6W3O0lj0rFZayE9yPtvdo8z797O5alfvPbZboycYJVP6HmYK5vS+gkun7Syr5D2WsKvsLjIaouikd3mcGVPFqRWq67ORk/khBAicdSQE0KIxFFDTgghEtdg+sj35vFjg8ePFqfnKsBvRWe9PO3Fohy71zyirf6U+B9jgoS03LclV2b4WlyKU4o7dkvdkQOBNj+5aPfYZL/+XL7fBrEvVfGI/XX4k7Z42y0jdW9tbgqXT554SUivVPJzAAYeeElI9zn1AVeW3FP8vzz8Bn9NC7PUtJoOoydyQgiROGrICSFE4hpM14qtPXmOTW/PL75Zreu3tJqCDwCp7YK5vOqFN4W0/vN5XFniJrdq3ZM4x1Nve3L5nz4Vh6L1epdfRVP1/CwuL1OJOw0xi/2PzrNj+Gek9Z+J07w1t6u32iKpPdYxiQYfn+eixZUTNxxP5Moiz4mrMQZ1HceV5d4pcGYVK0RP5IQQInHUkBNCiMRRQ04IIRLXYPvIa9tu7y5cvt1PS+wee+dXfneYxbmX7BxJnCXTxC9Na9aIQ0lVE/lp1n35FU4rlDfsNSE99zrfn75+9UDxHq/wSy+cx0Ih3fXTc1xZQ909SKXgmxdvlbgz100J7WK/QysObV5/Yg9XpugdLaS/cOvGlY3D4VqtlzV6IieEEImjhpwQQiSOulYq4KZQcfkXA8UNfbsldq3w3FuTXhHSYWn5zq0YqVRK/u9c3nJkn5BWvBhqe7hdph83cPnH/xI3+9UW8fdYliQON2v6zWCurPhQlpC+qS9GY/Bpywgu//JnnYT0qnf47sX38g65pE7VYbDaSJsZDXaP82FmV1TnvuiJnBBCJM6hhnzFihWIiopCVFQUliwp/1IvNTUV0dHRGDx4MJKSkmq1koQQQuyrtCFPTU3F0aNHsWPHDuzcuRPnz5/H3r17MX/+fKxcuRL79u1DRkYGjhypn5uSEkJIQ1dpH7m/vz/mzp0LN7fyaeUdOnSAWq1GSEgIgoPLp6VHR0cjOTkZkZGRtVtbF2ji7iWkL3Tnp93771xs9zzDBr7soRRxd5LC0jtOqh2prhe+EL+n+D6SX9VSESz2mQ9/bDpXdu52JpevKJY/pYm/L8/YlCVcEVfDLDJcrrS+DcHCIn5F0pdl4pDdGafiuLKYBHGpjDd28ktYbM79DXXpMf8HhbS802NcmUwuPgtfUvHfqblSpU/kDz74IHr06AEAUKvV2L9/P2QyGfz9xbUjAgICoNVq7V2CEEJILXL4y87Lly9j8uTJeO+99xAcHAyZTFzjmzHG5QkhhLiOQ8MPT58+jRkzZmD+/PmIiorCiRMnoNOJK4TpdDoEBATUWiVrk0Ku4PJZL3YU0p7xy7kyi9XH6nPh8VzZm6yQy1N3Sv1ivRrmuGf4YWJX9eLMyks3+Y1GGGMO3+NdfZqQHvnrv7iyxM97Cumfpqq5st8LNQ7fQ0psVxYNnvytkM587RRX5jZb/P+05jl+o4+Hx4urVX52i1/VtKDkdo3racvLzYPL//Sq1SYx/iFcmaVYnEG8uPi/Tq+Loyp9Is/Ly8O0adOwdOlSREVFAQDCwsJw9epVZGZmwmw2Y+/evYiIiKjkSoQQQmpDpU/k69evh16vR2KiuA7v+PHjkZiYiOnTp0Ov1yMyMhJDh9rf9ooQQkjtqbQhj42NRWxs7H3Ldu/e7fQKEUIIqZpGOUXfXSkOb7rSsz1X5rngSzFjs3nqmcfFFez6XadNk6Uq+Vrt9GVa7wjT5O8ra+UeUma94mHTz/n/P+O3LBDS69YN4cpmnxA3PJ7550mu7P0X9wrpZTm/OFyXDs1acfkZHg8J6ckf8WWqETFC2nZXqLwxc8V0Ef89mSvRFH1CCJE4asgJIUTiGmXXijZGnMnn8b7NOjFW3SmbHvmIK1otq7uPToQ0ZP+be1xIJz9znit7scXPQjpxA9/tknh0vpCePzWXK7t+2VtItx3Db7itnPgGl5e34GdxWzMmfy2k53zwF1e2Xvun3fNciZ7ICSFE4qghJ4QQiaOGnBBCJK5R9pEruoursBn3rePKds3LFtKvXP8ZhBDXst2YeUXOr0J62xi+/7yfb0ernB9Xtna0SUi7zbK/cikAFL0uDjH87Ewbruz7YnEnKPWt+rk4ID2RE0KIxFFDTgghEtcou1aavLa5rqtACKkG29mT/yqyP8P6X9aTa1cOqMJdpLfxBz2RE0KIxFFDTgghEkcNOSGESBw15IQQInHUkBNCiMTVeUOuVMrQ4QEfBLfx4v41aVLzATWtAj3RxKf8OsFtvCCv4NXKZUDrVp72D7DD21uJNvc5z9NDgeC2XpWe3/FvTSCXV23j6gB/DzRrqqr0OB8fpfB+tmntBXc314Wb4lp7cfX2EuPaupUnlErXbXxOca29uHp5KhDcxgvt2nojKMADVdnPvl4MP2QM0OSUCHmFQoZ2bb2h15fAYLBUcKbjrK9/P3KFDB7uigqPkRKVSoaWfu7Q5JTAbGbw8lQgKMgTmVnFLqsDxdX5ZDIgMMADmuxiGE0MTZuq4N/SA3nXSl1WB4qr88nlMgQEeCAnpwRGE0MLPze09HOHrkDv0Pn1oiG3ZTYzGI0WqFRyuLvJ4dtEBZlcBouFITevFE2aqNDUVwXZ/x+rK9DDaLRAoZAh0N8DCqUMJhODQiH+Sev4tyb4S10Ei4WheTM3NPFRggEwGi3Izy9DgH/5X8DgNl7Q5JRApZLDv4U75AoZZABu3jbgzp3yKb9+zd3QxEcl1LMyKpUM/i08IJfLoFDIoDeYoc0vw90N2lv4ucHdXQEZgIIbepSUlO/ybu912uPuJkeAvwc0OSVgDMjXlcFsLr+JXm+BUuG6J7f7objWPK53lT8VMshlMrC7N6wjFNeax9XLSwG93gKjqfwmt24bEdzWW9oNuYe7HCqVHPoyMzw9FXBzU0CdVQTGAA8PBXx9lMjJLW+sPD0VaBXogazsEvi3dEeZ3ozCawaolDIEt/W+59peXgo08VEhO7cYFgvQ0s8dTZu6IV9XhnZtvYX/LEGBHsjPL4PeYIFcBrRt4wWDwQKlQg5vbyWysovBWPnHwcr4NnHD7SIjiorKf7GC23jBy0uJ4uLyvNHIoLteAjeVHG1aeyFTUww3N7nd12mP3mAR6m8yMZhMZqGsZQt34X51heJa87gyBuiul6FtGy/hj3RObsVPr7WN4lrzuCoVcphMYqNvMjEo5DLIZIAjf6dd3pCbzeWNy92nQ6WivLLtrPqnLBbgekEZICv/2Hb3rzcANPFRQqWSI7iNeLxCIYObmwxenkrcvFUCpVIGBqCszAyFQib0ISqVgI+3EiWlJsjlMsjl5X+5ufooZVApZXBTyREY4CHcQyaXwctTAZVKjtJSs1Cf4hIjfJu43dNPqfj/JwOlUoZbtw3w8FDAr7kbVCo5lEo5lEqxXsUlRiiVMlhY+RODt7cCHu4Ku69TLit/InOkb1QmA1q28IBSIYM2v1R8L/6//nfjUVMUV9fEVaWSw6+5O3LySmAyMTTxUaFVkCdyrbpWnBlbiqtr4qpQADLZvccolTKhIa8orjLm4s9lp06dwgsvvODKW5IKbNq0CY899liNr0NxrX+cEVuKa/1z37gyFystLWUnT55karWaaTQadvLkSda9e3em0Wju+2/t2rVswoQJQn7Hjh2sX79+7L///S/TaDRsxYoVbODAgSwrK4tNmTKFxcbGMo1Gw06dOsUeffRRtnbtWqbRaFinTp1YRkYG27p1K3v66afZxYsXmUajYXPnzmVz5sxhZ86cYaGhoSwrK4v99ddfrH///uzrr79mGo2GnT59mvXq1Yvt37+fbd26lQ0cOJD9/vvvLDvahaMAABoJSURBVDMzk7366qvs2Wefvafee/bsYYMHD2YajYY98sgj7PDhw0yj0bBff/2VPfLII2z16tVCvZYvX840Gg376aefWO/evdn58+crfJ1vvvkm+/TTT+2+ZxqNhv3xxx8sMjKSJSQk3LdcrVazkydPstLSUoqrhOK6e/du1r9/f3bu3Dmm0WjYpk2b2JNPPllrsaW4uiauaWlprHfv3uzYsWNMo9GwuLg4Nn36dIfj6vKuFQ8Pj3v+msjlcrRt2/a+x/v5+cHT01Mob9u2LYqLi/H+++9DJpPBx8cHX331FYKDg5GYmIh58+ZhypQpCAoKQteuXeHn5yec26pVKzz88MO4efMm3nnnHQBAx44dsWDBAnh6eiIsLAxTp07Fpk2bsGbNGiQkJGD79u0wmUyYNWsWhg4dCgAoKCjA9OnT4evriy5duqCsrOye+ufk5EClUqFt27Z4++23ER8fDy8vL/j4+KB37964c+eOcM6dO3fw5ptvQiaTYdmyZejatSu6du1q93V6e3ujadOm99wzPT0dsbGx2LVrF1avXg2tVovjx4/j+HFxP8QNGzagefPmAICQkJBqxfB+KK6uiWvbtm1x+/ZtzJkzByqVCk2bNsWaNWvuOcdZsaW4ui6uS5YswaJFi2A0GtGuXTt8/PHHaNasGXeOvbi6vGuFEEKIc9X5hCBCCCE1Qw05IYRIXL0cR16ZhQsX4uTJkwCAK1euoE2bNvDwKB969MMPPwhpZ8nOzkZ0dDTOnj1b4XEDBw7EsmXL0K1bN4evvXz5cty4cQMffPCB3WNSU1Px8ccfC/mysjKo1Wps27YNoaGhDt+rvmtscQWA/fv3Y9WqVWCMoXnz5oiPj0f79u0dvo8UNMa47tu3D19++SUUCgWCgoIQFxeHNm3aVHhOTUiyIY+NjRXSAwcOxNKlS6sUDKnp27cvdu3aJeRnzJiBwYMHN6hGHGh8cdXpdIiLi8Pu3bsRFBSE7777DgsWLMD69evrumpO1djiqlarERcXh++++w6dO3fGyZMnMWPGDGzbtq3W7tkgu1ZCQ0Mxc+ZMDBkyBOnp6ejcuTMKC8Utoqzzhw4dwrPPPovRo0dj/Pjxlf4Vv379Ot544w0899xzGDhwICZMmICCggKhfPPmzRgzZgyioqKwdetW4edVvY9Wq8WoUaOg1fK7du/atQvZ2dmYOXOmw+9HQ9HQ4urv74+UlBQEBQXBZDIhJyfnnlEKjUFDi+vFixfRpUsXdO7cGQDQq1cv5OTkIDs7u8rvjcNqPNC0jg0YMIClpaVxP+vUqRPbsWMHly8oKLgnf/XqVTZixAhWWFjIGGPs0qVL7IknnmDFxcXc9TQaDevRowdjjLENGzaw1atXM8YYs1gs7NVXX2Xr168X6hIXF8cYY+zatWssPDycXbp0qcL7fPHFF+yjjz5y6LXq9XoWGRnJTp486ejbI1mNKa5paWmsb9++7NFHH2Vnzpxx9C2SpMYQ18zMTNa7d2924cIFxhhjP/30E+vcuXOtxlaSXSuOcGRGW0pKCvLz8/HSSy8JP5PJZMjKykKXLl3ue86kSZNw6tQp/POf/4Rarcbly5cRFhYmlI8fPx4AEBgYiCeeeALHjh2DQqGwe5+q+M9//oPg4GCnzMSUqoYY127duiElJQW//PILpk6dioMHD8LX17dK15C6hhTXdu3aYdGiRYiLi4PBYMCgQYPQpUsXqFSVL2VbXQ22Iffyuv/awgaDQUhbLBaEh4fj888/F36Wl5eHgIAAu9f95JNPkJaWhrFjx6JPnz4wmUzc6nNyq0WULRYLlEolzGaz3fscOHDA4de0b98+PPPMMw4f3xA1pLhqtVpcunQJ/fv3BwBERETAx8cHWVlZDe77j8o0pLgaDAaEhIRgy5YtQv6bb76xO4nKGRpkH7ktPz8/pKenAwD27t0r/Dw8PBwpKSm4cuUKAODIkSMYOXIkysrK7F7r6NGjmDRpEkaPHo0WLVogNTWVW8Rmx44dAIDc3FwcO3YM4eHh1bqPLcYYTp06hfDwcMdfeAMn9bgaDAbMnj0bmZmZAIDffvsNJpMJHTp0qMK70PA0hLg+//zzyMvLA1A+m7pnz561+v1Hg30itxYbG4v4+Hj4+vqib9++8Pf3B1A+3Tc+Ph6zZ88GYwxKpRKrVq2Ct/e9y2neNW3aNCxZsgTLli2DSqXCo48+yn3k0uv1GDNmDIxGI2JjY/HAAw8AQJXvo9VqERMTgzVr1iAwMBA3btxASUkJgoKCnPSuSJ/U4xocHIyFCxdi+vTpkMlk8PX1xVdffQVPz6rvfNOQSD2ugYGBWLBgAaZMmQKz2YwOHTpg8eLFTnp37o+m6BNCiMQ1iq4VQghpyKghJ4QQiaOGnBBCJK5GDfmePXswfPhwDB48GJs2bXJWnQghhFRBtUetaLVaJCUlYfv27XBzc8P48ePRp08fdOzYscLzysrKkJGRAX9/fygUiurentSQ2WyGTqdDaGioUxYtorjWH86MLcW1/qgortVuyFNTU/H4448LYyOHDBmC5ORkvPnmmxWel5GRQXsA1iPO2rOT4lr/OCO2FNf6535xrXZDnp+fL4zvBICAgACkpaVVet7dc7JzimEy08jHuqJUyNC2jTcXw5qguNYfzowtxbX+qCiu1W7ILRYLZDKZkGeMcXl77n48M5kZTCb6xahrzvq4THGtf5wRW4pr/XO/uFb7y86goCDodDohr9PpKlzzgBBCSO2odkPet29fHDt2DIWFhSgtLcWPP/6IiIgIZ9aNEEKIA6rdtRIYGIhZs2Zh4sSJMBqNGDduHLp37+7MuhFCCHFAjRbNio6ORnR0tLPqQgghpBpoZichhEgcNeSEECJx1JATQojEUUNOCCESRw05IYRIHDXkhBAicY1iz86q2OsnTmp6KiPB4fNMJ3Zz+a9eOy2k39Me5soszFK9yhFCyH3QEzkhhEgcNeSEECJx1JATQojENfo+coWcXxKy399LhDQzGbgydivf/nUeHcrl3zw9Qkh37mbiyp4rOiGkSwxljleWOMw2rgMCHhbSw9CSK/uf4Bwh7ffFDP467UL5C8vEZ58bf3+ZK8pICxTSMeYrXNnVW9ccqDWpLdcG8juXNdu4vlrXMWz6hMtP/7xASH+Te6xa13QGeiInhBCJo4acEEIkrtF3rfTzf4jLe8z7VEgbf0jiyprM3mn3OgmtBnD5GUs7C+mnzsRyZdrfxKGKHV/dxJcV3aikxuSup4PCuPwItBDSzW22JXv25AfVugez2A4VFfPNNq7lSvpZpdNP7OHK3p4hdqetzjlarbqQisW2fpLLvzPqlpD2iP2cK7ONqyn5azFz547de7i98C6X/7LXcSF9aNSfXJnmtg6uQk/khBAicdSQE0KIxFFDTgghEtfo+8jnGZtzeVZWLKRTF153+Drv5/3M518Q83eW8X1lqnHiELfO3vx51EdesW/9xe8ixv7nVa5M7tfa7nmsqFBIm9OPcGVFK/cJ6VNp9q9hq6WCHzoa9vPbQlrRm98565OZl4T0hvluXJneZpgrsc/LzYPL/9xM3F6y20/TuTKZj5+QNu5ZzZUN++C/XP5EwWUhbTAb7d5f9/N5Lu+zco2QDnBrypVpQH3khBBCHEQNOSGESFyj7Fpp31Scgdfvm35cmemgOBxwSKFzhomtSCzk8rPGiekfOvMfq1tpnXLLBqOlF/9xdcxycYambVeK+S9xxclTz2znypaoSoX0vmtnK7jjnxWUVexIpDjErdfRd7gy1fNit8vaT4q5sok6vnuN8NyVYldUin9XrqzLbx+LGZsukdL33hDS3ffy3RzZdxzvNrV2+y8Vl/exSnvK+S4zV6InckIIkThqyAkhROKoISeEEIlrlH3kXyk6CWmlzaqF+tRY28NJHbpecovLW9LThbQ5sB1X1j5K3NHJ9jxXiCwQV7+79S0/JdztjXghHdEmjz/RdaPUJOlMsLiMxt9++dTucaa967i83/e/O70uhQVeXN76W5pt3fnvuwIPOv32dtETOSGESJxDDXlRURFGjBiB7OxsAEBqaiqio6MxePBgJCUlVXI2IYSQ2lRp18q5c+cQGxsLtVoNACgrK8P8+fOxceNGtGrVClOnTsWRI0cQGRlZ23WttmBffy7f7ztxdqClmP8IPusbs9Pv38HA7JZdu9LE6fdrNMz8hh110Z1ij+LpKLtlpXfqbpiaFDzXug+X75CyVMzYbFxu+vOkkH5o3qFarRcAtO/F/47J5OKzsPfrI/mDD35W6/W5q9In8i1btiAuLg4BAQEAgLS0NISEhCA4OBhKpRLR0dFITk6u9YoSQgi5v0qfyBMSErh8fn4+/P3FJ9yAgABotTSLhRBC6kqVv+y0WCyQyWRCnjHG5QkhhLhWlYcfBgUFQacTx0vpdDqh26W+6uXdnssrwwYJaf0Kfrjh17mpTr//sC+62C17obR6U4Ubq8lfi5tjT1u11+5xtg8X/jZT/a0NaCru5rSvIJ0ru6MvsT3crtmtxe+JFA/0sHvcjGJ68LGlUohN0doYH77Qql/cdIbvxu05caOQru60+6rwmDCMy1vvNFT2rf3fx9pW5SfysLAwXL16FZmZmTCbzdi7dy8iIiJqo26EEEIcUOUncnd3dyQmJmL69OnQ6/WIjIzE0KFDKz+REEJIrXC4IT90SBzaEx4ejt27d1dwdN2zXjHt6zH2hxR+saF2Jrd+1Foc4qjozg/NNGdlCOlSM20qUBXb8sThZtsqOE7T50Eu3+Jf4sYC5svHuTLFg+JwN/PpfVxZ02f4IWRmi/i7ZDusNX6T1ZBDpTtXZrl2RUhn6qk7zdayluKnetUrfHen5cY1IT3vtRSu7I8b2bVbMRuqJ8ZyeeuulcupzVxaF2s0s5MQQiSOGnJCCJE4asgJIUTiGuzqh1H+YULa4/2lXJlh0ydCen3JhVq5/+yP2gpp251sbs/5QEhfvXUNxDm6tWgvpJt/we/Qc3PCFCHd92wRV7bZTVxFsceZj7kyTd+dXH7wBfE7jRWyNlyZomMvIW292TMAvDFqvZC+dCPnvvVvTBRyBZd/YWKZnSMBfZI4KXF5Tlqt1ckeT5V75QfVMXoiJ4QQiaOGnBBCJK7BdK1YDzcE+CGHxu/5xei7J4hD2NS3nLNOzNKggVxeET5CSJv+OsOVvX627oYpNWRPeohdJPJW/PDDX9PFrq6rtw5zZYtaiV0kW2yu2fz7NVz+JByT//z7XP6b3IsOntk4rGvBTyJ0m/GBnSOB9f9uWdvVqdCPvo84dNwJhU/lB9USeiInhBCJo4acEEIkjhpyQgiRuAbTR2493BAA3N76h5CePWAJV+aMfvFRrXpy+Td+nMLl5c2DhHTpR3z/37a8P2p8f3Kv72+cE9KL/+R7s4et6i6kV07nn19W6sXp82ZNBlemCA6tVl3i8/2qdV5jMfhBfmq99U47hu/4/6/vXKv9nX8q0n2SVUbG/+6YT4pLlXx0y9FvUJyPnsgJIUTiqCEnhBCJo4acEEIkrsH0kZ8vy+Pyw/rPF9JHtBm2h9fYC4YmXN52Gr61OSm242Cpj7w2XC8Rdzif+sJ2rmzd8cVC+uVTY7iyl016MWOz/CwrvcPlzSfFZW7lf+vGlcnbdhXSX/zwLFf2dX/n7zwlZU3fjuby1svBHvy4yPZwl2rq4c3lFYOeEjNWuxUBwOFXxX7xQpvfFVeiJ3JCCJE4asgJIUTiGkzXiu1OIX+gejuHtPIRh43NaMIPaXxUbxTS/X+dVuF1jCni/jXdzfzH9cWtBtgeLviV3RTS+66drbiyjVBXP3Ea/kT3jlxZT73YRdK1fb7da5iO7eDyMjcxPuajR7myiZv4HZx25Z0W0mnt+KnbD6ZY7SakaDD/tZzmqSBxCKg8jF/SglltnJwgr9vVIaf4PcblFd0H2TkS+MHDaLfMleiJnBBCJI4ackIIkThqyAkhROIabEfe6FZiP9cA5suVtTOIS9w+vY7vD4N3UyGpDI2s9v2td9uOOTO2giN5M7RXhTRTp3NlrZ5fJaTv6EuqXTcpCWvxAJc/PFoc9unx4T+4MutdeX4fuJgrm9JrtpD+pdD+krKlJr5P3Gwx2zkSaD+ZpuFXhY9MXGpa5s4P8bMUaIT06et/uqxOd/ULEIeOfrQ/xu5xhrXxXP5/tXU3Ld8aPZETQojEUUNOCCESJ+muFTeFSkhfG8/vCOMx920hXdGsS2cpi5/J5Ut/L3b6PUwVfMxvqH7w43dd8fhwmZA27VnNlT2/4JKQ3pvH78rkLCsCxWFzqglv2z2u+INldsvIvWQtgoX0h6354bkf5v7s9Pt915K/xzOHXhPr0jTA7nlrvmJc3mg2Obdi1URP5IQQInEONeQrVqxAVFQUoqKisGRJ+VrBqampiI6OxuDBg5GUlFSrlSSEEGJfpQ15amoqjh49ih07dmDnzp04f/489u7di/nz52PlypXYt28fMjIycOTIEVfUlxBCiI1K+8j9/f0xd+5cuLmVDx3q0KED1Go1QkJCEBxc3q8VHR2N5ORkREZWf7hedSgVCiFddJHvPy566UOHrvFhXgsur7LaAWTZiQS75xm+5ssC11/g8nqbYWykeoK/fc1u2VNx/BIGx3XOX1Xy/dZPcvmX1vYVM26eXJn11P8+J+tuJbz6amfeKSFt3MAPD3Wb/L6Qfmf141zZgy+L6Zdv8EsotPcV+7Mv3eCn9ndq3kZI//IIH6tmGz90rNIATH8cE+tWx7sV2VPpE/mDDz6IHj16AADUajX2798PmUwGf39/4ZiAgABotTXfPo0QQkjVOfxl5+XLlzF58mS89957CA4OhkwmE8oYY1yeEEKI6zg0/PD06dOYMWMG5s+fj6ioKJw4cQI6nU4o1+l0CAiwP2SntpQYyoR0u9PO+VhtOyzJWuYAccXDcE0mV0ZdKbXDsI4fYmg9/HDb3/ihYG11qDHrGcEAEJsylz/AauMJ84k9XNEXM8WZuM7Y4Lshm7iykMtvbP6lkFaOep0rG3tuqJAenXuZK5NZbQLBSm5yZey62NVi3LqLK0vpNp/LP35oqpCWWw2FBIAfx+2/9wXUM5U+kefl5WHatGlYunQpoqKiAABhYWG4evUqMjMzYTabsXfvXkRERNR6ZQkhhNyr0ify9evXQ6/XIzExUfjZ+PHjkZiYiOnTp0Ov1yMyMhJDhw6t4CqEEEJqS6UNeWxsLGJjY+9btnv3bqdXiBBCSNVIeoq+MwT7+nP5Ucs6C2nrYUcA8PdCsU++LjdabUw2/JvfuPq1f4g7srT43y+4sqKzB4R0+tRfubICs4fde/QbL2726zbT5qHFdjNmqw2e587gV75bkcffk9hnPRQRAB6enyWkf/hIzZV1+6qfmLHwmx9bf0tiOsTPZXluh1j6n2v8SqJTWj/B5fta9Yubz/PXebmEr2t9RFP0CSFE4qghJ4QQiWv0XSsKGf+3jF25IqQHzzrMldXFgveN3SybmXTNe4kfl8fG8d1i1sPWepwZ4ZT76z95j8v3+EbsAqAhhs6TdVvcLDscNhtnjz3h9PvN8rDfNTrvFf537laZ81cydTZ6IieEEImjhpwQQiSOGnJCCJG4Rt9HbtvP6fvBATtHkvrgJZ24W0zMTBVX1m/B70J6uRf/jBI8OdDuNW/uFJdbiM/jhzuutxkmxxi/LACRphdu8/3eR602f86y1P8+cVv0RE4IIRJHDTkhhEhco+9aIdJlMBu5/CGtOHvvYduD/+HoVZ2/OQWpf85ev8LlvXtMrKOaOAc9kRNCiMRRQ04IIRJHDTkhhEgcNeSEECJx1JATQojEUUNOCCESRw05IYRIHDXkhBAicdSQE0KIxLl8ZqfZbC6/sULm6lsTK3ff/7vxqCmKa/3hzNhSXOuPiuLq8oZcp9MBANq28Xb1rcl96HQ6hISEOOU6AMW1PnFGbCmu9c/94ipjLl6Xs6ysDBkZGfD394dCoXDlrYkVs9kMnU6H0NBQeHjY32HeURTX+sOZsaW41h8VxdXlDTkhhBDnoi87CSFE4qghJ4QQiaOGnBBCJI4ackIIkThqyAkhROKoISeEEImjhpwQQiSuThryPXv2YPjw4Rg8eDA2bdpUF1WoN1asWIGoqChERUVhyZIlAIDU1FRER0dj8ODBSEpKquMaOo7iKqK4Nlz1MrbMxa5du8YGDBjAbty4wYqLi1l0dDS7fPmyq6tRL6SkpLDnnnuO6fV6ZjAY2MSJE9mePXtYZGQky8rKYkajkU2ePJkdPny4rqtaKYqriOLacNXX2Co+/PDDD135h+PAgQOQy+WIioqCSqVCYWEh/vzzT/Tu3duV1agXysrK0KdPH7Rp0wYKhQLp6enQ6/UoLi7GpEmTIJfLwRjDkSNH8NRTT9V1dStEcRVRXBuu+hpbl3et5Ofnw9/fX8gHBARAq9W6uhr1woMPPogePXoAANRqNfbv3w+ZTCbJ94fiKqK4Nlz1NbYub8gtFgtkMnFJTMYYl2+MLl++jMmTJ+O9995DcHCwJN8fiuu9KK4NV32Lrcsb8qCgIGFpTKB8ScaAgABXV6PeOH36NF566SW8/fbbGDNmjGTfH6nWu7ZQXBuuehlbl/bIM/HLk4KCAlZSUsJGjhzJzp075+pq1Au5ubmsT58+LDU1VfhZWVkZi4iIYGq1mplMJvbKK6+wffv21WEtHUNxFVFcG676Gts6WcZ2z549WL16NYxGI8aNG4cpU6a4ugr1wsKFC7Ft2za0a9dO+Nn48ePRvn17LF68GHq9HpGRkZg3b54kPs5SXMtRXBuu+hpbWo+cEEIkjmZ2EkKIxFFDTgghEkcNOSGESBw15IQQInHUkBNCiMRRQ04IIRJHDTkhhEgcNeSEECJx/wceGZzogvQ+cAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 6 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Display some error results \n",
    "\n",
    "# Errors are difference between predicted labels and true labels\n",
    "errors = (Y_pred_classes - Y_true != 0)\n",
    "\n",
    "Y_pred_classes_errors = Y_pred_classes[errors]\n",
    "Y_pred_errors = Y_pred[errors]\n",
    "Y_true_errors = Y_true[errors]\n",
    "X_val_errors = X_val[errors]\n",
    "\n",
    "def display_errors(errors_index,img_errors,pred_errors, obs_errors):\n",
    "    \"\"\" This function shows 6 images with their predicted and real labels\"\"\"\n",
    "    n = 0\n",
    "    nrows = 2\n",
    "    ncols = 3\n",
    "    fig, ax = plt.subplots(nrows,ncols,sharex=True,sharey=True)\n",
    "    for row in range(nrows):\n",
    "        for col in range(ncols):\n",
    "            error = errors_index[n]\n",
    "            ax[row,col].imshow((img_errors[error]).reshape((28,28)))\n",
    "            ax[row,col].set_title(\"Predicted label :{}\\nTrue label :{}\".format(pred_errors[error],obs_errors[error]))\n",
    "            n += 1\n",
    "\n",
    "# Probabilities of the wrong predicted numbers\n",
    "Y_pred_errors_prob = np.max(Y_pred_errors,axis = 1)\n",
    "\n",
    "# Predicted probabilities of the true values in the error set\n",
    "true_prob_errors = np.diagonal(np.take(Y_pred_errors, Y_true_errors, axis=1))\n",
    "\n",
    "# Difference between the probability of the predicted label and the true label\n",
    "delta_pred_true_errors = Y_pred_errors_prob - true_prob_errors\n",
    "\n",
    "# Sorted list of the delta prob errors\n",
    "sorted_dela_errors = np.argsort(delta_pred_true_errors)\n",
    "\n",
    "# Top 6 errors \n",
    "most_important_errors = sorted_dela_errors[-6:]\n",
    "\n",
    "# Show the top 6 errors\n",
    "display_errors(most_important_errors, X_val_errors, Y_pred_classes_errors, Y_true_errors)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "# predict results\n",
    "results = model.predict(test)\n",
    "\n",
    "# select the index with the maximum probability\n",
    "results = np.argmax(results,axis = 1)\n",
    "\n",
    "results = pd.Series(results,name=\"Label\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "submission = pd.concat([pd.Series(range(1,28001),name = \"ImageId\"),results],axis = 1)\n",
    "\n",
    "submission.to_csv(\"cnn_mnist_datagen.csv\",index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}