CNN_on_MNIST_dataset.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "VAi3bJMeqzRU",
    "outputId": "7f191a8c-d1cd-4da1-dd70-d5a106712602"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from keras.datasets import mnist\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Flatten\n",
    "from keras.layers import Conv2D, MaxPooling2D, BatchNormalization\n",
    "from keras import backend as K\n",
    "from keras.utils import np_utils\n",
    "\n",
    "from matplotlib import pyplot as plt\n",
    "from prettytable import PrettyTable"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "-0AoZLnkvVhT"
   },
   "outputs": [],
   "source": [
    "def plot_loss(x, vy, ty, xlabel='Epoch', ylabel='Categorical Crossentropy Loss'):\n",
    "    _, = plt.plot(x, vy, 'b', label=\"Validation Loss\")\n",
    "    _, = plt.plot(x, ty, 'r', label=\"Train Loss\")\n",
    "    plt.xlabel(xlabel)\n",
    "    plt.ylabel(ylabel)\n",
    "\n",
    "    plt.grid()\n",
    "    plt.legend()\n",
    "    plt.grid()\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "owayIH7UOMFX"
   },
   "outputs": [],
   "source": [
    "pt = PrettyTable()\n",
    "pt.field_names = ['CNN Architecture', 'Mean Train Accuracy', 'Mean Validation Accuracy', 'Test Accuracy']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "0j8WywH2kvKk"
   },
   "outputs": [],
   "source": [
    "# Credits: https://github.com/keras-team/keras/blob/master/examples/mnist_cnn.py\n",
    "batch_size = 128\n",
    "num_classes = 10\n",
    "epochs = 12\n",
    "\n",
    "# input image dimensions\n",
    "img_rows, img_cols = 28, 28\n",
    "\n",
    "# the data, split between train and test sets\n",
    "(x_train, y_train), (x_test, y_test) = mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 364
    },
    "colab_type": "code",
    "id": "btj0i1AGkvQQ",
    "outputId": "d81fb0e4-1e87-44f1-817d-40c0b9a89683"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7fcd68a99be0>"
      ]
     },
     "execution_count": 5,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUsAAAFKCAYAAACU6307AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fcda6bde828>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(x_train[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 67
    },
    "colab_type": "code",
    "id": "oO_ZJvejkvS0",
    "outputId": "3cedd7e8-acb0-47e0-fbc0-1140e413b21c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "x_train shape: (60000, 28, 28, 1)\n",
      "60000 train samples\n",
      "10000 test samples\n"
     ]
    }
   ],
   "source": [
    "# Check for Theano vs Tensorflow Backend\n",
    "if K.image_data_format() == 'channels_first':\n",
    "    x_train = x_train.reshape(x_train.shape[0], 1, img_rows, img_cols)\n",
    "    x_test = x_test.reshape(x_test.shape[0], 1, img_rows, img_cols)\n",
    "    input_shape = (1, img_rows, img_cols)\n",
    "else:\n",
    "    x_train = x_train.reshape(x_train.shape[0], img_rows, img_cols, 1)\n",
    "    x_test = x_test.reshape(x_test.shape[0], img_rows, img_cols, 1)\n",
    "    input_shape = (img_rows, img_cols, 1)\n",
    "\n",
    "x_train = x_train.astype('float32')\n",
    "x_test = x_test.astype('float32')\n",
    "x_train /= 255\n",
    "x_test /= 255\n",
    "print('x_train shape:', x_train.shape)\n",
    "print(x_train.shape[0], 'train samples')\n",
    "print(x_test.shape[0], 'test samples')\n",
    "\n",
    "# convert class vectors to binary class matrices\n",
    "y_train = np_utils.to_categorical(y_train, num_classes)\n",
    "y_test = np_utils.to_categorical(y_test, num_classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 848
    },
    "colab_type": "code",
    "id": "xON5C4JlkvU2",
    "outputId": "239628f4-bdbb-4c70-dd88-b81fd2ab2163"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 18000 samples, validate on 42000 samples\n",
      "Epoch 1/12\n",
      "18000/18000 [==============================] - 6s 337us/step - loss: 0.5104 - acc: 0.8462 - val_loss: 0.1122 - val_acc: 0.9662\n",
      "Epoch 2/12\n",
      "18000/18000 [==============================] - 4s 228us/step - loss: 0.1669 - acc: 0.9502 - val_loss: 0.0766 - val_acc: 0.9772\n",
      "Epoch 3/12\n",
      "18000/18000 [==============================] - 4s 229us/step - loss: 0.1102 - acc: 0.9663 - val_loss: 0.0682 - val_acc: 0.9787\n",
      "Epoch 4/12\n",
      "18000/18000 [==============================] - 4s 230us/step - loss: 0.0854 - acc: 0.9728 - val_loss: 0.0637 - val_acc: 0.9814\n",
      "Epoch 5/12\n",
      "18000/18000 [==============================] - 4s 230us/step - loss: 0.0690 - acc: 0.9793 - val_loss: 0.0580 - val_acc: 0.9826\n",
      "Epoch 6/12\n",
      "18000/18000 [==============================] - 4s 229us/step - loss: 0.0563 - acc: 0.9818 - val_loss: 0.0526 - val_acc: 0.9845\n",
      "Epoch 7/12\n",
      "18000/18000 [==============================] - 4s 230us/step - loss: 0.0529 - acc: 0.9822 - val_loss: 0.0536 - val_acc: 0.9843\n",
      "Epoch 8/12\n",
      "18000/18000 [==============================] - 4s 230us/step - loss: 0.0453 - acc: 0.9856 - val_loss: 0.0592 - val_acc: 0.9826\n",
      "Epoch 9/12\n",
      "18000/18000 [==============================] - 4s 231us/step - loss: 0.0396 - acc: 0.9872 - val_loss: 0.0568 - val_acc: 0.9845\n",
      "Epoch 10/12\n",
      "18000/18000 [==============================] - 4s 230us/step - loss: 0.0352 - acc: 0.9891 - val_loss: 0.0567 - val_acc: 0.9845\n",
      "Epoch 11/12\n",
      "18000/18000 [==============================] - 4s 230us/step - loss: 0.0339 - acc: 0.9887 - val_loss: 0.0556 - val_acc: 0.9852\n",
      "Epoch 12/12\n",
      "18000/18000 [==============================] - 4s 231us/step - loss: 0.0360 - acc: 0.9880 - val_loss: 0.0538 - val_acc: 0.9857\n",
      "Test loss: 0.04481222015287385\n",
      "Test accuracy: 0.9871\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAe8AAAFYCAYAAAB6RnQAAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fcd6031da20>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 41.5 s, sys: 9.24 s, total: 50.7 s\n",
      "Wall time: 53.4 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "def get_cnn():\n",
    "    model = Sequential()\n",
    "    \n",
    "    model.add(Conv2D(filters=32, kernel_size=(3, 3), strides=(1, 1), padding='valid', \\\n",
    "                     activation='relu', kernel_initializer='he_normal', input_shape=input_shape))\n",
    "    \n",
    "    model.add(Conv2D(64, (3, 3), activation='relu', kernel_initializer='he_normal'))\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "    model.add(Dropout(0.25))\n",
    "    \n",
    "    model.add(Flatten())\n",
    "    model.add(Dense(128, activation='relu', kernel_initializer='he_normal'))\n",
    "    model.add(Dropout(0.5))\n",
    "    model.add(Dense(num_classes, activation='softmax'))\n",
    "    \n",
    "    model.compile(loss='categorical_crossentropy', optimizer='Adam', metrics=['accuracy'])\n",
    "    \n",
    "    return model\n",
    "\n",
    "model = get_cnn()\n",
    "history = model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs, verbose=1, validation_split=0.7)\n",
    "\n",
    "score = model.evaluate(x_test, y_test, verbose=0)\n",
    "print('Test loss:', score[0])\n",
    "print('Test accuracy:', score[1])\n",
    "\n",
    "x = list(range(1, epochs+1))\n",
    "vy = history.history['val_loss']\n",
    "ty = history.history['loss']\n",
    "plot_loss(x, vy, ty)\n",
    "\n",
    "pt.add_row(['2-Conv+Dropout+Kernel(3x3)', sum(history.history['acc'])/len(history.history['acc']), \\\n",
    "            sum(history.history['val_acc'])/len(history.history['val_acc']), score[1]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 848
    },
    "colab_type": "code",
    "id": "Pqkh5OmOkvdN",
    "outputId": "80577095-7531-47ff-9139-57162a2b04c6"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 18000 samples, validate on 42000 samples\n",
      "Epoch 1/12\n",
      "18000/18000 [==============================] - 4s 238us/step - loss: 0.3027 - acc: 0.9089 - val_loss: 0.1015 - val_acc: 0.9695\n",
      "Epoch 2/12\n",
      "18000/18000 [==============================] - 4s 219us/step - loss: 0.0659 - acc: 0.9808 - val_loss: 0.0771 - val_acc: 0.9761\n",
      "Epoch 3/12\n",
      "18000/18000 [==============================] - 4s 218us/step - loss: 0.0323 - acc: 0.9896 - val_loss: 0.0759 - val_acc: 0.9773\n",
      "Epoch 4/12\n",
      "18000/18000 [==============================] - 4s 218us/step - loss: 0.0196 - acc: 0.9938 - val_loss: 0.0804 - val_acc: 0.9777\n",
      "Epoch 5/12\n",
      "18000/18000 [==============================] - 4s 222us/step - loss: 0.0100 - acc: 0.9972 - val_loss: 0.0751 - val_acc: 0.9792\n",
      "Epoch 6/12\n",
      "18000/18000 [==============================] - 4s 230us/step - loss: 0.0071 - acc: 0.9984 - val_loss: 0.0784 - val_acc: 0.9809\n",
      "Epoch 7/12\n",
      "18000/18000 [==============================] - 4s 218us/step - loss: 0.0031 - acc: 0.9989 - val_loss: 0.0771 - val_acc: 0.9823\n",
      "Epoch 8/12\n",
      "18000/18000 [==============================] - 4s 219us/step - loss: 0.0036 - acc: 0.9990 - val_loss: 0.0842 - val_acc: 0.9808\n",
      "Epoch 9/12\n",
      "18000/18000 [==============================] - 4s 218us/step - loss: 0.0026 - acc: 0.9993 - val_loss: 0.0911 - val_acc: 0.9787\n",
      "Epoch 10/12\n",
      "18000/18000 [==============================] - 4s 219us/step - loss: 0.0033 - acc: 0.9992 - val_loss: 0.0840 - val_acc: 0.9815\n",
      "Epoch 11/12\n",
      "18000/18000 [==============================] - 4s 218us/step - loss: 0.0060 - acc: 0.9983 - val_loss: 0.0813 - val_acc: 0.9819\n",
      "Epoch 12/12\n",
      "18000/18000 [==============================] - 4s 219us/step - loss: 0.0055 - acc: 0.9981 - val_loss: 0.0959 - val_acc: 0.9799\n",
      "Test loss: 0.08748575288906003\n",
      "Test accuracy: 0.9799\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfUAAAFYCAYAAABKymUhAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ne+nndyzkAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fcd4ecd3ef0>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 38.6 s, sys: 8.37 s, total: 46.9 s\n",
      "Wall time: 49.4 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "def get_cnn():\n",
    "    model = Sequential()\n",
    "    \n",
    "    model.add(Conv2D(filters=32, kernel_size=(3, 3), strides=(1, 1), padding='valid', \\\n",
    "                     activation='relu', kernel_initializer='he_normal', input_shape=input_shape))\n",
    "    \n",
    "    model.add(Conv2D(64, (3, 3), activation='relu', kernel_initializer='he_normal'))\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "    \n",
    "    model.add(Flatten())\n",
    "    model.add(Dense(128, activation='relu', kernel_initializer='he_normal'))\n",
    "    model.add(Dense(num_classes, activation='softmax'))\n",
    "    \n",
    "    model.compile(loss='categorical_crossentropy', optimizer='Adam', metrics=['accuracy'])\n",
    "    \n",
    "    return model\n",
    "\n",
    "model = get_cnn()\n",
    "history = model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs, verbose=1, validation_split=0.7)\n",
    "\n",
    "score = model.evaluate(x_test, y_test, verbose=0)\n",
    "print('Test loss:', score[0])\n",
    "print('Test accuracy:', score[1])\n",
    "\n",
    "x = list(range(1, epochs+1))\n",
    "vy = history.history['val_loss']\n",
    "ty = history.history['loss']\n",
    "plot_loss(x, vy, ty)\n",
    "\n",
    "pt.add_row(['2-Conv+Kernel(3x3)', sum(history.history['acc'])/len(history.history['acc']), \\\n",
    "            sum(history.history['val_acc'])/len(history.history['val_acc']), score[1]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 848
    },
    "colab_type": "code",
    "id": "HiDQh863xCHm",
    "outputId": "009dbd2f-ef43-41af-954e-60d7c95d2b92"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 18000 samples, validate on 42000 samples\n",
      "Epoch 1/12\n",
      "18000/18000 [==============================] - 5s 259us/step - loss: 0.4113 - acc: 0.8718 - val_loss: 0.1139 - val_acc: 0.9656\n",
      "Epoch 2/12\n",
      "18000/18000 [==============================] - 4s 230us/step - loss: 0.1434 - acc: 0.9577 - val_loss: 0.0803 - val_acc: 0.9760\n",
      "Epoch 3/12\n",
      "18000/18000 [==============================] - 4s 229us/step - loss: 0.0992 - acc: 0.9686 - val_loss: 0.0695 - val_acc: 0.9794\n",
      "Epoch 4/12\n",
      "18000/18000 [==============================] - 4s 229us/step - loss: 0.0733 - acc: 0.9761 - val_loss: 0.0670 - val_acc: 0.9802\n",
      "Epoch 5/12\n",
      "18000/18000 [==============================] - 4s 229us/step - loss: 0.0580 - acc: 0.9821 - val_loss: 0.0667 - val_acc: 0.9808\n",
      "Epoch 6/12\n",
      "18000/18000 [==============================] - 4s 229us/step - loss: 0.0510 - acc: 0.9826 - val_loss: 0.0666 - val_acc: 0.9806\n",
      "Epoch 7/12\n",
      "18000/18000 [==============================] - 4s 228us/step - loss: 0.0445 - acc: 0.9838 - val_loss: 0.0692 - val_acc: 0.9816\n",
      "Epoch 8/12\n",
      "18000/18000 [==============================] - 4s 228us/step - loss: 0.0379 - acc: 0.9877 - val_loss: 0.0626 - val_acc: 0.9832\n",
      "Epoch 9/12\n",
      "18000/18000 [==============================] - 4s 228us/step - loss: 0.0368 - acc: 0.9881 - val_loss: 0.0614 - val_acc: 0.9831\n",
      "Epoch 10/12\n",
      "18000/18000 [==============================] - 4s 228us/step - loss: 0.0309 - acc: 0.9897 - val_loss: 0.0650 - val_acc: 0.9833\n",
      "Epoch 11/12\n",
      "18000/18000 [==============================] - 4s 229us/step - loss: 0.0251 - acc: 0.9918 - val_loss: 0.0637 - val_acc: 0.9840\n",
      "Epoch 12/12\n",
      "18000/18000 [==============================] - 4s 229us/step - loss: 0.0228 - acc: 0.9927 - val_loss: 0.0625 - val_acc: 0.9843\n",
      "Test loss: 0.05603565712792624\n",
      "Test accuracy: 0.9852\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfUAAAFYCAYAAABKymUhAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fcd4eb47a58>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 38.9 s, sys: 9.18 s, total: 48.1 s\n",
      "Wall time: 51.8 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "def get_cnn():\n",
    "    model = Sequential()\n",
    "    \n",
    "    model.add(Conv2D(filters=32, kernel_size=(2, 2), strides=(1, 1), padding='valid', \\\n",
    "                     activation='relu', kernel_initializer='he_normal', input_shape=input_shape))\n",
    "    \n",
    "    model.add(Conv2D(64, (2, 2), activation='relu', kernel_initializer='he_normal'))\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "    model.add(Dropout(0.25))\n",
    "    \n",
    "    model.add(Flatten())\n",
    "    model.add(Dense(128, activation='relu', kernel_initializer='he_normal'))\n",
    "    model.add(Dropout(0.5))\n",
    "    model.add(Dense(num_classes, activation='softmax'))\n",
    "    \n",
    "    model.compile(loss='categorical_crossentropy', optimizer='Adam', metrics=['accuracy'])\n",
    "    \n",
    "    return model\n",
    "\n",
    "model = get_cnn()\n",
    "history = model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs, verbose=1, validation_split=0.7)\n",
    "\n",
    "score = model.evaluate(x_test, y_test, verbose=0)\n",
    "print('Test loss:', score[0])\n",
    "print('Test accuracy:', score[1])\n",
    "\n",
    "x = list(range(1, epochs+1))\n",
    "vy = history.history['val_loss']\n",
    "ty = history.history['loss']\n",
    "plot_loss(x, vy, ty)\n",
    "\n",
    "pt.add_row(['2-Conv+Dropout+Kernel(2x2)', sum(history.history['acc'])/len(history.history['acc']), \\\n",
    "            sum(history.history['val_acc'])/len(history.history['val_acc']), score[1]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 848
    },
    "colab_type": "code",
    "id": "N7V4Dv3_xCWP",
    "outputId": "9afa0074-00ab-41ef-ab0b-872e0001e471"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 18000 samples, validate on 42000 samples\n",
      "Epoch 1/12\n",
      "18000/18000 [==============================] - 3s 157us/step - loss: 0.6812 - acc: 0.7836 - val_loss: 0.1518 - val_acc: 0.9549\n",
      "Epoch 2/12\n",
      "18000/18000 [==============================] - 2s 138us/step - loss: 0.1973 - acc: 0.9416 - val_loss: 0.0965 - val_acc: 0.9695\n",
      "Epoch 3/12\n",
      "18000/18000 [==============================] - 3s 141us/step - loss: 0.1364 - acc: 0.9611 - val_loss: 0.0757 - val_acc: 0.9760\n",
      "Epoch 4/12\n",
      "18000/18000 [==============================] - 3s 144us/step - loss: 0.1050 - acc: 0.9688 - val_loss: 0.0667 - val_acc: 0.9795\n",
      "Epoch 5/12\n",
      "18000/18000 [==============================] - 3s 143us/step - loss: 0.0840 - acc: 0.9745 - val_loss: 0.0562 - val_acc: 0.9830\n",
      "Epoch 6/12\n",
      "18000/18000 [==============================] - 3s 144us/step - loss: 0.0726 - acc: 0.9783 - val_loss: 0.0577 - val_acc: 0.9830\n",
      "Epoch 7/12\n",
      "18000/18000 [==============================] - 3s 142us/step - loss: 0.0649 - acc: 0.9801 - val_loss: 0.0557 - val_acc: 0.9839\n",
      "Epoch 8/12\n",
      "18000/18000 [==============================] - 3s 144us/step - loss: 0.0548 - acc: 0.9823 - val_loss: 0.0478 - val_acc: 0.9856\n",
      "Epoch 9/12\n",
      "18000/18000 [==============================] - 3s 147us/step - loss: 0.0470 - acc: 0.9852 - val_loss: 0.0490 - val_acc: 0.9852\n",
      "Epoch 10/12\n",
      "18000/18000 [==============================] - 3s 156us/step - loss: 0.0443 - acc: 0.9861 - val_loss: 0.0505 - val_acc: 0.9856\n",
      "Epoch 11/12\n",
      "18000/18000 [==============================] - 3s 156us/step - loss: 0.0359 - acc: 0.9889 - val_loss: 0.0590 - val_acc: 0.9839\n",
      "Epoch 12/12\n",
      "18000/18000 [==============================] - 3s 156us/step - loss: 0.0351 - acc: 0.9889 - val_loss: 0.0486 - val_acc: 0.9858\n",
      "Test loss: 0.035210203452127095\n",
      "Test accuracy: 0.9876\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAe8AAAFYCAYAAAB6RnQAAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fcd4dffd160>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 34.4 s, sys: 5.83 s, total: 40.3 s\n",
      "Wall time: 33.7 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "def get_cnn():\n",
    "    model = Sequential()\n",
    "    \n",
    "    model.add(Conv2D(filters=32, kernel_size=(5, 5), strides=(2, 2), padding='valid', \\\n",
    "                     activation='relu', kernel_initializer='he_normal', input_shape=input_shape))\n",
    "    \n",
    "    model.add(Conv2D(64, (5, 5), activation='relu', kernel_initializer='he_normal'))\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "    model.add(Dropout(0.25))\n",
    "    \n",
    "    model.add(Flatten())\n",
    "    model.add(Dense(128, activation='relu', kernel_initializer='he_normal'))\n",
    "    model.add(Dropout(0.5))\n",
    "    model.add(Dense(num_classes, activation='softmax'))\n",
    "    \n",
    "    model.compile(loss='categorical_crossentropy', optimizer='Adam', metrics=['accuracy'])\n",
    "    \n",
    "    return model\n",
    "\n",
    "model = get_cnn()\n",
    "history = model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs, verbose=1, validation_split=0.7)\n",
    "\n",
    "score = model.evaluate(x_test, y_test, verbose=0)\n",
    "print('Test loss:', score[0])\n",
    "print('Test accuracy:', score[1])\n",
    "\n",
    "x = list(range(1, epochs+1))\n",
    "vy = history.history['val_loss']\n",
    "ty = history.history['loss']\n",
    "plot_loss(x, vy, ty)\n",
    "\n",
    "pt.add_row(['2-Conv+Dropout+Kernel(5x5)+Stride(2x2)+Dense(128)', sum(history.history['acc'])/len(history.history['acc']), \\\n",
    "            sum(history.history['val_acc'])/len(history.history['val_acc']), score[1]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 848
    },
    "colab_type": "code",
    "id": "bTs60twc0zVG",
    "outputId": "dc2ef2fa-4637-482b-9929-9caa5a4a6dcb"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 18000 samples, validate on 42000 samples\n",
      "Epoch 1/12\n",
      "18000/18000 [==============================] - 3s 171us/step - loss: 0.5352 - acc: 0.8303 - val_loss: 0.1178 - val_acc: 0.9664\n",
      "Epoch 2/12\n",
      "18000/18000 [==============================] - 3s 143us/step - loss: 0.1483 - acc: 0.9543 - val_loss: 0.0862 - val_acc: 0.9728\n",
      "Epoch 3/12\n",
      "18000/18000 [==============================] - 3s 142us/step - loss: 0.1032 - acc: 0.9679 - val_loss: 0.0619 - val_acc: 0.9804\n",
      "Epoch 4/12\n",
      "18000/18000 [==============================] - 3s 143us/step - loss: 0.0727 - acc: 0.9780 - val_loss: 0.0621 - val_acc: 0.9798\n",
      "Epoch 5/12\n",
      "18000/18000 [==============================] - 3s 143us/step - loss: 0.0619 - acc: 0.9808 - val_loss: 0.0532 - val_acc: 0.9838\n",
      "Epoch 6/12\n",
      "18000/18000 [==============================] - 3s 146us/step - loss: 0.0457 - acc: 0.9857 - val_loss: 0.0522 - val_acc: 0.9843\n",
      "Epoch 7/12\n",
      "18000/18000 [==============================] - 3s 144us/step - loss: 0.0420 - acc: 0.9862 - val_loss: 0.0482 - val_acc: 0.9850\n",
      "Epoch 8/12\n",
      "18000/18000 [==============================] - 3s 145us/step - loss: 0.0363 - acc: 0.9877 - val_loss: 0.0438 - val_acc: 0.9857\n",
      "Epoch 9/12\n",
      "18000/18000 [==============================] - 3s 143us/step - loss: 0.0293 - acc: 0.9902 - val_loss: 0.0519 - val_acc: 0.9846\n",
      "Epoch 10/12\n",
      "18000/18000 [==============================] - 3s 144us/step - loss: 0.0271 - acc: 0.9912 - val_loss: 0.0499 - val_acc: 0.9854\n",
      "Epoch 11/12\n",
      "18000/18000 [==============================] - 3s 143us/step - loss: 0.0259 - acc: 0.9908 - val_loss: 0.0471 - val_acc: 0.9862\n",
      "Epoch 12/12\n",
      "18000/18000 [==============================] - 3s 144us/step - loss: 0.0207 - acc: 0.9928 - val_loss: 0.0434 - val_acc: 0.9871\n",
      "Test loss: 0.033515525764345695\n",
      "Test accuracy: 0.9892\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAe8AAAFYCAYAAAB6RnQAAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4yLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvNQv5yAAAIABJREFUeJzs3XlYVGX/P/D3mY1thn0AcSUEEbBc\nynJLS0jNyvQpNddKMzOznjRTy/TJMDV7vlZqWaaVmcuvyLBF01KfTNRMcwH3FLdkUfZ9Zs7vj4ER\nFDjDMvv7dV1cM3Nm+3AzzPvc59znPoIoiiKIiIjIYchsXQARERHVD8ObiIjIwTC8iYiIHAzDm4iI\nyMEwvImIiBwMw5uIiMjBKGxdgLkyM/NtXYLF+fl5Iju7yNZl2D22k3nYTuZhO5mH7STNEm2k1Wpq\nXM6etx1RKOS2LsEhsJ3Mw3YyD9vJPGwnadZsI4Y3ERGRg2F4ExERORiGNxERkYNheBMRETkYhjcR\nEZGDYXgTERE5GIY3ERGRg2F4ExGRVTz77FM4ceJ4tWUffbQU69Z9WePjDx48gNdfnw4AmDHj5Vvu\n/+abDfj00xW1vt+ZM6dx4UIaAGDOnJkoLS1paOlISJiL33//rcHPb2oMbyIisor4+H749ddt1Zbt\n3Pkr4uIekHzuggX/rff77dr1Ky5evAAA+M9/3oabm3u9X8NeOcz0qERE5Nj69n0Azz03DpMmTQEA\nnDhxHFqtFlptEP74Yx9WrvwISqUSGo0Gb765oNpzBw7six9++AUHDuzH+++/C3//AAQEBCI0tDl0\nOh0SEuYiMzMDxcXFePrpCQgJaYbvvkvErl2/ws/PD2+8MRNffLEBBQX5ePvtN1FeXg6ZTIYZM2ZD\nEAQkJMxFaGhznDlzGpGR7TBjxmyzfqfly9/D0aOHodPp8eSTY9CjR1/89NP3SEzcCIVCibZtIzF1\n6qs1LmsM1wzvoiK4bd6E0kFDAHfnWRMjIjLX3Llu2LzZ/AiQyQCDwavOxzz8sA5z55bWer+fnz9C\nQ5sjNfUYoqNj8euv2xAf3x8AkJ+fjzlz3kJoaHPMm/cG9u1Lhqen5y2vsWLFUsyePQ8REZGYNm0K\nQkObIz8/D1273oMBAx7C5cuXMHv2DKxa9SXuvrsb+vTpi+joWNPzV678CA89NAh9+z6AHTu2Y9Wq\njzFu3LM4efI4/vOf+fDz88fgwQ8iPz8fGk3N84pX+uuvg/j777P48MNVFSsNI9Cp0z1Yv/5LLFq0\nBMHBIfjhhySUlpbUuKwxWwJcMrxVv26H9wsTka/Xo2TEaFuXQ0TkMuLj++OXX7YhOjoWv//+P3z4\n4SoAgK+vLxYufAt6vR5XrlxGly531Rje//zzDyIiIgEAHTt2RmlpKTQabxw/noKkpEQIggx5ebm1\nvv/Jk8cxceJkAEDnznfis89WAgCaN2+JgIBAAEBgoBaFhQWS4X3iRCo6duwMAPDw8EDbtm1x8eJF\nxMX1w6xZr6BfvwGIi+sHNzf3Gpc1hkuGt6FFCwCA4uhhAAxvInI9c+eW1tlLvplWq0FmZmGj37d3\n7/vwxRerEB/fDy1btoK3tzcA4O235+Gdd5agTZsw/Pe/C2t9vkx2Y6iWKIoAgG3btiAvLw/Llq1E\nXl4exo+v63tdMD2vvFwHQTC+nlxe/aQilY+piyAIqPow46Z4AaNHP4X4+AHYuXM7pkx5DsuWfVzj\nMh8fX8n3qI1LDljTRUZBlMkgT02xdSlERC7F09ML4eER+OKL1aZN5gBQWFiA4OAQ5Ofn4+DBP1Fe\nXl7j8wMDtbhw4TxEUcShQ38CAHJyctCsWShkMhl27frV9FxBEKDX66s9v337aBw8eAAA8NdffyIq\nqn2Df5eoqBhTDUVFRbhw4QJatGiFFSuWITAwEMOHj0JsbAdcvXq1xmWN4ZI9b3h6Qn9bOBQpxwBR\nBATB1hUREbmM+Pj+eOutOZgzZ55p2ZAhj+O558ahZctWGDlyDFat+hgTJky65bkTJkzC66+/ipCQ\nZggKCgYA9OlzP2bMeBmpqccwcOAjCAoKwurVn+COOzphyZJ3qm1+Hz9+It5+ex42b94EhUKJmTNn\nQ6fTmVX3ihVLsW7dGgBAmza3Ydq0GWjXLgrPP/8MdDodpk6dCg8PD3h6euHZZ5+CWq1GaGhzRERE\nYv/+vbcsawxBNGfbgB3IzMxv0tfTjB8L96Rvce1gCgwtWjbpazeUcbNU0/6ezojtZB62k3nYTuZh\nO0mzRBtptTXvd3fJzeYAoI+OAQAoUo/ZuBIiIqL6cdnw1sV0AAAouN+biIgcjOuGd0XPW86eNxER\nORiXDW9Di5YwaLyNg9aIiIgciMuGNwQB+ugYyM+eAYqLbV0NERGR2Vw3vGHcdC4YDFCcOmHrUoiI\niMzmmsd5V6gctCZPTYHujk42roaIyLl98MH/4eTJ47h+/RpKSkoQGtoc3t4+mD//Hcnn/vjjZnh5\nqdG7932Sj508eQJefnk6brutbVOUbZdcO7yrHC5m/iSBRETUEC+88G8AxiD++++zmDz5JbOf++CD\nD1uqLIfk2uEdFQ0AHLRGRGRDBw8ewPr1X6KoqAiTJ/8bhw79iZ07f4HBYEC3bj3w9NMT8OmnK+Dr\n64uwsHAkJm6EIMiQlnYOffr0xdNPT5B8D51Oh0WLEnDlymWUlZVh/PiJ6Nr1Hnz55WfYtWsHZDIZ\nevTohTFjnq5xmb1x6fCGWg19mzDjRC2cJpWIXIjX3NfhtnmT+U+QCfA31D0hZ+nDj6Jw7lsNqufs\n2TNYty4RKpUKhw79ieXLV0Imk2Ho0EEYNmxEtcempqbgq6++gcFgwOOPP2xWeG/btgUqlQpLl36M\nrKxMTJ78LNavT8T69V9i06YtkMvl2LTpGwCocZm9sWh4z58/H4cPH4YgCJg1axZuv/120333338/\nQkJCTGdyWbx4MYKDgy1ZTo100bFw+3EzZOlXYQhpZvX3JyIioG3bCKhUKgCAu7s7Jk+eALlcjpyc\nHOTl5VV7bLt2UXB3r98pNU+ePI5OnboAMJ7cRKVSIi8vF3369MVLL01CfHx/PPCA8UQpNS2zNxYL\n7/379yMtLQ0bNmzA2bNnMWvWLGzYsKHaYz755BN4edV9cndL08UYw1ueeozhTUQuo3DuW/XqJWu1\nGly34NzmSqUSAHD16j/YsGEtVq1aC09PT4wePfSWx958+k7zCNVO81leXg5BkGHatJlISzuPX3/d\nhhdeeBYff/x5jcsUCvvaUG2xQ8WSk5MRFxcHAAgPD0dubi4KCgos9XYNpouOBQAoUjhNKhGRreXk\n5MDPzw+enp44efIErl69WuvpQeuj6qlA09OvQiaTQRAErF79CVq3boOnnnoGGo0PsrIyb1lWVNT4\n85g3NYutSmRlZSEmJsZ029/fH5mZmVCr1aZlc+bMweXLl9GlSxdMnToVgg32OZtGnKcctfp7ExFR\ndRERkfDw8MRzzz2NDh06YtCgIXj33YW4/fY76vU68+e/adq03qXLXRg9+ikcOvQnXnjhWeh05Xjl\nlVlQq9XIycnGM8+MgYeHJ2Jjb0dISLNblnl7+1jiV20Ui50SdPbs2ejdu7ep9/3EE09g/vz5CAsL\nAwBs2rQJvXr1go+PD55//nkMHjwY/fvXvm9Bp9NDoWjIphIJBgPg4wO0aQMcZYATEZH9s1jPOygo\nCFlZWabbGRkZ0Gq1ptuPPvqo6fq9996LU6dO1Rne2dlFlikUgG9UNBR/HUTWpSzAzc1i7yOF58s1\nD9vJPGwn87CdzMN2kuYU5/Pu0aMHtm7dCgBISUlBUFCQaZN5fn4+xo0bh7KyMgDAH3/8gYiICEuV\nIkkXHQtBp4P89Cmb1UBERGQui/W8O3fujJiYGAwfPhyCIGDOnDlITEyERqNBfHw87r33XgwbNgxu\nbm6Ijo6us9dtaVX3e+tjO9isDiIiInNYdOz7tGnTqt2OiooyXR87dizGjh1rybc3m2nEeWoKp0kl\nIiK759JnFaukj66YJjWV06QSEZH9Y3gDEL19oG/VGopUHutNRET2j+FdQRcdA1lmBoSMDFuXQkRE\nVCeGd4WqpwclIiKyZwzvClUHrREREdkzhncFvSm82fMmIiL7xvCuoA+7DaKHB+TseRMRkZ1jeFeS\ny6GLag/FqRNAE5zBhoiIyFIY3lXoomMhlJVBfua0rUshIiKqFcO7Cj1HnBMRkQNgeFfBEedEROQI\nGN5VVB7rLWfPm4iI7BjDuwrRzx/60ObseRMRkV1jeN9EFx0D+T9XIFy/ZutSiIiIasTwvome+72J\niMjOMbxvoovhTGtERGTfGN43qRxxzpnWiIjIXjG8b6IPbwtRpWLPm4iI7BbD+2YKBXTt2kNx4jig\n09m6GiIiolswvGugj46BUFIC+bm/bV0KERHRLRjeNeCgNSIismcM7xrcGLTG8CYiIvvD8K4B5zgn\nIiJ7xvCugRgYCH1QMBQp7HkTEZH9YXjXQh8dA/mlixByc2xdChERUTUM71roYjoAABTHU21cCRER\nUXUM71qYTg/KTedERGRnGN614KA1IiKyVwzvWugjIiEqFFCkHrV1KURERNUwvGujUkEf0c64z9tg\nsHU1REREJgzvOuhiYiEUFUF2/pytSyEiIjJheNeB+72JiMgeMbzrUDniXJHC/d5ERGQ/GN510Mew\n501ERPaH4V0HQ1AwDAEBPLsYERHZFYZ3XQQBuugOkKedh1CQb+tqiIiIADC8JZlmWuM0qUREZCcY\n3hJ0lfu9OU0qERHZCcnwLi8vx9WrVwEAJ06cwKZNm1BcXGzxwuyFvnLEOfd7ExGRnZAM7xkzZuCv\nv/5Ceno6XnjhBZw6dQozZsywRm12QRcZBVEu54hzIiKyG5LhnZ6ejv79++PHH3/EiBEjMH36dOTm\n5lqjNvvg7g592wjIU1MAUbR1NURERNLhXVZWBlEUsW3bNvTp0wcAUFRUZOm67IouOgaygnzILl6w\ndSlERETS4d21a1d06dIFWq0WYWFh+OyzzxAWFmaN2uyGaZpUDlojIiI7oJB6wLRp0zBhwgR4e3sD\nAOLi4jBy5EiLF2ZPqg5aKxsw0MbVEBGRq5Psee/atQs7duwAAEydOhVPP/206baU+fPnY9iwYRg+\nfDiOHDlS42PeffddjB49uh4lW58upgMATpNKRET2QTK8ly9fjl69emHXrl0wGAz49ttvsWbNGskX\n3r9/P9LS0rBhwwYkJCQgISHhlsecOXMGf/zxR8MqtyJDs1AYfH0h5+FiRERkByTD293dHf7+/ti1\naxcGDRoELy8vyGTSc7skJycjLi4OABAeHo7c3FwUFBRUe8yCBQvw73//u4GlW5EgQBcdC/nfZwEX\nG6xHRET2RzKFS0tLsXLlSvzvf/9Dt27dcP78eeTnS8/znZWVBT8/P9Ntf39/ZGZmmm4nJiaia9eu\naN68eQNLty5ddAwEUYTiBKdJJSIi25IcsDZv3jxs3LgRCxYsgJubG3bv3o1p06bV+43EKsdI5+Tk\nIDExEatXr0Z6erpZz/fz84RCIa/3+zaZu+8EVq6A38WzQL/7LPY2Wq3GYq/tTNhO5mE7mYftZB62\nkzRrtZFkeEdERGDs2LFITU3Ftm3bcP/99yM0NFTyhYOCgpCVlWW6nZGRAa1WCwDYu3cvrl+/jpEj\nR6KsrAwXLlzA/PnzMWvWrFpfLzvbtpurFa3awg9A0b4DKBw0zCLvodVqkJnJs5dJYTuZh+1kHraT\nedhO0izRRrWtDEhuNl+3bh3GjBmDH374AZs3b8bo0aPx7bffSr5hjx49sHXrVgBASkoKgoKCoFar\nAcA0Y9vGjRuxdOlSxMTE1Bnc9kDXrj1EQeCIcyIisjnJnvd3332Hn376CW5ubgCMs6s99dRTGDx4\ncJ3P69y5M2JiYjB8+HAIgoA5c+YgMTERGo0G8fHxTVO9NXl6Qn9buPEEJaIICIKtKyIiIhclGd4K\nhcIU3ADg6ekJpVJp1ovfvG88Kirqlse0aNHCrEPP7IE+OhaKzZsgu3IZhuYtbF0OERG5KMnwDgkJ\nwbx589C9e3cAwO7du9GsWTOLF2aPdNExcNu8yTjTGsObiIhsRHKf97x58xAcHIzExER8++23CA0N\nxbx586xRm92pnGlNzv3eRERkQ5I9bw8PD0yYMKHasn379uHuu++2WFH2SldljnMiIiJbkZ4qrQbL\nli1r6jocgqFlKxjUGo44JyIim2pQeFedcMWlCAL00TGQnzkNlJTYuhoiInJRDQpvwYUPk9LFxELQ\n66E4dcLWpRARkYuqdZ/3e++9V+NyURRx6dIlixVk73TRsQCMg9Z0t3e0cTVEROSKag1vubz2ecSH\nDBlikWIcgWnQWsoxlNq4FiIick21hvfkyZOtWYfD0LePBsAR50REZDsN2uftykS1BvrWbaBIOWqc\nJpWIiMjKGN4NoIvpANn165BlmHc6UyIioqZk1lnFCgoKrFGLw6jc7y1P4aZzIiKyPsnwPnnyJB55\n5BG8+uqrOHDggDVqsnuVI845WQsREdmC5PSoc+fOhcFgwL59+5CUlITFixejb9++GDp0KHx8fKxR\no925MeL8qI0rISIiV2TWPm+ZTIZWrVohJCQEZWVlSElJwciRI7F9+3ZL12eXDG3CIHp6sedNREQ2\nIdnz3rRpE7755hvk5OTg8ccfx+rVq+Hj44O8vDyMGjUKcXFx1qjTvshk0LWPhuLwIaCsDFCpbF0R\nERG5EMnw3r17N1588UXceeed1ZZ7e3tj7NixFivM3umiY6H88w/IT5+CPibW1uUQEZELMWuf96ZN\nm/D9999DEAS0a9cOjz76KNzd3fGvf/3LGjXapaqnB2V4ExGRNUnu8546dSqOHDmCqKgoREZG4sCB\nA3j55ZetUZtdqwxsBQ8XIyIiK5Pseefm5mLFihWm20888QRGjBhh0aIcgY7TpBIRkY1I9rxbtGiB\nzMxM0+2srCy0bt3aokU5AtHHF/qWrSDniHMiIrIyyZ73lStXEB8fj7Zt28JgMODcuXMIDw/HyJEj\nAQBr1661eJH2ShcdA7etP0HIzISo1dq6HCIichGS4f3SSy9Zow6HVBneiuMpKNf2sXU5RETkIiTD\nu2vXrjhw4ACOHj0KQRBwxx13oFOnTtaoze7po28MWiu/t49tiyEiIpchuc/7vffew6JFi5CRkYH0\n9HS89dZb1QawuTJdTAcAHLRGRETWJdnz3rdvH9avXw+ZzJjzOp0Oo0aNwrPPPmvx4uydPuw2iO7u\nHLRGRERWJdnzNhgMpuAGAIVCAUEQLFqUw5DLoYtqD8XJ44BOZ+tqiIjIRUj2vGNjYzFx4kR0794d\nALBnzx506NDB4oU5Cl10LJR/HYL87Bno20XZuhwiInIBkuE9a9Ys/PTTTzh8+DAEQcAjjzyCAQMG\nWKM2h6CvcnpQhjcREVmDZHivXLkSEyZMwMCBA61Rj8O5MWgtBaVDHrdxNURE5Aok93mfOnUKaWlp\n1qjFIVVOkyrniHMiIrISyZ73yZMnMXDgQPj4+ECpVEIURQiCgJ07d1qhPPsn+gdA3ywUCo44JyIi\nK5EM748++uiWZcXFxRYpxlHpomPg9ss2CNnXIfr527ocIiJycpKbzd944w00b9682s+rr75qjdoc\nhmmmNfa+iYjICmrteSclJWHZsmW4cuUK+vTpY1peXl6OwMBAa9TmMHSV5/ZOPYbyHr1sXA0RETm7\nWsP7kUcewcCBA/Haa6/hhRdeMC2XyWQICgqySnGOQlfR8+ZMa0REZA117vOWy+VYsGABTpw4gZyc\nHIiiCAA4f/48unXrZpUCHYE+vC1ElYpznBMRkVVIDlibMmUKjh8/jpCQENMyQRAY3lUpldBFRkFx\n4jig1wNyua0rIiIiJyYZ3pcuXcK2bdusUYtD00fHQHnsCOTn/oa+bYStyyEiIicmOdo8LCwMZWVl\n1qjFoVXOtMbJWoiIyNIke94ymQwDBw7E7bffDnmVzcGLFi2yaGGORlc5x3nqMZQ9MtjG1RARkTOT\nDO/u3bubzihGtdPxWG8iIrISyfAePHgwTp06hQsXLiAuLg55eXnw9va2Rm0ORdRqYdAGQZHCzeZE\nRGRZkuH92Wef4fvvv0dZWRni4uKwfPlyeHt7Y9KkSdaoz6HoomOg2rUDQl4uRG8fW5dDREROSnLA\n2vfff4+NGzfCx8cYRtOnTzf7pCTz58/HsGHDMHz4cBw5cqTafRs3bsTQoUMxfPhwzJ0713QMuSO7\nMWgt1caVEBGRM5MMby8vL8hkNx4mk8mq3a7N/v37kZaWhg0bNiAhIQEJCQmm+4qLi/HDDz9g7dq1\nWL9+Pf7++28cOnSogb+C/ag6aI2IiMhSJDebt2rVCkuXLkVeXh5+/vln/PjjjwgPD5d84eTkZMTF\nxQEAwsPDkZubi4KCAqjVanh4eODzzz8HYAzygoICaLXaRv4qtsdBa0REZA1mnVXMw8MDwcHBSEpK\nwh133IE5c+ZIvnBWVhb8/PxMt/39/ZGZmVntMR9//DHi4+PRv39/tGzZsgHl2xd9RCREhQKKlKO2\nLoWIiJyYZM9bqVRi2LBhGDduHDIzM5GWlgaFQvJpt6hpn/aECRMwZswYPPPMM+jSpQu6dOlS6/P9\n/DyhUNj7tKMaoH17KE+kQhvgBZixe+FmWq3GAnU5H7aTedhO5mE7mYftJM1abSSZwvPmzUNUVBTi\n4+MxYsQIxMTEICkpCW+++WadzwsKCkJWVpbpdkZGhmnTeE5ODk6fPo277roL7u7uuPfee3Hw4ME6\nwzs7u8jc38mmNJHt4X70KK4dOApD2G31eq5Wq0FmZr6FKnMebCfzsJ3Mw3YyD9tJmiXaqLaVAcmu\nYWpqKh5//HH89NNPePTRR7FkyRKkpaVJvmGPHj2wdetWAEBKSgqCgoKgVqsBADqdDjNmzEBhYSEA\n4OjRowgLCzP7l7Fn3O9NRESWJtnzrtzcvXPnTrz00ksAYNZc5507d0ZMTAyGDx8OQRAwZ84cJCYm\nQqPRID4+Hs8//zzGjBkDhUKBdu3aoW/fvo38VeyDLqbKNKkDH7ZxNURE5IwkwzssLAwPPvgg/P39\n0b59e2zatMl0zLeUadOmVbsdFRVluj5kyBAMGTKknuXaP31lz5szrRERkYVIhvdbb72FU6dOmQ4P\ni4iIwP3332/xwhyVITgEhoAAHutNREQWI7nP+/jx47h69SpUKhX+7//+D4sWLcKpU6esUZtjEgTo\nomMhP38OKCiwdTVEROSEJMP7rbfeQlhYGA4cOICjR49i9uzZeP/9961Rm8MyzbR2gtOkEhFR05MM\nbzc3N7Rp0wa//PILhg4dirZt25o1Paor44hzIiKyJMkULi4uxk8//YTt27ejZ8+eyMnJQV5enjVq\nc1j6yp43Z1ojIiILkAzvl19+GZs3b8bLL78MtVqNNWvW4Mknn7RCaY5L1649RJmMPW8iIrIIydHm\n99xzD26//XacO3cOqampGD9+PDw8PKxRm+Nyd4e+bQTkqSmAKAKCYOuKiIjIiUiG9/bt2zF37lyE\nhITAYDAgKysL8+bNQ+/eva1Rn8PSRcfA/dRJyC5dhKFlK1uXQ0RETkQyvFeuXImkpCT4+/sDANLT\n0/Hiiy8yvCXoo2OBTYlQpKagjOFNRERNSHKft1KpNAU3AAQHB0OpVFq0KGeg46A1IiKyEMmet5eX\nF1atWoXu3bsDAHbv3g0vLy+LF+bodDEdAMC435uIiKgJSYZ3QkIC3nvvPSQlJUEQBHTs2BHz58+3\nRm0OzRDaHAYfX06TSkRETU4yvI8dOyZ57m6qgSBAFx0D5b5koKgI8PS0dUVEROQkJPd5f/bZZ9Dp\ndNaoxenoo2MgGAxQnDph61KIiMiJSPa8NRoNBg4ciOjo6GoD1RYtWmTRwpxB5X5vRcox6Dp2tnE1\nRETkLCTD+7777sN9991njVqcTuWIczn3exMRUROqM7wvXryIwYMHm24XFxcjPT0dbdq0sXRdTkHX\nrj1EQeA0qURE1KRq3eednJyMJ554Avn5+aZlFy9exPjx43HsGHuSZvHygj7sNuOIc1G0dTVEROQk\nag3vpUuXYtWqVdBoNKZlkZGR+PDDD7FkyRKrFOcM9NGxkGVnQ/bPFVuXQkRETqLW8BZFEZGRkbcs\nj4iIQGlpqUWLcia6mMpze3NrBRERNY1aw7uoqKjWJ+Xk5FikGGekizaGN2daIyKiplJreEdERGDd\nunW3LP/kk09wxx13WLQoZ2Ka45w9byIiaiK1jjafPn06nn/+eXz33XeIjY2FwWDAwYMHoVarsWLF\nCmvW6NAMLVvBoNZwxDkRETWZWsNbq9Vi48aNSE5OxunTpyGXyzFgwADcdddd1qzP8clk0LePhuLg\nAaCkBHB3t3VFRETk4CQnaenWrRu6detmjVqcli4mFso/9kFx+iR0HbjLgYiIGkdybnNqPNOgtRTu\n9yYiosZjeFtBZXhzvzcRETWFWjebJycn1/lEbko3n759ewAMbyIiahq1hvfy5ctrfZIgCAzvehA1\n3tC3agNF6lHjNKmCYOuSiIjIgdUa3mvWrKn1SVu3brVIMc5MFxMLt5++h5CRATE42NblEBGRA5Mc\nbX7lyhV8+eWXyM7OBgCUlZVh37596Nevn8WLcya66Bi4/fQ9FKnHUM7wJiKiRpAcsDZ9+nT4+vri\nr7/+QmxsLLKzs7Fo0SJr1OZUOGiNiIiaimR4y+VyTJgwAYGBgRg5ciQ+/PBDrF271hq1ORV9DKdJ\nJSKipiEZ3qWlpbh69SoEQcDFixehUChw+fJla9TmVPStwyB6ekLBY72JiKiRJPd5jx8/Hnv27MG4\nceMwaNAgyOVyPPTQQ9aozbnI5dC1j4biyGGgrAxQqWxdEREROSjJ8I6Li4MoihAEAfv370deXh78\n/f2tUZvT0UXHQvnnAcjPnIa+4mxjRERE9SW52XzLli147rnnAAAKhQITJ07Eli1bLF6YM+LpQYmI\nqClIhvdnn32Gd955x3R71apVWL16tUWLclZ6jjgnIqImIBneoihCo9GYbqvVagicIaxBTD3vlKM2\nroSIiByZ5D7v2NhYvPTSS+jzd24HAAAgAElEQVTatStEUcRvv/2G2NhYa9TmdEQfX+hbtIScPW8i\nImoEyfB+/fXXkZSUhCNHjkAQBDz88MMYMGCANWpzSrroGLj9vAVCVhbEwEBbl0NERA6o1vDOyMhA\nUFAQLl26hM6dO6Nz586m+y5fvoyWLVtapUBno4uOhdvPW6A4noLyXr1tXQ4RETmgWsN74cKFePfd\ndzF27Nhq+7grDxv75ZdfrFKgs9FXGXHO8CYiooaoNbzfffddAMC6desQzBNpNBldTAcA4ExrRETU\nYJKjzV955ZUGv/j8+fMxbNgwDB8+HEeOHKl23969ezF06FAMHz4cM2fOhMFgaPD7OBJ92G0Q3d05\naI2IiBpMcsBamzZtMH36dHTq1AlKpdK0/LHHHqvzefv370daWho2bNiAs2fPYtasWdiwYYPp/jfe\neANffPEFQkJCMGXKFPz222/o3dsFNiMrFNC1aw/FiVRApwMUkn8CIiKiaiSTo7y8HHK5/Jaes1R4\nJycnIy4uDgAQHh6O3NxcFBQUQK1WAwASExNN1/39/U3nC3cFuugYKA8fgvzvs9BHtrN1OURE5GAk\nw/vtt98GAOTk5EAQBPj4+Jj1wllZWYiJuTF/t7+/PzIzM02BXXmZkZGB33//HS+++GKdr+fn5wmF\nQm7We9u9rl2AdV/C/+IZoMed1e7SajW1PImqYjuZh+1kHraTedhO0qzVRpLhffDgQUyfPh2FhYUQ\nRRG+vr5455130KFDh3q9kSiKtyy7du0aJk6ciDlz5sDPz6/O52dnF9Xr/eyZsnUEfAEU7T2Awr4D\nTcu1Wg0yM/NtV5iDYDuZh+1kHraTedhO0izRRrWtDEiG97vvvovly5cjMjISAJCamoqEhASsXbu2\nzucFBQUhKyvLdDsjIwNardZ0u6CgAM888wxeeukl9OzZ06xfwlno2hu3SMh5ghIiImoAydHmMpnM\nFNwAEB0dDblcevN1jx49sHXrVgBASkoKgoKCTJvKAWDBggUYO3Ys7r333obU7dDEgADoQ5rxBCVE\nRNQgkj1vmUyGn3/+Gd27dwcA/O9//zMrvDt37oyYmBgMHz4cgiBgzpw5SExMhEajQc+ePbFp0yak\npaXh66+/BgA89NBDGDZsWCN/Hcehj46B6tftEHKyIfrWvcuAiIioKsnw/s9//oN58+bhtddegyAI\n6NixI/7zn/+Y9eLTpk2rdjsqKsp0/dgx195krIuOherX7VCkpqC8u2vtNiAiosYx6zjvTz/91Bq1\nuBRdjPHMbPLUYwxvIiKqF8nwHjFixC3n75bL5QgLC8OkSZM4dWoD6aKN4c393kREVF+S4d29e3ec\nO3cO/fr1g0wmw/bt29GsWTP4+Phg5syZWLVqlTXqdDr6thEQlUooOOKciIjqSTK8//zzT6xevdp0\nOy4uDhMmTMDHH3/MM4s1hlIJfWQUFCeOA3o9YMYgQCIiIsCMQ8WuXbuG69evm27n5+fjypUryMvL\nQ34+D9hvDF1MLISiIsjP/23rUoiIyIFI9rzHjBmDAQMGoHnz5hAEAZcuXcKzzz6LHTt2uNShXZZQ\nud9bnpoCfXiEjashIiJHIRnejz32GPr374/z58/DYDCgVatW8PX1tUZtTk8XbZxpTZFyDGUPP2rj\naoiIyFFIbjbPzc3FsmXL8Nlnn+H222/HwYMHq21Gp4bjiHMiImoIyfB+/fXX0axZM1y6dAkAUFZW\nhldffdXihbkCMSgIhkAtw5uIiOpFMryvX7+OMWPGQKlUAgD69++PkpISixfmKnQxsZBfOA8hP8/W\npRARkYOQDG8AKC8vN03UkpWVhaIi5zk9p63dGLSWauNKiIjIUUiG98iRI/HYY4/hzJkzmDhxIgYN\nGoRx48ZZozaXYBq0xslaiIjITJKjzR988EF07twZhw4dgkqlwptvvglvb29r1OYSOGiNiIjqS7Ln\nPW7cOISEhGDAgAHo27cvgoKCMHLkSGvU5hL0ke0gKhTseRMRkdlq7XknJSVh2bJluHLlCvr06WNa\nXl5ejsDAQGvU5hrc3KCPiIQ8NQUwGGxdDREROYBaw/uRRx7BwIED8dprr+GFF14wLZfJZDyTWBPT\ntY+B+/FU4Px5QKO1dTlERGTn6txsLpfLsWDBAvj6+kIQBAiCgNLSUgwdOtRa9bmEyv3eOHLEtoUQ\nEZFDkBywtnLlSnz00UcoKyuDp6cnSktL8fDDD1ujNpehjzGOOMeRI0CPvrYthoiI7J7kgLUtW7Zg\nz549uOOOO7B3714sXrwYERE8iUZTYs+biIjqQzK8vby8oFKpUF5eDgDo27cvz+PdxAwhzWDw9wcO\nH7Z1KURE5AAkN5v7+PggKSkJkZGRmDlzJsLDw5GRkWGN2lyHIEAXHQvV778BBQWAWm3rioiIyI5J\n9rwXLlyIzp07Y+bMmWjdujWuXr2K//73v9aozaXoomMAUYTXovkA544nIqI61BneFy9ehIeHB1q0\naAEPDw+MHTsWo0aNQlRUlLXqcxklT44HWreG50dL4Xd/Dyj2Jtu6JCIislO1hndycjKeeOIJ5Ofn\nm5ZdvHgR48ePx7FjnA2sqenbRgDHjqHomYmQnz0Dv0f6Qf3qyzzbGBER3aLW8F66dClWrVoFjUZj\nWhYZGYkPP/wQS5YssUpxLketRmHCIuR8/zN0ke3gsXol/HrdDdW2LbaujIiI7Eit4S2KIiIjI29Z\nHhERgdLSUosW5ep0d92N7F92o3Dqq5BlZsBn5FBoJo6DkJVl69KIiMgO1BredZ2zOycnxyLFUBVu\nbih69TVkb/8N5Z27wD3x/8G/111w+3oDIIq2ro6IiGyo1vCOiIjAunXrbln+ySef4I477rBoUXSD\nvn00cn7YjoI350MoLob3pGfgPfJxyC5fsnVpRERkI4Io1tyNy8zMxPPPPw+ZTIbY2FgYDAYcPHgQ\narUaK1asgJeXl1ULzczMl36Qg9NqNXX+nrK089BMfRGq/+2AwUuNwtfnouSp8YBM8og/pyLVTmTE\ndjIP28k8bCdplmgjrVZT4/Jaw7tScnIyTp8+DblcjsjISNx1111NWpi5XOFDY9YfXhThtn4t1G/M\ngiw3B+V3d0P+fz+APuLW8QnOil8i5mE7mYftZB62kzRrhrfkDGvdunVDt27dmrQYagRBQOkTo1B2\nfzw0s16B2+ZN8LuvO4qmvoqiyS8BSqWtKyQiIgtzre2tTkQMDkbep18gd/VaGHz94PX2PPjF94bi\nr4O2Lo2IiCyM4e3gygY+jOzf/0DxqLFQpB6Db//74TX3daCOowWIiMixMbydgOjji4L/foCcbzbD\n0LIVPJe/D/8+3aDc/T9bl0ZERBbA8HYi5b164/quvSiaNAWyC2nwHfIQ1C+/ACGXx+UTETkThrez\n8fRE4dy3kLPlV+iiY+Hx5efw69kVqh+/t3VlRETURBjeTkrXsTOyt+1C4czZkGVfh8+TI+A9bgyE\n9HRbl0ZERI3kkuF99KgMHTt6YepUN5w+7cRNoFSi6N+vIHvHHpR3vQdumzcZp1hdv5ZTrBIROTAn\nTq7aeXuLcHMD1qxRoUcPL4we7YHkZLnT5pk+IhI5SVuQ//ZioFwH7ynPwWfoo5Clnbd1aURE1AAu\nGd6tW4vYs6cQq1YVo0sXPbZuVWDQIE/07++JpCQFdDpbV2gBMhlKxk1A9m/7UNo3HqpdO+Df+x54\nfLQU0OttXR0REdWDS4Y3AMjlwEMP6fDTT0X4/vtCPPhgOf76S4bx4z1wzz1eWLlSiYICW1fZ9Awt\nWiLvq6+Rt/wTiO7uUL8xC74PxUN+PNXWpRERkZlcNryr6trVgM8+K8GePYUYO7YMGRkCZs1yR+fO\nasyfr0J6umDrEpuWIKD0sWG4vvsASoY8DuWfB+AX1wueCxMAnqudiMjuMbyrCA8X8c47pTh4sBCv\nvFIKuVzEkiVu6NLFCy+95IaTJ52rucTAQOR/9Cly126EQRsEr3cXwi+uFxQH9tu6NCIiqoNzpVET\nCQwU8corZTh4sBDvvFOCFi1EfPWVCr16eWHECA/s3u1cg9vK4vsj+7d9KH5qPBQnT8B3YDy8XpsO\np9xvQETkBCwa3vPnz8ewYcMwfPhwHDlypNp9paWlePXVVzFkyBBLltAoHh7A2LHl2LOnEJ9/Xoy7\n79Zh+3YFhgzxxAMPeOLbb51ncJuo8UbBwv8iJ2kL9LeFw/OTj+Df+x6ofvmZh5UREdkZi4X3/v37\nkZaWhg0bNiAhIQEJCQnV7l+0aBHat29vqbdvUjIZMGCADps3F+PHHwvx0EPlOHJEhmef9UDXrl5Y\nscJ5BreV39Md2Tv2oPClaZBduQyfJx6D7wN94Pb/1gNlZbYuj4iIYMHwTk5ORlxcHAAgPDwcubm5\nKKiScP/+979N9zuSO+80YNWqEuzdW4inny7DtWsCZs92R8eOasybp8I//zjB4DZ3dxTNegPZ2/6H\n0oGPQHH0MLyfnwD/LrHw/O8iCFlZtq6QiMilCaJomW2is2fPRu/evU0BPWLECCQkJCAsLMz0mEuX\nLmHKlClITEyUfD2dTg+FQm6JUhvl2jXgww+BDz4AMjIApRIYMQKYOhXo0MHW1TWRc+eApUuBlSuB\nvDzAzQ0YNQp48UUn+iWJiByHwlpv1Nh1hOxs+z0/9bPPAmPHAl9/rcTy5Up8/rkcn38O3HefDpMm\nleHee/UQzOiQa7UaZGbmW77g+lIHAjPmQpg8FW7r18Lz4w8h//RT4NNPUdarD4qffQ5lcf2M+xes\nwG7byc6wnczDdjIP20maJdpIq9XUuNxi37ZBQUHIqrJ5NSMjA1qt1lJvZ3Pu7sCoUeXYvbsIa9YU\noVs3HXbsUODxxz3Rt68nvv5agfJyW1fZOKJag5LxE3E9+SBy12xAWa/eUP22Ez6jhsGvexe4f7qC\nI9SJiKzAYuHdo0cPbN26FQCQkpKCoKAgqNVqS72d3ZDJgH799Pjuu2Js2VKIQYPKkZoqw6RJHrjr\nLi8sX65EvqOvvMrlKOs3ALnfbMb1HXtQPGI05JcvQTPzFQR0bA+vOa9BdiHN1lUSETkti+3zBoDF\nixfjwIEDEAQBc+bMQWpqKjQaDeLj4zFlyhRcvXoVp0+fRmxsLIYOHYqHH3641tdy5M01aWkCPv5Y\nhbVrlSgqEqDRiBg9uhzPPFOG5s1vNL8jb5YSMjPh8cUqeKz6BLLMDIgyGcoefBhFEyZBd/c9MGu/\ngZkcuZ2sie1kHraTedhO0qy52dyi4d2UnOFDk50NfPGFCp98okRGhgwKhYhHHzXuF4+NNTjHP0dp\nKdy+S4THiuVQHj0MACi/oxOKJzyH0kFDAJWq0W/hFO1kBWwn87CdzMN2ksbwroEzfWhKS4FvvlFg\n+XIVTp0yjqDv3VuHGTMU6NgxH3L7G1Rff6II5b5keKxYDtVP30MwGKAPDkHJU+NRPOZpiIGBDX5p\nfomYh+1kHraTedhO0hjeNXDGD43BAPz6qxzLl6uwe7dx4H9wsAGPPqrD4MHl6NTJ0JRbm21GlnYe\nHp9+DPe1X0CWnwfRzQ0ljw1D8TPPQR8dU+/X45eIedhO5mE7mYftJI3hXQNn/9AcPizDxo1e+H//\nT0ROjjGx27QxYPDgcgwerENUlMHGFTaeUJB/41Cz8+cAoEGHmvFLxDxsJ/OwnczDdpLG8K6BK3xo\ntFoNLl/Ox86dciQmKrFliwJFRcYgb99ejyFDdHj00XK0bu0Qf7La6fVQbf8ZHh8vh+q3XQAA3W3h\nKH5mIkqGjQQkjkrgl4h52E7mYTuZh+0kjeFdA1f40Nz8hy8sBLZtUyAxUYFff1WgrMwY5F266PGv\nf5XjkUd0CApyiD9freQpx+DxyYdw/2YjhNJSGLx9UDJyDIrHTYChVesan8MvEfOwnczDdjIP20ka\nw7sGrvChqesPn5MD/PijAomJSuzeLYfBIEAmE9Gzpx5DhpRj4EAdfHysXHATqs+hZvwSMQ/byTxs\nJ/OwnaQxvGvgCh8ac//w6ekCkpKMQf7nn8ah6SqViPvv12HIEB0eeEAHT09LV2shZhxqxi8R87Cd\nzMN2Mg/bSRrDuwau8KFpyB8+LU3Apk1KJCYqcPy4Mcg9PUX076/DkCHl6NNH3xSHVltfHYeaeU2Z\nhEyF88/W11j8sjUP28k8bCdpDO8auMKHprF/+BMnZPj2W2OPPC3NOHLbz0/EQw8ZR6x366Z3yGPI\nbz7UDAAMAQHQh0dAFxEJfXgE9G0joI+IgL5VG+Op3YhftmZiO5mH7SSN4V0DV/jQNNUfXhSBQ4dk\n+PZbJTZtUiA93RjkISEGDBpk7JF37Oh4x5BXHmqm+X0XdKnHIb+QBkGvr/YYUaGAPuw2U6DfCPe2\nEP0DbFS5bfDL1jxsJ/OwnaQxvGvgCh8aS/zh9XogOVmOb79VYPNmZbVjyIcMMfbI27VzrGPITe1U\nVgb5+XOQnz4F+dnTUJw5bbouy8m55Xmm3nrbCOjbRjp9b51ftuZhO5mH7SSN4V0DV/jQWPqfo6wM\nNR5DHh194xjyVq3s/+Mg2U6iCOHaNSjOnIK8SqDLz5yGPO18zb31NmE3Ar1tBHRtIx2+t84vW/Ow\nnczDdpLG8K6BK3xorPnPUdsx5HfeaTz0zJ6PIW9UO1X21s+chvzMKWNvveJ6jb11f3/o20Yae+vh\nEdBHVAR86zZ231vnl6152E7mYTtJY3jXwBU+NLb656jrGPK4OB1UKuN+dMB4Kf0j1PPx1R9b8/ME\n0/WAABUCAkrQqpUBrVoZ0Ly5CIWikY1wc2+9ItCle+vGTfC69tEo73qPcWIZOxlMwC9b87CdzMN2\nksbwroErfGjs4Z+jpmPI7Z1cLqJ5c9EU5q1aVb8eFCSaO216zerRW9cHh6D87m7Qdb3beBnTAY1f\ns2gYe/g8OQK2k3nYTtIY3jVwhQ+Nvf1zpKUJ+OsvOUTR2Jms/AGq3zb+iDUsq+vxUo8Va32+UumF\nI0eKceGCDGlpMly4IMPFiwKuXq05od3dRbRoUT3UW7e+cd3Xt4Gd5Sq9dcXhQ1Du3wfFvmTIM9Jv\nPMTTC+Vd7kR513tQ3vUe6O68C6LGuwFvVn/29nmyV2wn87CdpDG8a+AKHxr+c5intnYqKQEuXRKq\nhfqFC0LFpQzZ2TUntEZTe6+9VSsDvLzqUZwoQpZ2Hsp9yVDu3wflH3uhOHH8xt0yGXTRsaaeeXnX\ne2Bo3qK+TWAWfp7Mw3YyD9tJGsO7Bq7woeE/h3ka2k75+agW6hcv3rieliYzjb6/WWBgzaHeqpUB\nLVqIkjPYCdnXoTyw39QzV/51EEJJiel+fYuWKO96N8rvugfld3eDvn00mmI2HX6ezMN2Mg/bSRrD\nuwau8KHhP4d5LNFOoghcuyZU66mnpd24fumSgPLyW8NdEESEhooIDBQhlxtPSa5QGK9X/igUxv3y\nlbfdZWUIzz2E6OzfEZW1B5GZe+BdmmV6zWKVN86H3I3zzbshrWU3XG7eFaKnl1mvrVBU1gCEhHhA\nry+Etzfg7S1CrRbh5WU34+nshj3935WWGsedCAKgUhl3+ahUgJub2ae7txh7aid7Zc3wts1IGiI7\nIwhAYKAxhDt3vnXSGr0euHpVuGVTfOX106dl0OuNj9PppNJRCaBnxQ8AiIjAafTA7+iJ3ehR9jva\nX9iG9he2AcmADnIcQifsRk/8jh74HT1wFc3M/M2qb/OXyURoNMYw12iMP97eqHLdeH/V21Xvr3yu\nnR8lZ5cMBiArS8DlywIuXZLh8mUBly9Xv8zIqD2hlUoRbm6Am5vx8uZwv7FchLs7KpbfeI7x8cb7\njcvqd7+3N0zjX8j22PO2I1yzNY8jtJPBAOh0MAV61WC/9T6h4r4bz5Nfz4T3sf3wS02Gf2oyfM8e\nhFxfbnr9PG0YroZ3w5WwbrjcphsyA9tDL1ZdgQDkcnf8808p8vIEFBQIyMsTkJ+PisvKH8BgqP+3\nsYeHsSdfU/BX9vJvDX4RPj6Aj48IHx9jwNiDpvo8FRTgljC+dEmGK1duXFbOp3Azlcq4BadFCwNC\nQoyDNcvKjD3x0lKh2mVZGVBSIlTcX3mf8XNkaTKZCA8P49/f09N46eFhPBlS5fKql3Utr/r8my9t\nvZWhobjZvAb2/mXdFBwhlOyBS7ZTSQkUfx2Ccn8ylPv3Qrl/b7XD1Ay+vii/627jiPa7u6H8jk7Q\ntgqSbCdRNE7Yk59fc7jn5aHK9ZrvLygQUFxc/+Aw9ubEijC/Eep1LfP1Na4Q+Pg0Xe/fnM+TTmfc\n8lJbj/nyZZlp6uGaBAUZ5yNo3rz6ZYsWxsvAwEYezlhRY2W43xz4xuUCSkpuvd+4MmC8/+bnVL1f\nr1ciN1eH4mIBxcW46bJpVxzc3WtaCbh1xUGtBtRq44qhWn1jRbFyufE+43JrnF2R4V0DV/iydslQ\nagC2EwCDAfLTp4xBvs8Y6PLz50x3i0olhDZtUO6phqhWQ9RoIKo1Fde9IarVMJiWaYz3a2667ull\ndheorAymnnzVYL856PPyjCsBublVf4Dc3JrHFNTF01O8KdyNvX5fX6llxhWAyjGBgYEanDqVXy2I\nq/aYL18WcPWqUOsWCi+vGyFcNZxbtBARGmpAaKhx07Ojq+v/zmAwBnxR0a3BXlR0c9DfuCwsrHl5\n5WXV1ysqathWokoqlTHYvbyqh7zxsvp1jabqSsGtj6ttxZHhXQNX+LJmKJmH7VQzIT3d1CtX7k+G\n8p8rEPPyIRQVNvg1DabArwh0r6pBr4ahYkXAFPrqm+5Xe0PUaAAPjzp3looiUFyMKsFuvJ6TY7xd\nuTwvDzUuy80V6r3ZuPILPCdHhqKimh8jlxs3Z4eGVu8pV+05e3u7xn5gW//fiaJxJbFqmBcUGFcM\nCwpQcXnjduX1/HzjdePPjZXLwkLjzI0N4eEhwsureshrNEB0tALTp+c36amXGd4OwNb/HI6C7WQe\nUzvpdBAKCyAUFEDIz4eQn2e8XpAPWX4+hIJ84/LK+wvyqjw2H7KCKo8pLW1QLaJMZuzx+/rCEBgI\nQ4DxRwyovB4AscpyQ0Ag6nOAfeXm/5p69DX19G8Ev4CAABmCg8ur9Zgrwzk4WGzSL2JH5mz/dwYD\nqq0AVIb8zeFffcXgxopA1edUHmbq6QkcPJgPf/+mq5OjzYlclUIB0ccXoo9v41+rrMwY5FXDvTDf\ndN24PM94vbByZaBiJSE/D0J2NhRHDkMoL5d8K9HDo0rIB9wI9sCK0A/UwlCxXAwMhFqtgVoNNG9e\nv/6IMZRKpB9ITkUmg2nfeEhI4/qwer1x5bFZMw0KG76hq14Y3kRkPpUKon9A406VKorGIM/Kguxa\nFmTXrkF2LQvCtSzITMuyTPcrTh6vNqlNrS+rUt0a9oFVe/eBpqA3BAQYV2YcdVgz2RW5HPD2Nva8\nGd5E5JwEAaK3D0RvHxhuC5d+fMU28cpQNwb9tepBX3lf1jXIz/0N2bEj0i8rlxs3z4c2g3eAFoaQ\nZjAEB8MQ3OzG9ZBmMGiD7P70r+R6GN5EZN8EAVCrYVCrYWjdxrznFBdDdv1atR58ZS/f2MPPhCzL\nGPo4fRpuf/1V60uJgmDsvQeHQB8SAkNwCAwhITAEhRjDvXJZUDBDnqyG4U1EzsfDA4bmLcw66YtW\nq0HWuSuQpV+F7OpVyK7+A1l6uvEyo2JZ+lXIzp+DIuVona9lCAw09tyDg6GvrSfPkKcmwPAmIpcn\nqjXQqzXQh0fU+TihIP9GyFe9TP+n/iEfZOzB62/uyQcHQ/RSQ3RzA9zdIbq5m67b6tzwZH/4SSAi\nMlO9Q76yB1815CuXXUiDIvVY/d5foYDo5g64u1ULddHNDXCrCHr3yutuED08qtx368qA6OZW/fXc\n3Wt/jM6Dk5vbEYY3EVETa3DIp6dDln4VQnERUFoKoaTEeGx9qfFSKCkBSkoglJZAKCmFUFQEIfu6\n8TElJRAsPG2HFsYVCCgUEBVKQCEHFEqISmXFMuN9N5ZV3F+xDAp5xXJllccqjMvkCojKysdVXJcr\nAKWy4r1uWqZSGecN8PWH6O8Pg58/RF/fJjmdriNgeBMR2Yi5IW/ei4lAeTmE0hKgpNQY8NWulwLF\nxcaVgNLKlYDSWx9TuWJQWnLjekkxVDKgrKjEeIy+XgeU6yDodcb31OmMkwEVFVUsq7itKzfeb7j1\nTH2WIAoCRB8fiL5+MFQGup8/DH5+FZdVgt7Pz3Rb9FI73BYFhjcRkTOoOAm4qFIBGqCp++BarQa5\nDZ1hrfJ0eZWBrqsI/4rrxkt9xYpA5TJdxePKq6wQVKwolJcbVzxyciDLvg4h+zpk2dkVl9chXL8O\nxeVLZk0GBBjPBVDfwDf4+cMqZzupBcObiIgsSyYzBp1KVW2lwqIb+SvnB6gS6LKcbONlLYEvy0iH\n/NRJs3c/iJ5e1QIft8cAs960ysBChjcRETmfqvMDtGxl/vP0egi5lT367BvBXi3wsyG7XnE7JxuK\ns2eMJwA6+AeEyVMbNwOhmRjeREREleRyiP4B0Nc3gEtLoQ1UQ8w3b1N9Y3FiXyIiosaqPBbfShje\nREREDobhTURE5GAY3kRERA6G4U1ERORgGN5EREQOxqLhPX/+fAwbNgzDhw/HkSNHqt23Z88ePPbY\nYxg2bBiWLVtmyTKIiIicisXCe//+/UhLS8OGDRuQkJCAhISEave/9dZb+OCDD7Bu3Tr8/vvvOHPm\njKVKISIicioWC+/k5GTExcUBAMLDw5Gbm4uCggIAwMWLF+Hj44NmzZpBJpOhd+/eSE5OtlQpRERE\nTsViM6xlZWUhJibGdNvf3x+ZmZlQq9XIzMyEv79/tfsuXrxY5+v5+XlCoXD+U71ptRpbl+AQ2E7m\nYTuZh+1kHraTNGu1kdWmRxUbeZ7Z7OyiJqrEfmm1GmQ29Kw9LoTtZB62k3nYTuZhO0mzRBvVtjJg\nsc3mQUFByMrKMt3OyPDY9ZwAAAcVSURBVMiAVqut8b709HQEBQVZqhQiIiKnYrGed48ePfDBBx9g\n+PDhSElJQVBQENRqNQCgRYsWKCgowKVLlxASEoIdO3Zg8eLFdb6eq2yucZXfs7HYTuZhO5mH7WQe\ntpM0a7WRIDZ2e3YdFi9ejAMHDkAQBMyZMwepqanQaDSIj4/HH3/8YQrsBx54AOPGjbNUGURERE7F\nouFNRERETY8zrBERETkYhjcREZGDYXgTERE5GIY3ERGRg2F424lFixZh2LBh+Ne//oWff/7Z1uXY\ntZKSEsTFxSExMdHWpditpKQkPPLIIxgyZAh27txp63LsTmFhISZPnozRo0dj+PDh+O2332xdkt05\ndeoU4uLi8OWXXwIA/vnnH4wePRojRozAiy++iLKyMhtXaHs1tdGTTz6JUaNG4cknn0RmZqbF3pvh\nbQf27t2L06dPY8OGDVi5ciXmz59v65Ls2ocffggfHx9bl2G3srOzsWzZMnz11Vf46KOP8Msvv9i6\nJLvz7bffIiwsDGvWrMF77713y4mTXF1RURHmzZuHbt26mZa9//77GDFiBL766iu0bt0aX3/9tQ0r\ntL2a2mjJkiUYOnQovvzyS8THx2P16tUWe3+Gtx2466678N577wEAvL29UVxcDL1eb+Oq7NPZs2dx\n5swZ9OnTx9al2K3k5GR069YNarUaQUFBmDdvnq1Lsjt+fn7IyckBAOTl5cHPz8/GFdkXlUqFTz75\npNrMl/v27UPfvn0BAPfdd5/Ln0yqpjaaM2cO+vXrB6D6Z8wSGN52QC6Xw9PTEwDw9ddf495774Vc\n7vwnYWmIhQsXYsaMGbYuw65dunQJJSUlmDhxIkaMGOHyX7I1GThwIK5cuYL4+HiMGjUKr776qq1L\nsisKhQLu7u7VlhUXF0OlUgEAAgICLLpJ2BHU1Eaenp6Qy+XQ6/X46quv8PDDD1vu/S32ylRv27dv\nx9dff41Vq1bZuhS7tGnTJnTs2BEtW7a0dSl2LycnB0uXLsWVK1cwZswY7NixA4Ig2Losu/Hdd98h\nNDQUn376KU6cOIFZs2ZxDEU9cG6v2un1ekyfPh333HNPtU3qTY3hbSd+++03fPTRR1i5ciU0Gs4f\nXJOdO3fi4sWL2LlzJ65evQqVSoWQkBB0797d1qXZlYCAAHTq1AkKhQKtWrWCl5cXrl+/joCAAFuX\nZjcOHjyInj17AgCioqKQkZEBvV7PLV518PT0RElJCdzd3XkyqTrMnDkTrVu3xuTJky36Ptxsbgfy\n8/OxaNEirFixAr6+vrYux24tWbIE33zzDTZu3IjHH38ckyZNYnDXoGfPnti7dy8MBgOys7NRVFTE\nfbo3ad26NQ4fPgwAuHz5Mry8vBjcErp3746tW7cCAH7++Wf06tXLxhXZn6SkJCiVSkyZMsXi78W5\nze3Ahg0b8MEHHyAsLMy0bOHChQgNDbVhVfbtgw8+QPPmzTFkyBBbl2KX1q9fbxoN/Nxzz5kGGpFR\nYWEhZs2ahWvXrkGn0+HFF1+06CZOR3Ps2DEsXLgQly9fhkKhQHBwMBYvXowZM2agtLQUoaGhePvt\nt6FUKm1dqs3U1EbXrl2Dm5ub6Qya4eHhmDt3rkXen+FNRETkYLjZnIiIyMEwvImIiBwMw5uIiMjB\nMLyJiIgcDMObiIjIwXCSFiIXcenSJfTv3x+dOnWqtrx3794YP358o19/3759WLJkCdatW9fo1yKi\nujG8iVyIv78/1qxZY+syiKiRGN5EhOjoaEyaNAn79u1DYWEhFixYgMjISBw+fBgLFiyAQqGAIAh4\n44030LZtW5w/fx6zZ8+GwWCAm5sb3n77bQCAwWDAnDlzcPz4cahUKqxYsQJeXl42/u2InA/3eRMR\n9Ho9IiIisGbNGjzxxBN4//33AQDTp0/HzJkzsWbNGjz1/9u7Y1TFwSgMw6+a1k7BwsomnYhFNmJt\ncbG2ESxT2Ig27sBSBRegC1DQxkbQBdgrmAXIFAGZmTtMN/eSyfuUKUJOdXLOD//38cFoNALS6MNe\nr8disaDT6bDdboE0srXf77NerwmCgP1+/201Sf8zJ28pRx6PB91u95dnw+EQ4B3U0W63mc/nJEnC\n/X6n2WwCEEURg8EAgPP5TBRFQBqvCemZd6PRoFKpAFCr1UiS5N8XJeWQzVvKkb+def98U3KhUPgU\nIfr7Tcqv1+vTOwz3kL6Ga3NJAByPRwBOpxNhGFIul6lWq+/0rcPhQKvVAtLpfLfbAbDZbJjNZt/z\n0VJOOXlLOfKntXm9Xgfger2yWq14Pp9Mp1MgTbebTCaUSiWKxeI7ISmOY+I4ZrlcEgQB4/GY2+32\npbVIeWaqmCTCMORyuRAE/s9LWeDaXJKkjHHyliQpY5y8JUnKGJu3JEkZY/OWJCljbN6SJGWMzVuS\npIyxeUuSlDE/AHrXjrt8QoV/AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fcd4d8d7208>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 33.3 s, sys: 5.6 s, total: 38.9 s\n",
      "Wall time: 33.1 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "def get_cnn():\n",
    "    model = Sequential()\n",
    "    \n",
    "    model.add(Conv2D(filters=32, kernel_size=(5, 5), strides=(2, 2), padding='valid', \\\n",
    "                     activation='relu', kernel_initializer='he_normal', input_shape=input_shape))\n",
    "    \n",
    "    model.add(Conv2D(64, (5, 5), activation='relu', kernel_initializer='he_normal'))\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "    model.add(Dropout(0.25))\n",
    "    \n",
    "    model.add(Flatten())\n",
    "    model.add(Dense(256, activation='relu', kernel_initializer='he_normal'))\n",
    "    model.add(Dropout(0.5))\n",
    "    model.add(Dense(num_classes, activation='softmax'))\n",
    "    \n",
    "    model.compile(loss='categorical_crossentropy', optimizer='Adam', metrics=['accuracy'])\n",
    "    \n",
    "    return model\n",
    "\n",
    "model = get_cnn()\n",
    "history = model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs, verbose=1, validation_split=0.7)\n",
    "\n",
    "score = model.evaluate(x_test, y_test, verbose=0)\n",
    "print('Test loss:', score[0])\n",
    "print('Test accuracy:', score[1])\n",
    "\n",
    "x = list(range(1, epochs+1))\n",
    "vy = history.history['val_loss']\n",
    "ty = history.history['loss']\n",
    "plot_loss(x, vy, ty)\n",
    "\n",
    "pt.add_row(['2-Conv+Dropout+Kernel(5x5)+Stride(2x2)+Dense(256)', sum(history.history['acc'])/len(history.history['acc']), \\\n",
    "            sum(history.history['val_acc'])/len(history.history['val_acc']), score[1]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1117
    },
    "colab_type": "code",
    "id": "9j5_shSM0zk6",
    "outputId": "1751fced-8fa9-40bf-b4be-9cd4ac621521"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 18000 samples, validate on 42000 samples\n",
      "Epoch 1/20\n",
      "18000/18000 [==============================] - 5s 281us/step - loss: 0.7778 - acc: 0.7494 - val_loss: 0.1374 - val_acc: 0.9581\n",
      "Epoch 2/20\n",
      "18000/18000 [==============================] - 4s 221us/step - loss: 0.1939 - acc: 0.9412 - val_loss: 0.1104 - val_acc: 0.9658\n",
      "Epoch 3/20\n",
      "18000/18000 [==============================] - 4s 223us/step - loss: 0.1343 - acc: 0.9597 - val_loss: 0.0648 - val_acc: 0.9791\n",
      "Epoch 4/20\n",
      "18000/18000 [==============================] - 4s 222us/step - loss: 0.0983 - acc: 0.9711 - val_loss: 0.0670 - val_acc: 0.9806\n",
      "Epoch 5/20\n",
      "18000/18000 [==============================] - 4s 221us/step - loss: 0.0792 - acc: 0.9767 - val_loss: 0.0663 - val_acc: 0.9807\n",
      "Epoch 6/20\n",
      "18000/18000 [==============================] - 4s 220us/step - loss: 0.0641 - acc: 0.9803 - val_loss: 0.0553 - val_acc: 0.9840\n",
      "Epoch 7/20\n",
      "18000/18000 [==============================] - 4s 219us/step - loss: 0.0546 - acc: 0.9824 - val_loss: 0.0544 - val_acc: 0.9848\n",
      "Epoch 8/20\n",
      "18000/18000 [==============================] - 4s 220us/step - loss: 0.0462 - acc: 0.9862 - val_loss: 0.0513 - val_acc: 0.9850\n",
      "Epoch 9/20\n",
      "18000/18000 [==============================] - 4s 220us/step - loss: 0.0400 - acc: 0.9867 - val_loss: 0.0483 - val_acc: 0.9875\n",
      "Epoch 10/20\n",
      "18000/18000 [==============================] - 4s 220us/step - loss: 0.0364 - acc: 0.9888 - val_loss: 0.0586 - val_acc: 0.9849\n",
      "Epoch 11/20\n",
      "18000/18000 [==============================] - 4s 221us/step - loss: 0.0327 - acc: 0.9893 - val_loss: 0.0520 - val_acc: 0.9867\n",
      "Epoch 12/20\n",
      "18000/18000 [==============================] - 4s 222us/step - loss: 0.0308 - acc: 0.9900 - val_loss: 0.0531 - val_acc: 0.9860\n",
      "Epoch 13/20\n",
      "18000/18000 [==============================] - 4s 222us/step - loss: 0.0282 - acc: 0.9915 - val_loss: 0.0442 - val_acc: 0.9883\n",
      "Epoch 14/20\n",
      "18000/18000 [==============================] - 4s 223us/step - loss: 0.0246 - acc: 0.9919 - val_loss: 0.0519 - val_acc: 0.9873\n",
      "Epoch 15/20\n",
      "18000/18000 [==============================] - 4s 221us/step - loss: 0.0239 - acc: 0.9926 - val_loss: 0.0600 - val_acc: 0.9862\n",
      "Epoch 16/20\n",
      "18000/18000 [==============================] - 4s 222us/step - loss: 0.0278 - acc: 0.9913 - val_loss: 0.0617 - val_acc: 0.9857\n",
      "Epoch 17/20\n",
      "18000/18000 [==============================] - 4s 221us/step - loss: 0.0216 - acc: 0.9932 - val_loss: 0.0520 - val_acc: 0.9874\n",
      "Epoch 18/20\n",
      "18000/18000 [==============================] - 4s 220us/step - loss: 0.0196 - acc: 0.9934 - val_loss: 0.0468 - val_acc: 0.9887\n",
      "Epoch 19/20\n",
      "18000/18000 [==============================] - 4s 220us/step - loss: 0.0168 - acc: 0.9943 - val_loss: 0.0549 - val_acc: 0.9866\n",
      "Epoch 20/20\n",
      "18000/18000 [==============================] - 4s 221us/step - loss: 0.0185 - acc: 0.9937 - val_loss: 0.0463 - val_acc: 0.9885\n",
      "Test loss: 0.037904893789874904\n",
      "Test accuracy: 0.9884\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAe8AAAFYCAYAAAB6RnQAAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4yLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvNQv5yAAAIABJREFUeJzs3Xd8FHX+x/HX7G56D9kUmkBIQIOo\nWO4AC2rwUOw/T2LDOwv2csopoh6cGOyenr2hP8sP8e5QsWI59U5FkaJSpIgQQEgT0ntmfn9sskkk\nYRNga97PxyOP7Mzuzn6+2STv/c585zuGZVkWIiIiEjRs/i5AREREekbhLSIiEmQU3iIiIkFG4S0i\nIhJkFN4iIiJBRuEtIiISZBz+LqC7Skoq/V3CPpeUFM3OnTX+LmOfUpuCRyi2KxTbBKHZLrWpe5zO\nuE7Xq+ftRw6H3d8l7HNqU/AIxXaFYpsgNNulNu3la3lz47Nnz+a7777DMAymT5/OyJEj3fe98sor\nLFiwAJvNxogRI7j11lu9WYqIiEjI8FrPe/HixRQUFDBv3jzy8/PJz89331dVVcVzzz3HK6+8wty5\nc9mwYQPffvutt0oREREJKV4L70WLFpGbmwtAZmYm5eXlVFVVARAWFkZYWBg1NTU0NTVRW1tLQkKC\nt0oREREJKV7bbV5aWkpOTo57OTk5mZKSEmJjY4mIiOCqq64iNzeXiIgIJk6cyODBg3e7vaSk6JA8\nRtLVYIRgpjYFj1BsVyi2CUKzXWrTnvPZaPP21z+pqqriqaee4v333yc2NpYLL7yQNWvWMHz48C6f\nH2qjEsH1JofaKHq1KXiEYrtCsU0Qmu1Sm7q/zc54bbd5amoqpaWl7uXi4mKcTicAGzZsYMCAASQn\nJxMeHs5hhx3GypUrvVWKiIhISPFaeI8dO5aFCxcCsGrVKlJTU4mNjQWgX79+bNiwgbq6OgBWrlzJ\noEGDvFWKiIhISPFaeI8aNYqcnBzy8vK48847mTFjBvPnz+fDDz8kJSWFiy++mMmTJ3POOeew//77\nc9hhh3mrFBERCQCXXfZH1qz5ocO6J598lLlzX+708cuWLeG2224CYNq0G3a5/1//msdzzz3V5ev9\n+ON6Nm8uAGDGjFuor6/b09LJz5/JF1/8d4+fv6959Zj31KlTOyy3P6adl5dHXl6eN19eREQCyPjx\nv+Pf//6Q4cP3d6/79NN/88gjT3p87t13P9jj1/vss38zfPgBDBy4H3/96109fn4gC5rpUUVEJLgd\nf/wJXHHFxVx55bUArFnzA06nE6czlW+++Zpnn32SsLAw4uLiuOOOuzs8d+LE43nnnY9ZsmQxf//7\nAyQn96FPnxT69u1HU1MT+fkzKSkppra2losumkJ6egZvvjmfzz77N0lJSfzlL7fw4ovzqKqq5K67\n7qCxsRGbzca0abdjGAb5+TPp27cfP/64nuzsYUybdnu32vT44w+zYsV3NDU184c/TGbs2ON57723\nmT//NRyOMIYOzebGG2/udN3e6J3hXV1NxNtvUn/qGRAV5e9qRER8bubMCN56a99GwCmnNDFzZn2X\n9yclJdO3bz9Wr17JMceM5t///pDx4ycAUFlZyYwZd9K3bz9mzfoLX3+9iOjo6F228dRTj3L77bPI\nyspm6tRr6du3H5WVFRxxxG858cST+fnnrdx++zTmzHmZ3/xmNOPGHc8BB4xwP//ZZ5/k5JNP4/jj\nT+CTTz5izpynufjiy1i79gf++tfZJCUlc8YZJ1FZWUlc3O5P+/r222X89NMGnnhiTsuHhnM55JDf\n8uqrL3PvvQ+RlpbOO+8soL6+rtN1ERGRe/iT7qVzm4d//AHx11xOxNtv+rsUEZFeZfz4CXz88YcA\nfPHFfxg37ngAEhMTueeeO7n66iksX76UioryTp+/fft2srKyATj44FEAxMXF88MPq7jiiovIz5/Z\n5XMB1q79gUMOORSAUaMOY/36tQD06zeAPn1SsNlspKQ4qa6u8tiWNWtWu2uIiopi6NChbNmyhdzc\n3zF9+p957bX/Y/TosURERHa6bm/0yp63Fev6NGXfstnPlYiI+MfMmfW77SV7yzHHHMuLL85hxYoV\nDBgwkPj4eADuumsW9933EIMGDebBB+/p8vk2W1ufs3X+kA8/fJ+Kigoee+xZKioquOSSC3ZTgeF+\nXmNjE4bh2p7d3nESsPZzk3S5JcOg/cNcu+INLrjgj4wffyKffvoR1157BY899nSn6xISEj2+Rld6\nZc/bTM8AwFa43c+ViIj0LtHRMWRmZvHUU0+5d5kDVFdXkZaWTmVlJcuWLaWxsbHT56ekONm8eROW\nZbF8+VIAysrKyMjoi81m47PP/u1+rmEYNDc3d3j+/vsfwLJlSwD49tulHQbP9dTw4TnuGmpqati8\neTP9+w/kqaceIyUlhby88xkx4kAKCws7Xbc3emXP20xPB8C2lz88ERHpufHjJ5CfP4Np02a41515\n5u+54oqLGTBgIOedN5k5c55mypQrd3nulClXctttN5OenkFqahoA48Ydx7RpN7B69UomTjyV1NRU\nnn/+GQ466BAeeui+DsfOL7nkcu66axZvvfUGDkcYt9xyO01NTd2q+6mnHmXu3JcAGDRoCFOnTmPY\nsOFcddWlNDU1ceONNxIVFUV0dAyXXfZHYmNj6du3H1lZ2Sxe/NUu6/aGYXVn30AA2KdTzlkWKQOc\nNOWMoGzhp/tuuz2k6QGDQyi2CUKzXaHYJgjNdqlN3d9mZ3rlbnMMAzM9Qz1vEREJSr0zvHEd97YV\nF8GvjoeIiIgEul4b3s3pGRjNzdhKS/xdioiISI/02vBuG7SmEeciIhJcem94p7WeLqbj3iIiElx6\nb3hn6FxvEREJTr3yPG9oN1HL9m1+rkREpHd45JG/sXbtD+zY8QsNDfWkp/clPj6B2bPv8/jcd999\ni5iYWI455liPj7366inccMNNDBkydF+UHZAU3kXabS4i4gvXXPMnwBXEhYVbuOiiXSdh6cpJJ53i\nrbKCUi8Obw1YExEJBMuWLeHVV1+mpqaGq6/+E8uXL+XTTz/GNE1Gjx7LRRdN4bnnniIxMZHBgzOZ\nP/81DMNGQcFGxo07nosumuLxNZqamrj33ny2bfuZhoYGLrnkco444re8/PILfPbZJ9hsNsaOPYrJ\nky/qdF2g6bXhbcXGYcbEYteANRHphWJm3kbEW2/s023Wn3I61TPv3KPnbtjwI3Pnzic8PJzly5fy\n+OPPYrPZOPvs05g06dwOj129ehX/93//wjRNfv/7U7oV3h9++D7h4eE8+ujTlJaWcPXVl/Hqq/N5\n9dWXeeON97Hb7bzxxr8AOl0XaHpteINr0JqtUMe8RUT8bejQLMLDwwGIjIzk6qunYLfbKSsro6Ki\nosNjhw0bTmRkzy6p2f5SoCkpTsLDw6ioKGfcuOO5/vorGT9+Aiec4LpQSmfrAk3vDu/0DBw/rof6\neoiI8Hc5IiI+Uz3zzj3uJXtDWFgYAIWF25k37xXmzHmF6OhoLrjg7F0e++vLd3aP0eEyn42NjRiG\njalTb6GgYBP//veHXHPNZTz99P92us7hCKy47LWnigGYaS3HvYuL/FyJiIiA6/KeSUlJREdHs3bt\nGgoLC7u8PGhPtL8UaFFRITabDcMweP75Z9hvv0H88Y+XEheXQGlpyS7ramqq9/r197XA+ijhY+2v\n620OGOjnakREJCsrm6ioaK644iIOPPBgTjvtTB544B5GjjyoR9uZPfsO9671Qw89nAsu+CPLly/l\nmmsuo6mpkT//eTqxsbGUle3k0ksnExUVzYgRI0lPz9hlXXx8gjeauld65yVBW0Q9/Tixt02j/LmX\naDjltH2+fU90SbzgEIptgtBsVyi2CUKzXWpT97fZmV6927y5pedt16A1EREJIr06vDW/uYiIBKPe\nHd6aqEVERIJQ7w7v1tHm6nmLiEgQ6dXhTWQkZnKyJmoREZGg4tVTxWbPns13332HYRhMnz6dkSNH\nAlBUVMTUqVPdj9uyZQs33ngjp5zi+4nnzbQMbFu3+Px1RURE9pTXwnvx4sUUFBQwb948NmzYwPTp\n05k3bx4AaWlpvPTSS4BrsvgLLriA4447zlul7JaZno7jh1VQVQWxsX6pQUREpCe8ttt80aJF5Obm\nApCZmUl5eTlVVVW7PO7111/nd7/7HTExMd4qZbfcp4sV67i3iIgEB6/1vEtLS8nJyXEvJycnU1JS\nQuyverf/+Mc/mDNnjsftJSVF43DsyXy2HmQOctVXVwFdnAzvTV2dgB/M1KbgEYrtCsU2QWi2S23a\ncz6bHrWzidyWL1/OkCFDdgn0zuzcWeONsoiMSyYOqPjhR+oPGOWV1+iKZhgKDqHYJgjNdoVimyA0\n26U2dX+bnfHabvPU1FRKS0vdy8XFxTidzg6P+fTTTxk9erS3SuiWtvnNtdtcRESCg9fCe+zYsSxc\nuBCAVatWkZqauksPe8WKFQwfPtxbJXSLJmoREZFg47Xd5qNGjSInJ4e8vDwMw2DGjBnMnz+fuLg4\nxo8fD0BJSQl9+vTxVgnd4u55Fym8RUQkOHj1mHf7c7mBXXrZb731ljdfvltMZyqWzabd5iIiEjR6\n9wxrAA4HpjMV+3bNsiYiIsFB4Y1r17mtqBCC49LmIiLSyym8cQ1aM+rqMMrL/F2KiIiIRwpvwEzv\nC+h0MRERCQ4Kb9qdLqbj3iIiEgQU3rQ/XUw9bxERCXwKb9p63nZN1CIiIkFA4Q00p7VOkarwFhGR\nwKfwBswMDVgTEZHgofAGrORkrLAwbIUasCYiIoFP4Q1gGK6JWtTzFhGRIKDwbmGmpbtGm5umv0sR\nERHZLYV3CzOjL0ZzM0a7a5CLiIgEIoV3i+bW08V0aVAREQlwCu8WZuvpYpplTUREApzCu4V7ilQN\nWhMRkQCn8G7hniJVE7WIiEiAU3i3cE/UovnNRUQkwCm8W+jKYiIiEiwU3i2s2Dis6Bgd8xYRkYCn\n8G5lGDSnp+vKYiIiEvAU3u2YGX2xlZZAY6O/SxEREemSwrsdM63luHdxkZ8rERER6ZrCux336WIa\ntCYiIgFM4d2OJmoREZFgoPBux93z1vzmIiISwBze3Pjs2bP57rvvMAyD6dOnM3LkSPd927dv54Yb\nbqCxsZEDDjiAO+64w5uldEtzumuiFrt63iIiEsC81vNevHgxBQUFzJs3j/z8fPLz8zvcf/fdd3PR\nRRfxz3/+E7vdzrZt/j/OrIlaREQkGHgM78bGRgpbeqJr1qzhjTfeoLa21uOGFy1aRG5uLgCZmZmU\nl5dTVVUFgGmaLF26lOOOOw6AGTNm0Ldv3z1uxL7iHm2unreIiAQwj+E9bdo0vv32W4qKirjmmmtY\nt24d06ZN87jh0tJSkpKS3MvJycmUlJQAsGPHDmJiYrjrrrs455xzeOCBB/aiCftQVBRmYqKOeYuI\nSEDzeMy7qKiICRMm8Pzzz3Puuefyxz/+kT/84Q89fiHLsjrcLioqYvLkyfTr148pU6bw6aefMm7c\nuC6fn5QUjcNh7/Hr9lj//ti2bsXpjPP+a4HPXseX1KbgEYrtCsU2QWi2S23acx7Du6GhAcuy+PDD\nD93HrWtqajxuODU1ldLSUvdycXExTqcTgKSkJPr27cvAgQMBGD16NOvXr99teO/c6fk194WElFTC\nV66kpKAIoqO9+lpOZxwlJZVefQ1fU5uCRyi2KxTbBKHZLrWp+9vsjMfd5kcccQSHHnooTqeTwYMH\n88ILLzB48GCPLzh27FgWLlwIwKpVq0hNTSU2NhYAh8PBgAED2LRpk/v+7mzTF3RdbxERCXQee95T\np05lypQpxMfHA5Cbm8t5553nccOjRo0iJyeHvLw8DMNgxowZzJ8/n7i4OMaPH8/06dOZNm0almWR\nnZ3tHrzmb80t4W0vKsQckunnakRERHblMbw/++wzysrKOO2007jxxhtZsWIFU6dO5YQTTvC48alT\np3ZYHj58uPv2fvvtx9y5c/egZO9qG3GunreIiAQmj7vNH3/8cY466ig+++wzTNPk9ddf56WXXvJF\nbX5hZrhOWdPpYiIiEqg8hndkZCTJycl89tlnnHbaacTExGCzhe6sqm3zm6vnLSIigcljCtfX1/Ps\ns8/yn//8h9GjR7Np0yYqK0NrhGB7bQPWNMuaiIgEJo/hPWvWLIqKirj77ruJiIjg888/3+VYdigx\nnalYhqHd5iIiErA8DljLysriwgsvZPXq1Xz44Yccd9xxATGVqdeEhWGlOLXbXEREApbHnvfcuXOZ\nPHky77zzDm+99RYXXHABr7/+ui9q85vmjL7Yiwqh3axwIiIigcJjz/vNN9/kvffeIyIiAnDNrvbH\nP/6RM844w+vF+YuZno7x/bcYFeVYCYn+LkdERKQDjz1vh8PhDm6A6OhowsLCvFqUv5lprYPWdNxb\nREQCj8eed3p6OrNmzWLMmDEAfP7552RkZHi9MH9qf7pY87DhHh4tIiLiWx7De9asWbz00kvMnz8f\nwzA46KCDuOCCC3xRm9+0TdSiQWsiIhJ4PIZ3VFQUU6ZM6bDu66+/5je/+Y3XivI3d8+7SLvNRUQk\n8OzRVGmPPfbYvq4joDS3HPO2b9dELSIiEnj2KLytED+Fqm2WNfW8RUQk8OxReBuGsa/rCChWnz5Y\nYWE65i0iIgGpy2PeDz/8cKfrLcti69atXisoINhsmGnpOuYtIiIBqcvwttvtXT7pzDPP9EoxgcRM\nS8fx/bdgmhDCV1ETEZHg02V4X3311b6sI+CY6RkYS7/B+OUXLKfT3+WIiIi4qUvZBV3XW0REApXC\nuwvNLRO12IsU3iIiEli6dVWxqqoqX9QSUMy01p63Bq2JiEhg8Rjea9eu5dRTT+Xmm29myZIlvqgp\nILjP9dZELSIiEmA8To86c+ZMTNPk66+/ZsGCBdx///0cf/zxnH322SQkJPiiRr/QRC0iIhKounXM\n22azMXDgQNLT02loaGDVqlWcd955fPTRR96uz2/a5jfXMW8REQksHnveb7zxBv/6178oKyvj97//\nPc8//zwJCQlUVFRw/vnnk5ub64s6fc6KT8CKjlbPW0REAo7H8P7888+57rrrOOywwzqsj4+P58IL\nL/RaYX5nGDSnpetUMRERCTjdOub9xhtv8Pbbb2MYBsOGDeP0008nMjKS//mf//FFjX5jpmcQ9tWX\n0NgIYWH+LkdERAToxjHvG2+8ke+//57hw4eTnZ3NkiVLuOGGG3xRm9+Z6ekYloWtpNjfpYiIiLh5\n7HmXl5fz1FNPuZfPOecczj333G5tfPbs2Xz33XcYhsH06dMZOXKk+77jjjuO9PR09xzq999/P2lp\naT2t36vMdNdELbbC7Zh9+/m5GhEREReP4d2/f39KSkpwtszvXVpayn777edxw4sXL6agoIB58+ax\nYcMGpk+fzrx58zo85plnniEmJmYPS/c+nS4mIiKByGN4b9u2jfHjxzN06FBM02Tjxo1kZmZy3nnn\nAfDKK690+rxFixa5R6JnZmZSXl5OVVUVsbGx+7B873KfLqaJWkREJIB4DO/rr79+jzZcWlpKTk6O\nezk5OZmSkpIO4T1jxgx+/vlnDj30UG688UYMw9ij1/IWd89b1/UWEZEA4jG8jzjiCJYsWcKKFSsw\nDIODDjqIQw45pMcvZFlWh+Vrr72Wo446ioSEBK666ioWLlzIhAkTunx+UlI0DkfX1xj3igOGAhBT\nVkqMM84rL+H00nb9SW0KHqHYrlBsE4Rmu9SmPecxvB9++GG++OILDj30UADuvPNOTjjhBC677LLd\nPi81NZXS0lL3cnFxsfu4OcDpp5/uvn300Uezbt263Yb3zp01nkrd98LicAINmzZTXlK5zzfvdMZR\n4oXt+pPaFDxCsV2h2CYIzXapTd3fZmc8nir29ddf8+qrr3LzzTdz8803M2/ePD755BOPLzh27FgW\nLlwIwKpVq0hNTXXvMq+srOTiiy+moaEBgG+++YasrKxuN8ZnoqMxExK121xERAKKx563aZrYbG0Z\n73A4unVsetSoUeTk5JCXl4dhGMyYMYP58+cTFxfH+PHjOfroo5k0aRIREREccMABu+11+5OZnq4B\nayIiElA8hveIESO4/PLLGTNmDABffvklBx54YLc2PnXq1A7Lw4cPd9++8MILg2J6VTMtA8faNVBb\nC1FR/i5HRETEc3hPnz6d9957zz3ZyqmnnsqJJ57oi9oCgpnRNuLcHDTYz9WIiIh0I7yfffZZpkyZ\nwsSJE31RT8BpP1GLwltERAKBxwFr69ato6CgwBe1BKTmlola7Lqut4iIBAiPPe+1a9cyceJEEhIS\nCAsLw7IsDMPg008/9UF5/memtfS8NWhNREQChMfwfvLJJ3dZV1tb65ViApF7ilTNby4iIgHC427z\nv/zlL/Tr16/D18033+yL2gKCmdF2ZTEREZFA0GXPe8GCBTz22GNs27aNcePGudc3NjaSkpLii9oC\ngpnqukypJmoREZFA0WV4n3rqqUycOJFbb72Va665xr3eZrORmprqk+ICQlgYZopTx7xFRCRg7PaY\nt91u5+6772bNmjWUlZW5Ly6yadMmRo8e7ZMCA0FzegaOnzaAZUGAXflMRER6H48D1q699lp++OEH\n0lsGbgEYhtGrwtvMyMBY+T1GVSVWXLy/yxERkV7OY3hv3bqVDz/80Be1BKz2E7U0K7xFRMTPPI42\nHzx4sPvqX72VmdZ6uphGnIuIiP957HnbbDYmTpzIyJEjsdvt7vX33nuvVwsLJO6etwatiYhIAPAY\n3mPGjHFfUay30kQtIiISSDyG9xlnnMG6devYvHkzubm5VFRUEB/fu477uidq0fzmIiISADyG9wsv\nvMDbb79NQ0MDubm5PP7448THx3PllVf6or6A0Nwyv7ldPW8REQkAHgesvf3227z22mskJCQAcNNN\nN/Wai5K0slJSsOx2DVgTEZGA4DG8Y2JisNnaHmaz2Tos9wo2G2ZausJbREQCgsfd5gMHDuTRRx+l\noqKCDz74gHfffZfMzExf1BZQzIwMHN9/p1nWRETE77p1VbGoqCjS0tJYsGABBx10EDNmzPBFbQHF\nTMvAaGzE2LHD36WIiEgv57HnHRYWxqRJk7j44ospKSmhoKAAh8Pj00JO2+li22nu08fP1YiISG/m\nsec9a9Ys3nvvPcrKyjj33HN5+eWXmTlzpg9KCyytE7XYCzVRi4iI+JfH8F69ejW///3vee+99zj9\n9NN56KGHKCgo8EVtAaW53fzmIiIi/uQxvFsvA/rpp59y3HHHAfTKuc7bLk6iEeciIuJf3bowyUkn\nnUR1dTX7778/b7zxhvuc797EVM9bREQChMeRZ3feeSfr1q1znx6WlZXl7oH3Ju4Ba5oiVURE/Mxj\nz/uHH36gsLCQ8PBw/va3v3Hvvfeybt06X9QWUKyERKzISGzbFd4iIuJfHsP7zjvvZPDgwSxZsoQV\nK1Zw++238/e//71bG589ezaTJk0iLy+P77//vtPHPPDAA1xwwQU9q9ofDEOzrImISEDwGN4REREM\nGjSIjz/+mLPPPpuhQ4d2a3rUxYsXU1BQwLx588jPzyc/P3+Xx/z444988803e1a5HzRn9MVWUgxN\nTf4uRUREejGPKVxbW8t7773HRx99xJFHHklZWRkVFRUeN7xo0SJyc3MByMzMpLy8nKqqqg6Pufvu\nu/nTn/60h6X7npmejmGa2EpL/F2KiIj0Yh4HrN1www28+OKL3HDDDcTGxvLII4/whz/8weOGS0tL\nycnJcS8nJydTUlJCbGwsAPPnz+eII46gX79+3So0KSkah8Percd6zeD9AOhTXwHO7H2ySaczbp9s\nJ5CoTcEjFNsVim2C0GyX2rTnPIb3b3/7W0aOHMnGjRtZvXo1l1xyCVFRUT1+odbzxQHKysqYP38+\nzz//PEVFRd16/s6dNT1+zX0tKr4PsUD56h9p2G/YXm/P6YyjpKRy7wsLIGpT8AjFdoVimyA026U2\ndX+bnfEY3h999BEzZ84kPT0d0zQpLS1l1qxZHHPMMbt9XmpqKqWlpe7l4uJinE4nAF999RU7duzg\nvPPOo6Ghgc2bNzN79mymT5/ekzb5nJmhiVpERMT/PIb3s88+y4IFC0hOTgagqKiI6667zmN4jx07\nlkceeYS8vDxWrVpFamqqe5f5hAkTmDBhAgBbt27llltuCfjghnYTtehcbxER8aNuXVWsNbgB0tLS\nCAsL87jhUaNGkZOTQ15eHoZhMGPGDObPn09cXBzjx4/fu6r9pO3KYpplTURE/MdjeMfExDBnzhzG\njBkDwOeff05MTEy3Nj516tQOy8OHD9/lMf379+ell17q1vb8rTnVFd727bqymIiI+I/H8M7Pz+fh\nhx9mwYIFGIbBwQcfzOzZs31RW+CJjcWMi1fPW0RE/MpjeK9cuZI77rjDF7UEBTMjQ8e8RUTErzxO\n0vLCCy/QpBnF3My0DGw7dkB9vb9LERGRXspjzzsuLo6JEydywAEHdBiodu+993q1sEDVdnWxQsyB\n+/m5GhER6Y08hvexxx7Lscce64tagoL7dLHt2xXeIiLiF7sN7y1btnDGGWe4l2traykqKmLQoEHe\nritgNWfoXG8REfGvLo95L1q0iHPOOYfKyrap3rZs2cIll1zCypUrfVJcIDLTXOFt1yxrIiLiJ12G\n96OPPsqcOXOIi2ubVzU7O5snnniChx56yCfFBSJN1CIiIv7WZXhblkV29q5XzsrKyqK+F4+0bjvm\nrYlaRETEP7oM75qarq/iVVZW5pVigoGZ1jbaXERExB+6DO+srCzmzp27y/pnnnmGgw46yKtFBbTw\ncMyUFF1ZTERE/KbL0eY33XQTV111FW+++SYjRozANE2WLVtGbGwsTz31lC9rDDhmWga2gk3+LkNE\nRHqpLsPb6XTy2muvsWjRItavX4/dbufEE0/k8MMP92V9Aak5PR3HqhUYVZVYsZ1fKF1ERMRbPE7S\nMnr0aEaPHu2LWoKGe9BaYSHNQxXeIiLiWx7nNpddtYW3jnuLiIjvKbz3gMJbRET8qcvd5osWLdrt\nE3vzrvT2u81FRER8rcvwfvzcWMgzAAAgAElEQVTxx7t8kmEYvTy8W8/1Vs9bRER8r8vwfumll7p8\n0sKFC71STLBof2UxERERX/M42nzbtm28/PLL7Ny5E4CGhga+/vprfve733m9uEBlpjix7HZdnERE\nRPzC44C1m266icTERL799ltGjBjBzp07uffee31RW+Cy2zFT03TMW0RE/MJjeNvtdqZMmUJKSgrn\nnXceTzzxBK+88oovagtoZnq665i3Zfm7FBER6WU8hnd9fT2FhYUYhsGWLVtwOBz8/PPPvqgtoJlp\nGRj19Rg7d/i7FBER6WU8HvO+5JJL+PLLL7n44os57bTTsNvtnHzyyb6oLaCZGe1mWUvu4+dqRESk\nN/EY3rm5uViWhWEYLF68mIqKCpKTk31RW0BrP1FL8wE5fq5GRER6E4+7zd9//32uuOIKABwOB5df\nfjnvv/++1wsLdM2t4a3reouIiI95DO8XXniB++67z708Z84cnn/++W5tfPbs2UyaNIm8vDy+//77\nDve99tprnH322eTl5TFz5kysIBv4Zaa5JmrR6WIiIuJrHsPbsizi4tqunBUbG4thGB43vHjxYgoK\nCpg3bx75+fnk5+e776utreWdd97hlVde4dVXX+Wnn35i+fLle9gE/2ibqGWbnysREZHexuMx7xEj\nRnD99ddzxBFHYFkW//3vfxkxYoTHDS9atIjc3FwAMjMzKS8vp6qqitjYWKKiovjf//1fwBXkVVVV\nOJ3OvWyKb7UfsCYiIuJLHnvet912G8ceeywbNmxg48aNnHLKKUyfPt3jhktLS0lKSnIvJycnU1JS\n0uExTz/9NOPHj2fChAkMGDBgD8r3HysxCSsiQvObi4iIz3XZ8y4uLiY1NZWtW7cyatQoRo0a5b7v\n559/7nHYdnZMe8qUKUyePJlLL72UQw89lEMPPbTL5yclReNw2Hv0ml7Xty9hxUU4nXGeH9uFvXlu\noFKbgkcotisU2wSh2S61ac91Gd733HMPDzzwABdeeGGHY9ytp419/PHHu91wamoqpaWl7uXi4mL3\nrvGysjLWr1/P4YcfTmRkJEcffTTLli3bbXjv3FnT7Ub5SqIzDcfSbygtLAN7zz9YOJ1xlJRUeqEy\n/1GbgkcotisU2wSh2S61qfvb7EyX4f3AAw8AMHfuXNLS0nr8gmPHjuWRRx4hLy+PVatWkZqaSmxs\nLABNTU1MmzaNBQsWEBMTw4oVKzj11FN7/Br+1pzRl7DmZmylJe7R5yIiIt7mccDan//8Z1588cUe\nb3jUqFHk5OSQl5eHYRjMmDGD+fPnExcXx/jx47nqqquYPHkyDoeDYcOGcfzxx+9RA/zJfV3vwu0K\nbxER8RmP4T1o0CBuuukmDjnkEMLCwtzrzzrrLI8bnzp1aofl4cOHu2+feeaZnHnmmT2pNeCYae1G\nnB/k52JERKTX8BjejY2N2O32XSZZ6U54h7r2PW8RERFf8Rjed911F+AaZGYYBgkJCV4vKlhoohYR\nEfEHj+G9bNkybrrpJqqrq7Esi8TERO677z4OPPBAX9QX0NwTtWh+cxER8SGP4f3AAw/w+OOPk52d\nDcDq1avJz8/nlVde8Xpxga51kJp2m4uIiC95nGHNZrO5gxvggAMOwL4H5zSHIis2DjM2DrumSBUR\nER/qVnh/8MEHVFVVUVVVxbvvvqvwbsdMT9cUqSIi4lMew/uvf/0r8+bN49hjj+W4447jjTfe4K9/\n/asvagsKZkZfbKWlUF/v71JERKSX6NZ53s8995wvaglK7uPexUWYAwb6uRoREekNPIb3ueeeu8v1\nu+12O4MHD+bKK6/co6lTQ4n7dLHC7QpvERHxCY/hPWbMGDZu3Mjvfvc7bDYbH330ERkZGSQkJHDL\nLbcwZ84cX9QZsNomatGgNRER8Q2P4b106VKef/5593Jubi5Tpkzh6aef9nhlsd6guaXnbS/URC0i\nIuIbHges/fLLL+zYscO9XFlZybZt26ioqKCyMrQu57YnzPS+gHreIiLiOx573pMnT+bEE0+kX79+\nGIbB1q1bueyyy/jkk0+YNGmSL2oMaJrfXEREfM1jeJ911llMmDCBTZs2YZomAwcOJDEx0Re1BYW2\nWdbU8xYREd/wuNu8vLycxx57jBdeeIGRI0eybNmyDrvRe72ICMzkZE3UIiIiPuMxvG+77TYyMjLY\nunUrAA0NDdx8881eLyyYmOl9sW1XeIuIiG94DO8dO3YwefJkwsLCAJgwYQJ1dXVeLyyYmOnp2Cor\noKrK36WIiEgv4DG8ARobG90TtZSWllJTU+PVooKN+3SxYh33FhER7/M4YO28887jrLPOoqSkhMsv\nv5wVK1Zw6623+qK2oNF+opbmIUP9XI2IiIQ6j+F90kknMWrUKJYvX054eDh33HEH8fHxvqgtaJhp\nbVOkioiIeJvH3eYXX3wx6enpnHjiiRx//PGkpqZy3nnn+aK2oGFmtEzUokFrIiLiA132vBcsWMBj\njz3Gtm3bGDdunHt9Y2MjKSkpvqgtaGiiFhER8aUuw/vUU09l4sSJ3HrrrVxzzTXu9TabrddfSezX\n3FcW07neIiLiA7vdbW6327n77rtJTEzEMAwMw6C+vp6zzz7bV/UFBTPFiWWzaZY1ERHxCY8D1p59\n9lmefPJJGhoaiI6Opr6+nlNOOcUXtQUPhwPTmYp9u64sJiIi3udxwNr777/Pl19+yUEHHcRXX33F\n/fffT1ZWli9qCypmRga2okKwLH+XIiIiIc5jeMfExBAeHk5jYyMAxx9/vK7j3QkzPQOjrg6jvMzf\npYiISIjzuNs8ISGBBQsWkJ2dzS233EJmZibFxcXd2vjs2bP57rvvMAyD6dOnM3LkSPd9X331FQ8+\n+CA2m43BgweTn5+PzdatCd8CUtu53oU0Jyb5uRoREQllHtPynnvuYdSoUdxyyy3st99+FBYW8uCD\nD3rc8OLFiykoKGDevHnk5+eTn5/f4f6//OUv/P3vf+fVV1+lurqa//73v3veigCg08VERMRXdtvz\n3rJlCwMGDKB///4AXHjhhRQVFTFo0CCPG160aBG5ubkAZGZmUl5eTlVVFbGxsQDMnz/ffTs5OZmd\nO3fuTTv8zj1Ri8JbRES8rMue96JFizjnnHOorKx0r9uyZQuXXHIJK1eu9Ljh0tJSkpLadh8nJydT\nUlLiXm4N7uLiYr744guOOeaYPWpAoGhu6XnbFd4iIuJlXfa8H330UebMmUNcXJx7XXZ2Nk888QT3\n3HMPzz77bI9eyOpkFPYvv/zC5ZdfzowZMzoEfWeSkqJxOOw9ek2f2t91QZKY8l+IccZ5eHAbZw8e\nGyzUpuARiu0KxTZBaLZLbdpzXYa3ZVlkZ2fvsj4rK4v6+nqPG05NTaW0tNS9XFxcjNPpdC9XVVVx\n6aWXcv3113PkkUd63N7OnYF9GVIjPJ4UoPGbpZQVloHd8wcNpzOOkpJKj48LJmpT8AjFdoVimyA0\n26U2dX+bnelyt/nurtldVub5dKixY8eycOFCAFatWkVqaqp7VznA3XffzYUXXsjRRx/tcVvBwEpO\npmH0WMIWf0XcdVdCc7O/SxIRkRDVZc87KyuLuXPncs4553RY/8wzz3DQQQd53PCoUaPIyckhLy8P\nwzCYMWMG8+fPJy4ujiOPPJI33niDgoIC/vnPfwJw8sknM2nSpL1sjh8ZBhUvvUrC2acT+dpcsCwq\n//5Et3rgIiIiPWFYnR2MBkpKSrjqqquw2WyMGDEC0zRZtmwZsbGxPPXUU8TExPi00GDZvWJUlJMw\n6QzCli6h7vd5uw1w7TYKDqHYJgjNdoVimyA026U2dX+bnemy5+10OnnttddYtGgR69evx263c+KJ\nJ3L44Yfv08JCjRWfQPm810mYdCaR/3gVQD1wERHZpzzOsDZ69GhGjx7ti1pChivA57cFuGVR+ciT\nCnAREdkngnc+0gBnxSdQ/trrNB56OJH/nEfc1ZdpEJuIiOwTCm8vsuLi2wL8X68pwEVEZJ9QeHuZ\nO8APO8IV4FdNUYCLiMheUXj7gBUXT/m8+TQe/hsi5//DFeBNTf4uS0REgpTC20esuHjKX/1XW4Bf\nfZkCXERE9ojC24d+3QNn8mQFuIiI9JjC28es2Dh3gDN3LnFXXaoAFxGRHlF4+0FrgDN2LJGv/0sB\nLiIiPaLw9hMrNg7ee4/G34xWgIuISI8ovP0pLo7yuf9sC/ArL1GAi4iIRwpvP7Ni2wX4G/MV4CIi\n4pHCOwBYsXGUzf0XDb8d4wrwKxTgIiLSNYV3oIiNpfz//ukK8DcV4CIi0jWFdyBpDfDRY10BfvnF\n0Njo76pERCTA9Mrw/vlng8svj+Qf/3AEXjbGxlL+yj9cAb7gdVcPPOCKFBERf+qV4V1SYrBggYOr\nrorisMNieOSRcMrL/V1VO78K8PhL/4Cxc4e/qxIRkQDRK8P74INNFi2qZsqUBsrLDWbNiuDgg2O5\n/fYItmwx/F2eS+su9LFHEfHuWySPOZTI/3sJTNPflYmIiJ/1yvAG2G8/izvvrOfbb6u47bZ64uIs\nnnoqnCOOiOGyyyL59tsA+NHExFA+73Wq/jILo7aOuOuvInHieBzff+vvykRExI8CIKH8KzERrr22\ngSVLqnnkkVqys01efz2ME06I4fTTo1i40O7fzm54OLVXX8eOL5dQd9qZhC39hsQTxhE77UaMsp1+\nLExERPyl14d3q/BwmDSpiU8/reG112oYN66JL790cMEF0Rx5ZDQvvRRGXZ3/6jP79qPymRco+8eb\nNA/JJGrOMySPOZSIV1/RrnQRkV5G4f0rhgHjxjXz2mu1fPJJNZMmNVJQYOPGGyMZNSqG++4Lp7TU\nf8fFG485lp2fLqLqtr9i1NQQf+0VJJ7yO+wrvvdbTSIi4lsK793IyTF55JE6li6t5tpr62lsNLjv\nvghGjYph6tQINmzwU4iHh1N77Z/Y8cUS6k85nbBvviZp/NHE3jIVo7zMPzWJiIjPKLy7IT3d4rbb\nGli+vIr8/DpSUy1efDGcMWNimDw5kq++smNZvq/L7NefiudepGze6zQPHkLUc0+TPFq70kVEQp3C\nuwdiY+HSSxv56qtqnn22lkMOMXn//TBOPTWaE0+M5s03HX6Z0bTx2ONbdqXPxKipdu1KP3UC9pUr\nfF+MiIh4ncJ7DzgccOqpTbz3Xg0LFtQwYUIjy5fbuPTSKE46KZrNm/2wOz0igtprb2DH599Qf/Jp\nhC3+iqTco4i59SaMikCagUZERPaWV8N79uzZTJo0iby8PL7/vuOAqvr6em6++WbOPPNMb5bgVYYB\nv/1tMy++WMeXX1Zz5pmNfPutndzcGD74wO6Xmsz+A6iY8xJlr86nedBgop950rUr/bW5+GXfvoiI\n7HNeC+/FixdTUFDAvHnzyM/PJz8/v8P99957L/vvv7+3Xt7nMjMtnnyyjoceqqWuDs4/P5o77wz3\n24XBGo/LZednX1E9/S8YVZXEX32Za1f6qpX+KUhERPYZr4X3okWLyM3NBSAzM5Py8nKqqqrc9//p\nT39y3x9Kzj23iXfeqWHQIJO//z2Cs86KoqjIT6PSIyKouX6qa1f6SacQ9vUi167026dpV7qISBDz\nWniXlpaSlJTkXk5OTqakpMS9HBsb662X9rsDDzT56KNqJk5s5MsvHRx3XDSff+6f3egA5oCBVLzw\nCuVz/4k5YCDRTz1O8uhDiZ59B47vlmt3uohIkHH46oWsvQyIpKRoHA7/BWBPOZ3w1lvw8MPw5z/b\nOOusaGbNgmnTwGZr/7g43xWV9z9w+kS4/35sd91FzEP3E/PQ/TBwIJxxBpx5JowdC/a9+zn7tE0+\nEoptgtBsVyi2CUKzXWrTnvNaeKemplJaWupeLi4uxul07vH2du6s2Rdl+dx550FWlo0pU6K49VYb\nn3zSxKOP1pKc7HqTS0oqfV/UZdfB+ZcQ/snHRLz7FuEfvI/t4Yfh4YcxU1KonzCRhomn0HDkMRAR\n0aNN+61NXhSKbYLQbFcotglCs11qU/e32Rmv7TYfO3YsCxcuBGDVqlWkpqaG9K7y3TniCJOPP3bN\nl/7RRw5yc2NYtszPZ+nFxNBw8qlUPv4Mv6zeQNmr86mdfBEYNqJe/l8SzjmLPvsPIe7yiwhf8Dq0\nG68gIiL+ZVh7uz97N+6//36WLFmCYRjMmDGD1atXExcXx/jx47n22mspLCxk/fr1jBgxgrPPPptT\nTjmly22Fwie05mb429/Cue++cBwOePBBg7PPrsQIkEuIA9DcjOObxUS8+xYR776FfXMBAFZEBA3H\nHk/9SafQcMIErOQ+nT5dn6aDRyi2KxTbBKHZLrWp+9vsjFfDe18KpTf500/tXHllJKWlNk47rZG/\n/a2OgNwpYVnYV64g4p0FRLz7Fo41P7hW2+00jjmK+pNOpuGkkzEz+rqfoj/I4BGK7QrFNkFotktt\n6v42O6Pw9pPt2w2uvDKWL76AoUObee65OvbfP7DnI7dvWE/4O28T8e4CwpYtda9vPPRw6k86hfqJ\np9DnNweH3HsViv9kIDTbFYptgtBsl9rU/W12RuHtR4mJcVx3XQNPPBFOVJTFPffUkZfnp1ldesi2\n7WfC33ubiHfeImzRFxjNza47srOp/e2RNB55FA1jjsJKTfVvoftAKP6TgdBsVyi2CUKzXWpT97fZ\nGYW3H7W+0e++6+DaayOpqDA4//wG8vPriYryd3XdZ/zyC+EfvOc6Tv7FfzsMbmsaNpzGMUfScOTR\nNI4+EislxY+V7plQ/CcDodmuUGwThGa71Kbub7MzCm8/av9Gb9xocMklUaxYYWfEiGaefbaWIUOC\n4q3pwJkYyc6P/0vYF/8l/PP/ELb4K4yattP8mvY/gIaxR9E49mgax4zFSkr2Y7XdE4r/ZCA02xWK\nbYLQbJfa1P1tdkbh7Ue/fqPr6uDWWyN46aVw4uIsHn64jpNPDo7d6K12+eVtaMDx7XLCv/gPYZ//\nl7BvvsKoqwPAMgyaDxhBw5EtYT56DFZCop8q71oo/pOB0GxXKLYJQrNdalP3t9kZhbcfdfVGv/aa\ng5tuiqSmxuDyyxu4/fZ6wsI6PsayoKHBFfi1tQY1NVBXZ1Bb2/a9trbj99b1NTUG9fWuq6LZ7a4v\nmw3sdst927Xc/n7rV8vt77fcy9nZUfTrV0lyVx3q+nrCli8l7PP/EPbFfwlbshijvt7VJsOg6cCD\naBx7FI1HHkXjb0ZjxSfs4596z4XiPxkIzXaFYpsgNNulNnV/m51RePvR7t7oNWtsXHRRJD/+aGfA\nAJOoKGuXMDbNQDpBvKOMDJOcHJOcnOaW7yZDhpi7zrxaV0fY0m9cYf7l54Qt/QajoQEAy2aj6aCD\nafzNGJqH709TVjbN2cN83jsPxX8yEJrtCsU2QWi2S23q/jY747O5zaVnhg83+eCDGqZNi+S99xzU\n1kJUFCQnW0RGWkRFWURGQlSURXQ0Letwr4uK6vj91+sjIly99+bmtu+tX6Zp/Gq59bbRxeNd9zU1\nwY4dkXzzTROrVtn46CMHH33U9isWFWUxfHjHQD/ggEjixx5F49ijXA+qqSFsyWLCvvgP4V98jmPZ\nEsKWL+vwszGdqTRlD6M5K5vmrGyasobRnD3Mdb55QM14I7J3TBNKSw2qqmDbNhv19a49bg0NBnV1\nv77t2qPmeozhfmx9fdvtujrD/Zw+fSyyspoZNswkK8tk4ECrw3UXJLCp5+1Hof7J85dfDFavtrFq\nlY2VK+2sWmVj3TobjY0dA3bgwI6BnpPT3PaPpLoax4rvcaxfi339Ouzr1+JYvw7bls0Yv/rVNWNi\nac7Konmoq4felNUS8IOHsMtxhz1sUzCrqoJ162ysWWPjxx9tpKREkJBQR0aGSd++Fn37msTH+7vK\nvROM71V9PWzebGPTJoNNm2wUFNjYtMm1vHmzjbo633wgjYqyGDrUFeTDhplkZ7u+Bg0y9+bPp0vB\n+F55ot3mnQi1Nxl65y9vQwOsX+8K9FWrXIG+erWN0tKOH/ljYy0OOMAV6EOHmoSHg8PhOr7ucEBE\ncw3JpetJLlpLUtFaEravJWHbWmK3/4i9sb7Dtky7g7r+Q6gblE3DkGE0ZGbTmDkMx/5DiMpI9NhZ\nD7b3qb7e9TNes8bG2rU21qyx88MPNjZv9tytio11hXhGhkXfvpY72Pv1a11nkpAQuDs4AvW92rmT\nlkBuDWfDvbx9u4Fl7foDTUy02G8/kwEDTFJSwjDNBiIiICLCIjwc9+2ICFqWO7tttTxu19tFRQZr\n19pZt87m/vrxRxu1tR1rCQuzyMx0hXp2tunuqWdmmkRG7vnPxB/vlWW5OhUFBQYOBwwZYhK3Dy8C\npvDuRCD+Qe6tQP1Hszf2pE2WBcXFhruH3tpbX7/e1uPj+jaaGcxGhrOG/fmB/fnBfTuJsl0eX0IK\nG21D2RyeybaoIRTFDeWXpCHs7JOJmdSH2DhITQ3HZqsnNtZq+aLL2+HhPSp3rzQ1wcaNrpBu//XT\nTzaamzv+3FJSTPbf32T4cNdXVpZJVFQ0P/xQy7ZtNrZtM9i+vfW7wY4dXQd9dLTlDvLWHnvrcnw8\nNDa6vpqaoLHRdailbZ3R7r625faPb73P9TwDw3DtncnMdH0NGWISE9N5bf76m2pqgm3bDHevuaDA\n6BDW5eW7/h4bhusD0qBBZsuXK6xblxPbDe3wVbtME7ZsMVi3zvXBb/16u/t2VVXHNthsFoMGWWRn\nN5Od3RboTqeF0+k6nLc73mpTUxP8/LPRYQ9G+/eisnLXv43Bgy2GDDE7fA0ebPZ42mqFdydCLeRA\n4e1JbS2sXevqMTY1uf4om5td//DbbreuN371GNfj3LcbLWKqi0nfuZaMsjWkl68lrWID6dUb6Fu/\nkTCrcZfXLyOBHxna6VcRacCu/5DDw9uCPCbGIiam66CPiXHd3t1jHI62f6iucLa7Q3r9ehsNDR1r\niI+3GD68meHD28J62DCTlJRd/8x3917V1EBhodFpsLeu+/XeEl/KyGgL8/Zfo0bFUla27/+mWj9g\nbt7s2pXt+nLdLiiw8fPPxi4fmMA1FsUVyK6Qbh/OAwZY3b7irr//V1iW6/dh7Vpbh576unW2Lj/o\nRUe7QjwlxcLpbAt117LF0KFRhIVVk5Li+qDSk+Pt1dVQULDrXoxNm2xs2WLQ1LTrexEV1fYe7Lef\nRXMz/PST68Puli2dv39OZ2uYWx1CffDgzj9AKrw7EWohB/7/g/SGoGxTUxO2n7di3/gT9o0/YWz4\nCTa4bkds3YitoX6Xp9SHxVCSkElhbCZbIzMpCBvKT8ZQfjIHsak+g501kVRVuQYadfZPobsiIy0M\ng112ZUZFWQwb1tqTbgvr9HSr27u09/a9qqtz/UNvDfaff7ZRXe36wBEWBg6H1fLdtRwW5jrdsG1d\nx/s7e7zD4fowtmmTjQ0bOn5t3brrf3uHAwYNaiYz0/XPNjPTddglM9MkNXX3P5vyctxh/OuQ3rJl\n193JrdLSXIO9Bg5sH86u0E5L2zeDwAL576q0tK2nXlBgo6TEoLTUcH8vLe08TNtzOCz69GkL9vbf\nk5IsCguNDr3o4uLOf6gpKa5gbv8hadAgi8GDd//+NzS4PiC3hvlPP9nYuNH1fetWo9M9gGlp7Xvp\nrt+3Y4+NIiZG4d1BoP7i7o1A/oPcUyHXJtPE2VBB2ZLv3eFu37SxLehrqjt/Wp8+mGkZNGdk0OjM\noDapL9UJGVTE9aUsOoNfIvvyiy2V6lpbS8i7gr797epqg+pq1x6FoUPbgnrYMNc/qL0NhWB/r2pr\nXYcN2gf65s1hrFljsXPnrv9sY2Isd5gPGmRSW9uxJ93Zrm1wHXseONBs+WoL6YEDLfr3N30ylXEw\nv1em6fpgVFJi6xDq1dWRFBQ0tCzb2q3vOuhtNov+/a0OH5LaQnrfHr9u1Tqg8KefOob7pk2uYG8/\nXsHhgO+/r+p0T9eeUngHoGD+g+xKr2qTZWEUF7cE+k/Yf9qAfesWbEWF2LZvw1ZYiK2q65+F5XBg\npqVjpmdgpruCvvW2mZ6BmdEXMz0dK847Q8BD+b3asYNdeuobNrh6U78evR0VZXUI5l/fTvD/PEEh\n/V79Wk0N7h57SYnBzp0GqamukO7f3/LpuBJP6upcu+9dgW4QHx/J+edX7tMBnTrPW2RfMwystDSa\n0tJo+u3ozh9SVekK8e3bsBVux7Z9O7ai7di3b3ctF27H8f23GEu/6fJlrOhozIRErPh4rPgEzPh4\nrIQErLgErISW5fgE1/0JCZgt6634eMz4BIiODtzh4V6SnAzJySaHH97xMrum6RrM9NNPNqKjLfbb\nz7V7tpf9eAJadDQtH54Cv18ZGQnDhrnGlQA4nZGUlPjmtRXeIl5kxcbRPDSO5qFZXT/INDF++QVb\n4Xbshdvawr61B19cjK28DFtJMcaP69suv9rdGhwOV7DHxbs+BCQkuHrzaSnEhEe7lhMSMOMTXB8C\nWm+3rLdi43o2miiA2WwwYIDFgAE9+xmKBBqFt4i/2WxYTifNTifNB47c/WMtC2pqsFWUY1RUYJSX\nY6ssxyhvWa4ox9ay3mhZb2tZb5SX4/hxfYfj9B7O5nG9pGG4Q931AaBdyHe4nYgVEwORkVgRri8i\nI1puR7Ssj2hZH8muc+WKSHcpvEWCiWFATAxmTAxk9N2zbTQ2YlRUkBLWzM6NP7d9CGgJeKO8zPUh\noLzcFfoVFe7bts0FOCor9klTLIejy4CnJfytyJbb0dEtXzFt32NidllHfyf2Bjquc+jfnIQe/VaL\n9DZhYVh9+oAzjqb41J4/v7kZo7LC3dt3h35FOUZVJUZ9A0Z9HdTXYdTVu27X1WHU1bmuIFfv+m7U\n1UG9636jrg6jthajbKfrMbW1u0x/212/vqCdFR7eEvS/Cvu4OPdYAbPdGILWsQKuMQSJmHGusQR7\nNZ2YL1kWVFdjKy9zfQc+6scAAA6MSURBVCgrL8MoK8MoL2u7XVEOkVEtgyRdAyPNjL6YqWl7NZWw\n+I7CW0R6xm7HSkzCSkwCwCtHjy0LmppcwV9bh1FTjVFT4/pe3e52u+/U1BBjNlD7S1m7+zo+x7Zz\nB8bPW12P72lJ4eEdgz0+sV3wtwwijI4Buw3Lbgdb2zV0rV9dR7fj/Z083uZaj93u2ttiayJ883Zs\nZa7wtVWUtwTyTte6luXWwDaamvbsx24YWClOmtMzMDMyMNNavrcsN6f3xUzPwEpO7nWDIAONwltE\nAo9huPYQhIVBbBwWzm49LcYZR1V3TqkyTYzqqo57DyrKf7Vc0e6wQVmHvQ32bT+79hz40O7OWLMi\nIjATEjGT+2ANzsRMTHTtVUhMbLmd6Lqd4PpuxcdDTW3LmQ/bXGdBFLY7A2LDeowV33X9euHhbac4\ntg/6lBSs6GjXoY6WsQ9ERWJFRrkOh0RFta23ejj3qHSg8BaR3sdmc/WWW86h36O9B/X1rmBvP2Cw\npqblOrnNrrMC2l03t+Ny6/2m58dbFtHpKVSGRWG1nC1gJiS5QrhloOA+nynGsjAqK9pCffs27K3h\n3nK6o237dhxLvyGsh2c/tJcS2RLykVEQEYHVEu7u8G+5z4ppGeMQEwvR0e7b7de7D420rtvX4x2a\nmtoO87iutdrue8vtoQMhpf++e83dUHiLiOyJiAj3WQLeFu2Mo86Xk7S0nGHQHJ9A87DhXT+uuRlb\naUlbqP9S2ja+oa7WFWy1tS3jHeqgruV2XR3hzY00VVW7xztQX+/a49G6vA9YkZFt4d4a+tGuoMcw\noCV0jfp61+36BteYDPe6BowG10XSDdP0/IKGgbFqA1ZKyj6pf3cU3iIismfsdtcsgWnpcNAhPXqq\n0xlHWVcfSCzLFZwtYU91u7EO1VWu2+7v1e3GNXSxvroaW1Gh63ZDw64vZ7O5PoxFRGCFu66basbH\nQ3gEVkR4y/cIrPDwljMhWtZ1uD+cmAOGuQaD+oDCW0REAothtIXpvp6etrERo7oKLMsd1Ptq93qM\nMw58tIdE4S0iIr1HWJj7TIlg5tU5D2fPns2kSZPIy8vj+++/73Dfl19+yVlnncWkSZN47LHHvFmG\niIhISPFaeC9evJiCggLmzZtHfn4++fn5He6/8847eeSRR5g7dy5ffPEFP/74o7dKERERCSleC+9F\nixaRm5sLQGZmJuXl5VRVVQGwZcsWEhISyMjIwGazccwxx7Bo0SJvlSIiIhJSvHbMu7S0lJycHPdy\ncnIyJSUlxMbGUlJSQnJycof7tmzZstvtJSVF43CE3oUMurpWazBTm4JHKLYrFNsEodkutWnP+WzA\nmrWH8xS32rmz59MZBrquLkYfzNSm4BGK7QrFNkFotktt6v42O+O13eapqamUlpa6l4uLi3G2TGbw\n6/uKiopITd2DCySIiIj0Ql4L77Fjx7Jw4UIAVq1aRWpqKrGxrrls+/fvT1VVFVu3bqWpqYlPPvmE\nsWPHeqsUERGRkOK13eajRo0iJyeHvLw8DMNgxowZzJ8/n7i4OMaPH8/MmTO58cYbATjppJMYPHiw\nt0oREREJKV495j116tQOy8OHt82Re/jhhzNv3jxvvryIiEhI8uokLSIiIrLvKbxFRESCjGHt7Tlc\nIiIi4lPqeYuIiAQZhbeIiEiQUXiLiIgEGYW3iIhIkFF4i4iIBBmFt4iISJDx2VXFerN7772XpUuX\n0tTUxGWXXcYJJ5zgvu+4444jPT0du911udP777+ftLQ0f5XaLV9//TXXXXcdWVlZAGRnZ3P77be7\n7//yyy958MEHsdvtHH300Vx11VX+KrVH/vGPf7BgwQL38sqVK1m+fLl7OScnh1GjRrmXX3jhBff7\nFojWrVvHlVdeyR/+8AfOP/98tm/fzk033URzczNOp5P77ruP8PDwDs+ZPXs23333HYZhMH36dEaO\nHOmn6jvXWZtuueUWmpqacDgc3Hfffe4LIIHn39VA8Os2TZs2jVWrVpGYmAjAxRdfzLhx4zo8J9Df\nJ9i1Xddeey07d+4EoKysjIMPPphZs2a5Hz9//nwefvhhBg4cCMCYMWP+v707jY3pewM4/p12KFP9\n0VGUiJ1SESpae2uNLbYQTFIVqYiiglKVqI4XmApijaVBovbUG0ulYnmBtEOQWhNbREtS1doqFTo9\n/xcTk047Rfm3d64+n3e9Z6Z5npxz77nnuXfuJS4uTpPYq1P5WN6rVy/t9iklalV2draaN2+eUkqp\n4uJiFRUV5dY+fPhwVVJSokFkfy4nJ0fFx8dX2z5u3Dj15s0b5XA4lMViUU+fPq3D6P4/7Ha7slqt\nbtsiIiI0iqbmvnz5oqKjo9WaNWtUenq6UkqppKQklZmZqZRSasuWLero0aNu37Hb7Wr+/PlKKaWe\nPXumZsyYUbdB/4KnnBITE9X58+eVUkodOXJEpaamun3nV2NVa55yWrVqlbpy5Uq13/H2flLKc14V\nJSUlqdzcXLdtp0+fVjabra5CrDFPx3It9ykpm9ey8PBwtm/fDsB///1HaWkpDodD46hqT15eHk2b\nNqV169b4+PgQFRVFdna21mHV2O7du1m4cKHWYfyxhg0bkpaW5vaqXbvdzsiRIwEYPnx4lX7Jzs5m\n1KhRAHTu3JmPHz9SUlJSd0H/gqecUlJSGDNmDACBgYF8+PBBq/D+iKecfsXb+wl+nteLFy/4/Pmz\nV1YLfsbTsVzLfUom71rm6+uLyWQCICMjg8jIyCql1pSUFCwWC5s3b0bp5IF3z549Y8GCBVgsFm7c\nuOHaXlhYiNlsdv1tNpspLCzUIsQ/du/ePVq3bu1WfgX49u0bCQkJzJo1i0OHDmkU3e8xGo00atTI\nbVtpaamrpNe8efMq/fLu3TsCAwNdf3tb33nKyWQy4evri8Ph4NixY0ycOLHK96obq97AU04AR44c\nISYmhmXLllFcXOzW5u39BNXnBXD48GGio6M9tt28eZPY2FjmzJnDo0ePajPEGvN0LNdyn5Jr3nXk\n0qVLZGRkcPDgQbftS5YsYejQoTRt2pRFixaRlZXF2LFjNYry93To0IHFixczbtw48vLyiImJ4eLF\ni1Wu9ehVRkYGU6dOrbI9MTGRSZMmYTAYiI6Opl+/fvTq1UuDCP/e75wk6uVE0uFwkJiYyIABAxg4\ncKBbmx7H6uTJk2nWrBk9evRg//797Nq1i7Vr11b7eb30EzhPgG/fvo3Vaq3S1rt3b8xmM8OGDePu\n3busWrWKs2fP1n2Qv1DxWF7x/qW63qdk5V0Hrl27xt69e0lLSyMgIMCtbcqUKTRv3hyj0UhkZCRP\nnjzRKMrf16pVK8aPH4/BYKBdu3YEBQVRUFAAQMuWLXn37p3rswUFBTUqCXoDu91OWFhYle0WiwV/\nf39MJhMDBgzQRV9VZDKZ+Pr1K+C5Xyr33du3b6tUH7zR6tWrad++PYsXL67S9rOx6q0GDhxIjx49\nAOcNrZXHmV77CeDWrVvVlss7d+7sujEvLCyM4uJir7vEWPlYruU+JZN3Lfv8+TObNm1i3759rrtH\nK7bFxsby7ds3wDmwf9wV683OnDnDgQMHAGeZvKioyHWHfNu2bSkpKSE/P5+ysjKuXr3K4MGDtQy3\nRgoKCvD396+yMnvx4gUJCQkopSgrK+POnTu66KuKBg0aRFZWFgAXL15k6NChbu2DBw92tT98+JCW\nLVvSpEmTOo+zJs6cOUODBg1YsmRJte3VjVVvFR8fT15eHuA8kaw8zvTYTz/cv3+f7t27e2xLS0vj\n3LlzgPNOdbPZ7FW/5vB0LNdyn5KyeS3LzMzk/fv3LF261LWtf//+hISEMHr0aCIjI5k5cyZ+fn6E\nhoZ6fckcnKuBFStWcPnyZb5//47VauXcuXMEBAQwevRorFYrCQkJAIwfP56OHTtqHPHvq3zNfv/+\n/YSHhxMWFkZwcDDTp0/Hx8eHESNGePUNNw8ePCA1NZXXr19jNBrJyspi8+bNJCUlcfLkSdq0acOU\nKVMAWLZsGRs3bqRv37707NmTWbNmYTAYSElJ0TgLd55yKioqws/Pj9mzZwPO1ZvVanXl5GmselPJ\n3FNO0dHRLF26lMaNG2Mymdi4cSOgn34Cz3nt3LmTwsJC10/BfoiLi2PPnj1MnDiRlStXcuLECcrK\nyli/fr1G0Xvm6Vhus9lYs2aNJvuUvBJUCCGE0BkpmwshhBA6I5O3EEIIoTMyeQshhBA6I5O3EEII\noTMyeQshhBA6Iz8VE6KeyM/PZ+zYsVUeQBMVFcW8efP++v/b7Xa2bdvG8ePH//p/CSF+TiZvIeoR\ns9lMenq61mEIIf6STN5CCEJDQ1m4cCF2u50vX75gs9no1q0bubm52Gw2jEYjBoOBtWvX0qVLF16+\nfElycjLl5eX4+fm5HiRSXl5OSkoKjx8/pmHDhuzbtw9/f3+NsxPi3yPXvIUQOBwOunbtSnp6OhaL\nhR07dgDOl7GsXr2a9PR05s6dy7p16wDnm/BiY2M5evQo06ZN48KFCwA8f/6c+Ph4Tp06hdFo5Pr1\n65rlJMS/TFbeQtQjxcXFrkeJ/rBy5UoAhgwZAkDfvn05cOAAnz59oqioyPUY2IiICJYvXw44X5sa\nEREBwIQJEwDnNe9OnToRFBQEQHBwMJ8+far9pISoh2TyFqIe+dk174pPSjYYDBgMhmrbwVkir8yb\nXiQhxL9MyuZCCABycnIAuH37NiEhIQQEBNCiRQtyc3MByM7Opk+fPoBzdX7t2jXA+cKGrVu3ahO0\nEPWUrLyFqEc8lc3btm0LwKNHjzh+/DgfP34kNTUVgNTUVGw2G76+vvj4+GC1WgFITk4mOTmZY8eO\nYTQa2bBhA69evarTXISoz+StYkIIQkJCePjwIUajnM8LoQdSNhdCCCF0RlbeQgghhM7IylsIIYTQ\nGZm8hRBCCJ2RyVsIIYTQGZm8hRBCCJ2RyVsIIYTQGZm8hRBCCJ35HywekcpnUGAQAAAAAElFTkSu\nQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fcd4cf1c710>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 1min 12s, sys: 14 s, total: 1min 26s\n",
      "Wall time: 1min 23s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "epochs = 20\n",
    "def get_cnn():\n",
    "    model = Sequential()\n",
    "    \n",
    "    model.add(Conv2D(filters=32, kernel_size=(5, 5), strides=(2, 2), padding='same', \\\n",
    "                     activation='relu', kernel_initializer='he_normal', input_shape=input_shape))\n",
    "    \n",
    "    model.add(Conv2D(64, (5, 5), activation='relu', padding='same', kernel_initializer='he_normal'))\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "    model.add(Dropout(0.25))\n",
    "    \n",
    "    model.add(Conv2D(64, (5, 5), activation='relu', padding='same', kernel_initializer='he_normal'))\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "    model.add(BatchNormalization())\n",
    "    model.add(Dropout(0.25))\n",
    "    \n",
    "    model.add(Flatten())\n",
    "    model.add(Dense(128, activation='relu', kernel_initializer='he_normal'))\n",
    "    model.add(Dropout(0.5))\n",
    "    model.add(Dense(num_classes, activation='softmax'))\n",
    "    \n",
    "    model.compile(loss='categorical_crossentropy', optimizer='Adam', metrics=['accuracy'])\n",
    "    \n",
    "    return model\n",
    "\n",
    "model = get_cnn()\n",
    "history = model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs, verbose=1, validation_split=0.7)\n",
    "\n",
    "score = model.evaluate(x_test, y_test, verbose=0)\n",
    "print('Test loss:', score[0])\n",
    "print('Test accuracy:', score[1])\n",
    "\n",
    "x = list(range(1, epochs+1))\n",
    "vy = history.history['val_loss']\n",
    "ty = history.history['loss']\n",
    "plot_loss(x, vy, ty)\n",
    "\n",
    "pt.add_row(['3-Conv+BN+Dropout+Kernel(5x5)+Stride(2x2)', sum(history.history['acc'])/len(history.history['acc']), \\\n",
    "            sum(history.history['val_acc'])/len(history.history['val_acc']), score[1]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 849
    },
    "colab_type": "code",
    "id": "fx0JcBmK0zoU",
    "outputId": "18c617ce-0acb-4af5-85fd-a25adeccbe44"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 18000 samples, validate on 42000 samples\n",
      "Epoch 1/12\n",
      "18000/18000 [==============================] - 9s 509us/step - loss: 0.7721 - acc: 0.7626 - val_loss: 0.1316 - val_acc: 0.9598\n",
      "Epoch 2/12\n",
      "18000/18000 [==============================] - 8s 435us/step - loss: 0.2239 - acc: 0.9314 - val_loss: 0.0894 - val_acc: 0.9714\n",
      "Epoch 3/12\n",
      "18000/18000 [==============================] - 8s 437us/step - loss: 0.1502 - acc: 0.9537 - val_loss: 0.1346 - val_acc: 0.9584\n",
      "Epoch 4/12\n",
      "18000/18000 [==============================] - 8s 436us/step - loss: 0.1117 - acc: 0.9654 - val_loss: 0.0647 - val_acc: 0.9800\n",
      "Epoch 5/12\n",
      "18000/18000 [==============================] - 8s 437us/step - loss: 0.0966 - acc: 0.9704 - val_loss: 0.0750 - val_acc: 0.9777\n",
      "Epoch 6/12\n",
      "18000/18000 [==============================] - 8s 436us/step - loss: 0.0881 - acc: 0.9738 - val_loss: 0.0483 - val_acc: 0.9852\n",
      "Epoch 7/12\n",
      "18000/18000 [==============================] - 8s 436us/step - loss: 0.0745 - acc: 0.9767 - val_loss: 0.0514 - val_acc: 0.9844\n",
      "Epoch 8/12\n",
      "18000/18000 [==============================] - 8s 435us/step - loss: 0.0684 - acc: 0.9788 - val_loss: 0.0495 - val_acc: 0.9850\n",
      "Epoch 9/12\n",
      "18000/18000 [==============================] - 8s 437us/step - loss: 0.0580 - acc: 0.9817 - val_loss: 0.0445 - val_acc: 0.9866\n",
      "Epoch 10/12\n",
      "18000/18000 [==============================] - 8s 445us/step - loss: 0.0537 - acc: 0.9833 - val_loss: 0.0441 - val_acc: 0.9866\n",
      "Epoch 11/12\n",
      "18000/18000 [==============================] - 8s 436us/step - loss: 0.0486 - acc: 0.9842 - val_loss: 0.0401 - val_acc: 0.9879\n",
      "Epoch 12/12\n",
      "18000/18000 [==============================] - 8s 436us/step - loss: 0.0473 - acc: 0.9847 - val_loss: 0.0594 - val_acc: 0.9835\n",
      "Test loss: 0.04915771906846421\n",
      "Test accuracy: 0.9862\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAe8AAAFZCAYAAACxGqelAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fcd4c13fcc0>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 1min 7s, sys: 15.1 s, total: 1min 22s\n",
      "Wall time: 1min 38s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "epochs = 12\n",
    "def get_cnn():\n",
    "    model = Sequential()\n",
    "    \n",
    "    model.add(Conv2D(filters=32, kernel_size=(3, 3), strides=(1, 1), padding='same', \\\n",
    "                     activation='relu', kernel_initializer='he_normal', input_shape=input_shape))\n",
    "    \n",
    "    model.add(Conv2D(64, (3, 3), activation='relu', padding='same', kernel_initializer='he_normal'))    \n",
    "    model.add(Conv2D(64, (3, 3), activation='relu', padding='same', kernel_initializer='he_normal'))\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "    model.add(BatchNormalization())\n",
    "    model.add(Dropout(0.25))\n",
    "    \n",
    "    model.add(Conv2D(16, (3, 3), activation='relu', padding='same', kernel_initializer='he_normal'))    \n",
    "    model.add(Conv2D(16, (3, 3), activation='relu', padding='valid', kernel_initializer='he_normal'))\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "    model.add(BatchNormalization())\n",
    "    model.add(Dropout(0.25))\n",
    "    \n",
    "    model.add(Flatten())\n",
    "    model.add(Dense(128, activation='relu', kernel_initializer='he_normal'))\n",
    "    model.add(Dropout(0.5))\n",
    "    model.add(Dense(num_classes, activation='softmax'))\n",
    "    \n",
    "    model.compile(loss='categorical_crossentropy', optimizer='Adam', metrics=['accuracy'])\n",
    "    \n",
    "    return model\n",
    "\n",
    "model = get_cnn()\n",
    "history = model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs, verbose=1, validation_split=0.7)\n",
    "\n",
    "score = model.evaluate(x_test, y_test, verbose=0)\n",
    "print('Test loss:', score[0])\n",
    "print('Test accuracy:', score[1])\n",
    "\n",
    "x = list(range(1, epochs+1))\n",
    "vy = history.history['val_loss']\n",
    "ty = history.history['loss']\n",
    "plot_loss(x, vy, ty)\n",
    "\n",
    "pt.add_row(['5-Conv+BN+Dropout+Kernel(3x3)', sum(history.history['acc'])/len(history.history['acc']), \\\n",
    "            sum(history.history['val_acc'])/len(history.history['val_acc']), score[1]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 848
    },
    "colab_type": "code",
    "id": "Y_W_kKR2xCZz",
    "outputId": "5e0d6938-83f2-43f9-d1eb-7aae275308db"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 18000 samples, validate on 42000 samples\n",
      "Epoch 1/12\n",
      "18000/18000 [==============================] - 19s 1ms/step - loss: 0.9229 - acc: 0.7188 - val_loss: 0.2046 - val_acc: 0.9390\n",
      "Epoch 2/12\n",
      "18000/18000 [==============================] - 17s 945us/step - loss: 0.2146 - acc: 0.9329 - val_loss: 0.0907 - val_acc: 0.9723\n",
      "Epoch 3/12\n",
      "18000/18000 [==============================] - 17s 946us/step - loss: 0.1349 - acc: 0.9586 - val_loss: 0.0920 - val_acc: 0.9718\n",
      "Epoch 4/12\n",
      "18000/18000 [==============================] - 17s 946us/step - loss: 0.1016 - acc: 0.9681 - val_loss: 0.0653 - val_acc: 0.9789\n",
      "Epoch 5/12\n",
      "18000/18000 [==============================] - 17s 957us/step - loss: 0.0863 - acc: 0.9733 - val_loss: 0.0952 - val_acc: 0.9723\n",
      "Epoch 6/12\n",
      "18000/18000 [==============================] - 17s 954us/step - loss: 0.0727 - acc: 0.9774 - val_loss: 0.0507 - val_acc: 0.9838\n",
      "Epoch 7/12\n",
      "18000/18000 [==============================] - 17s 947us/step - loss: 0.0627 - acc: 0.9814 - val_loss: 0.0525 - val_acc: 0.9834\n",
      "Epoch 8/12\n",
      "18000/18000 [==============================] - 17s 945us/step - loss: 0.0548 - acc: 0.9834 - val_loss: 0.0446 - val_acc: 0.9868\n",
      "Epoch 9/12\n",
      "18000/18000 [==============================] - 17s 947us/step - loss: 0.0521 - acc: 0.9837 - val_loss: 0.0418 - val_acc: 0.9880\n",
      "Epoch 10/12\n",
      "18000/18000 [==============================] - 17s 955us/step - loss: 0.0369 - acc: 0.9885 - val_loss: 0.0592 - val_acc: 0.9844\n",
      "Epoch 11/12\n",
      "18000/18000 [==============================] - 17s 957us/step - loss: 0.0360 - acc: 0.9881 - val_loss: 0.0382 - val_acc: 0.9890\n",
      "Epoch 12/12\n",
      "18000/18000 [==============================] - 17s 953us/step - loss: 0.0396 - acc: 0.9872 - val_loss: 0.0369 - val_acc: 0.9895\n",
      "Test loss: 0.027266340670636418\n",
      "Test accuracy: 0.9917\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAe8AAAFYCAYAAAB6RnQAAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nKWUMb0mSUsbwliQpZb4Blw7zlh7PWScAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fcd4afb6da0>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 1min 56s, sys: 33.4 s, total: 2min 29s\n",
      "Wall time: 3min 32s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "epochs = 12\n",
    "def get_cnn():\n",
    "    model = Sequential()\n",
    "    \n",
    "    model.add(Conv2D(filters=32, kernel_size=(3, 3), strides=(1, 1), padding='same', \\\n",
    "                     activation='relu', kernel_initializer='he_normal', input_shape=input_shape))\n",
    "    \n",
    "    model.add(Conv2D(128, (3, 3), activation='relu', padding='same', kernel_initializer='he_normal'))    \n",
    "    model.add(Conv2D(128, (3, 3), activation='relu', padding='same', kernel_initializer='he_normal'))\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "    model.add(BatchNormalization())\n",
    "    model.add(Dropout(0.25))\n",
    "    \n",
    "    model.add(Conv2D(64, (3, 3), activation='relu', padding='same', kernel_initializer='he_normal'))    \n",
    "    model.add(Conv2D(64, (3, 3), activation='relu', padding='same', kernel_initializer='he_normal'))\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "    model.add(BatchNormalization())\n",
    "    model.add(Dropout(0.25))\n",
    "        \n",
    "    model.add(Conv2D(32, (3, 3), activation='relu', padding='same', kernel_initializer='he_normal'))    \n",
    "    model.add(Conv2D(32, (3, 3), activation='relu', padding='valid', kernel_initializer='he_normal'))\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "    model.add(BatchNormalization())\n",
    "    model.add(Dropout(0.25))\n",
    "    \n",
    "    model.add(Flatten())\n",
    "    model.add(Dense(256, activation='relu', kernel_initializer='he_normal'))\n",
    "    model.add(Dropout(0.5))\n",
    "    model.add(Dense(num_classes, activation='softmax'))\n",
    "    \n",
    "    model.compile(loss='categorical_crossentropy', optimizer='Adam', metrics=['accuracy'])\n",
    "    \n",
    "    return model\n",
    "\n",
    "model = get_cnn()\n",
    "history = model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs, verbose=1, validation_split=0.7)\n",
    "\n",
    "score = model.evaluate(x_test, y_test, verbose=0)\n",
    "print('Test loss:', score[0])\n",
    "print('Test accuracy:', score[1])\n",
    "\n",
    "x = list(range(1, epochs+1))\n",
    "vy = history.history['val_loss']\n",
    "ty = history.history['loss']\n",
    "plot_loss(x, vy, ty)\n",
    "\n",
    "pt.add_row(['7-Conv+BN+Dropout+Kernel(3x3)', sum(history.history['acc'])/len(history.history['acc']), \\\n",
    "            sum(history.history['val_acc'])/len(history.history['val_acc']), score[1]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 218
    },
    "colab_type": "code",
    "id": "qAsnTIolxCdP",
    "outputId": "37eea4bc-c573-4543-e182-8e717d2673aa"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+---------------------------------------------------+---------------------+--------------------------+---------------+\n",
      "|                  CNN Architecture                 | Mean Train Accuracy | Mean Validation Accuracy | Test Accuracy |\n",
      "+---------------------------------------------------+---------------------+--------------------------+---------------+\n",
      "|             2-Conv+Dropout+Kernel(3x3)            |  0.9681203703659551 |    0.9814444444444445    |     0.9871    |\n",
      "|                 2-Conv+Kernel(3x3)                |  0.9884722222531283 |    0.9788234126984127    |     0.9799    |\n",
      "|             2-Conv+Dropout+Kernel(2x2)            |  0.9727175926146683 |    0.9801706349206348    |     0.9852    |\n",
      "| 2-Conv+Dropout+Kernel(5x5)+Stride(2x2)+Dense(128) |  0.9599398148148146 |    0.9796527777777776    |     0.9876    |\n",
      "| 2-Conv+Dropout+Kernel(5x5)+Stride(2x2)+Dense(256) |  0.9696481481613937 |    0.9817837301587301    |     0.9892    |\n",
      "|     3-Conv+BN+Dropout+Kernel(5x5)+Stride(2x2)     |  0.972183333378368  |    0.983097619047619     |     0.9884    |\n",
      "|           5-Conv+BN+Dropout+Kernel(3x3)           |  0.9538796296296296 |    0.9788769841269843    |     0.9862    |\n",
      "|           7-Conv+BN+Dropout+Kernel(3x3)           |  0.953444444431199  |    0.9782599206349206    |     0.9917    |\n",
      "+---------------------------------------------------+---------------------+--------------------------+---------------+\n"
     ]
    }
   ],
   "source": [
    "print(pt)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "M5yMt90WcaNb"
   },
   "source": [
    "# Conclusion :\n",
    "- Dropout Layer is increasing the performance of the NN. Also, from the Error plots it is observed that Dropout layer is reducing Overfitting in the NN.\n",
    "- There isn't much evidence to suggest any significant changes in performance with different Kernel Sizes.\n",
    "- Increasing the Neurons in the Dense Layer (ANN) is improving the performance of the NN.\n",
    "- A deeper NN (7-Conv CNN) with dropouts and a high number of Neurons in the Dense Layer pushed the Accuracy beyond 99%. "
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "CNN on MNIST dataset.ipynb",
   "provenance": [],
   "version": "0.3.2"
  },
  "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}