dense-vs-cnn.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "d61cc5be-f2fe-48f6-a3f0-2c595b54dd9d",
   "metadata": {},
   "source": [
    "# The importance of Convolutional Neural Networks\n",
    "\n",
    "## Deep Learning\n",
    "\n",
    "### This notebook presents some comparison between <b>Fully-connected neural networks</b> and <b>Convolutional Neural Networks</b> on two distinct Classification problems.\n",
    "### The datasets used in this example are <b>[MNIST](https://en.wikipedia.org/wiki/MNIST_database)</b> and <b>[SVHN](http://ufldl.stanford.edu/housenumbers)</b>."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4b25d4d2-7774-495a-b8af-24bac82b94cf",
   "metadata": {},
   "source": [
    "#### This notebook uses Python version 3.6 or newer."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8ba194a8-e331-4ee7-8e80-0b782e13d20c",
   "metadata": {},
   "source": [
    "Let's start by checking the Python version."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "aeaf2199-581e-48e6-aa9a-13076ed9eb20",
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "\n",
    "if int(sys.version[0]) < 3 or int(sys.version[2]) < 6:\n",
    "    assert(\"Python version 3.6 or newer is required\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4e981609-14eb-4923-80eb-8ea5dee814a3",
   "metadata": {},
   "source": [
    "We continue by importing the necessary libraries and toolkits."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "8a64516a-4cba-4ae6-8649-eb4d9c56550b",
   "metadata": {},
   "outputs": [],
   "source": [
    "from IPython.display import clear_output\n",
    "from math import ceil\n",
    "from matplotlib.ticker import MaxNLocator\n",
    "from scipy.io import loadmat\n",
    "from sklearn.metrics import f1_score, confusion_matrix, accuracy_score, log_loss, precision_score, recall_score, precision_recall_curve\n",
    "from sklearn.model_selection import train_test_split\n",
    "from tensorflow.keras.activations import relu, sigmoid, softmax, tanh\n",
    "from tensorflow.keras import backend as B\n",
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "from tensorflow.keras.datasets import mnist\n",
    "from tensorflow.keras.layers import Conv2D, Dense, Dropout, Flatten, Input, MaxPool2D\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.optimizers import Adam, SGD\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "from time import time\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import os\n",
    "import random\n",
    "import requests\n",
    "\n",
    "%matplotlib inline\n",
    "\n",
    "random.seed(13)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3dab952a-3129-4920-875e-e793ebe2df08",
   "metadata": {},
   "source": [
    "We define the functions through which we will download and load the SVHN dataset from the remote host."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "5f64418c-481f-40ac-b232-61e33a24730c",
   "metadata": {},
   "outputs": [],
   "source": [
    "def update_progress(progress, file_name):\n",
    "    bar_length = 100\n",
    "    block = int(progress)\n",
    "    clear_output(wait = True)\n",
    "    text = file_name + ' progress: [' + '#' * block + '-' * (bar_length - block) + '] ' + '{:.1f}'.format(progress) + '%'\n",
    "    print(text)\n",
    "\n",
    "def download_svhn_train(path):\n",
    "    url = 'http://ufldl.stanford.edu/housenumbers/train_32x32.mat'\n",
    "    file_name = url.split(\"/\")[-1]\n",
    "    \n",
    "    if not os.path.exists(path + file_name):\n",
    "        response = requests.get(url, allow_redirects=True, verify=False, stream=True)\n",
    "\n",
    "        with open(path + file_name, 'wb') as f:\n",
    "            total_length = response.headers.get('content-length')\n",
    "\n",
    "            if total_length is None: # no content length header\n",
    "                f.write(response.content)\n",
    "            else:\n",
    "                dl = 0\n",
    "                total_length = int(total_length)\n",
    "                for data in response.iter_content(chunk_size=8192):\n",
    "                    dl += len(data)\n",
    "                    f.write(data)\n",
    "                    done = (100. * dl / total_length)\n",
    "                    update_progress(done, file_name)\n",
    "        update_progress(100, file_name)\n",
    "    \n",
    "    return path + '/' + file_name\n",
    "\n",
    "def load_svhn_train(path=\"./train_32x32.mat\"):\n",
    "    # Load data from .mat file\n",
    "    data_raw = loadmat(path)\n",
    "    data = np.array(data_raw['X'])\n",
    "    \n",
    "    # Transform data into the correct shape and rescale pixel values\n",
    "    data = np.moveaxis(data, -1, 0) / 255.0\n",
    "    labels = data_raw['y']\n",
    "    \n",
    "    # Split dataset into training and validation set\n",
    "    X_train, X_valid, y_train, y_valid = train_test_split(data, labels, test_size=7326)\n",
    "    \n",
    "    # Labels --> one-hot encodings\n",
    "    y_train = to_categorical(y_train.reshape([-1, 1]))\n",
    "    y_valid = to_categorical(y_valid.reshape([-1, 1]))\n",
    "    \n",
    "    return X_train, y_train, X_valid, y_valid, data, labels"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "85c6ecca-ba6e-4177-aa44-e32678d19a7d",
   "metadata": {},
   "source": [
    "Let's download the SVHN dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "f858af5b-2984-4695-a896-e7122acfe345",
   "metadata": {},
   "outputs": [],
   "source": [
    "svhn_path = download_svhn_train('./')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "31a155d2-b748-48e7-baf6-70a8cb1ce665",
   "metadata": {},
   "source": [
    "Now we define a function through which we will build the fully-connected Neural Network Model.\n",
    "\n",
    "This function creates a fully connected (dense) neural network for multi-class classification.\n",
    "\n",
    "We use the <b>categorical_crossentropy</b> loss function and we measure the <b>accuracy</b> of the model.\n",
    "\n",
    "For the gradient descent optimization, we use <b>Stochastic Gradient Descent</b>."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "c5cf5646-6ebe-454a-b3f5-4b632938a86f",
   "metadata": {},
   "outputs": [],
   "source": [
    "def build_dense_model(X, y, activation=relu, units=[32], dropout_rate=0.1, learning_rate=0.01):\n",
    "    \"\"\"\n",
    "    Build a Neural Network model based on the data and the parameters\n",
    "    \"\"\"\n",
    "    # Model creation\n",
    "    model = Sequential()\n",
    "    \n",
    "    # Input layer\n",
    "    model.add(Input(X.shape[1:]))\n",
    "    \n",
    "    # Flatten input (transform it into 1D array) for feeding it to the dense layers\n",
    "    model.add(Flatten())\n",
    "    \n",
    "    # Dense (intermediate) layers with dropout\n",
    "    for i in range(len(units)):\n",
    "        model.add(Dropout(rate=dropout_rate))\n",
    "        model.add(Dense(units[i], activation=activation))\n",
    "    \n",
    "    # Output layer\n",
    "    model.add(Dense(y.shape[1], activation=softmax))\n",
    "\n",
    "    # Model compilation and description\n",
    "    model.compile(optimizer=SGD(learning_rate=learning_rate), \n",
    "                  loss='categorical_crossentropy', \n",
    "                  metrics=['accuracy'])\n",
    "    model.summary()\n",
    "    \n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cc774110-002d-49b9-9719-7a10c7cb884c",
   "metadata": {},
   "source": [
    "Now we define a function through which we will build the Neural Network Model.\n",
    "\n",
    "This function creates a convolutional neural network for multi-class image classification.\n",
    "\n",
    "We use the <b>categorical_crossentropy</b> loss function and we measure the <b>accuracy</b> of the model.\n",
    "\n",
    "For the gradient descent optimization, we use <b>Adam</b>."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "348db6eb-bb64-43d2-9d1c-7ae385476018",
   "metadata": {},
   "outputs": [],
   "source": [
    "def build_cnn_model(X, y, activation=relu, kernel=(3,3), filters=[9,36,49], units=[32], dropout_rate=0.1, learning_rate=0.01):\n",
    "    \"\"\"\n",
    "    Build a Neural Network model based on the data and the parameters\n",
    "    \"\"\"\n",
    "    # Model creation\n",
    "    model = Sequential()\n",
    "    \n",
    "    # Input layer\n",
    "    model.add(Input(X.shape[1:]))\n",
    "    \n",
    "    # Convolutional and max pooling layers\n",
    "    for i in range(len(filters)):\n",
    "        model.add(Conv2D(filters[i], kernel, padding='same', activation=activation))\n",
    "        model.add(MaxPool2D(kernel))\n",
    "    \n",
    "    # Flatten data to feed to the dense layers\n",
    "    model.add(Flatten())\n",
    "    \n",
    "    # Dense layers\n",
    "    for i in range(len(units)):\n",
    "        model.add(Dropout(rate=dropout_rate))\n",
    "        model.add(Dense(units[i], activation=activation))\n",
    "    \n",
    "    # Output layer\n",
    "    model.add(Dense(y.shape[1], activation=softmax))\n",
    "\n",
    "    # Model compilation and description\n",
    "    model.compile(optimizer=Adam(learning_rate=learning_rate), \n",
    "                  loss='categorical_crossentropy', \n",
    "                  metrics=['accuracy'])\n",
    "    model.summary()\n",
    "            \n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "311a9b70-a7a2-4b44-981d-c4d4b5cd5ad7",
   "metadata": {},
   "source": [
    "## Experimentation"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0aa2df51-dae6-4bd7-8c15-2d1a856b0c24",
   "metadata": {},
   "source": [
    "#### Experiment #1\n",
    "For the first experiment we are going to use the MNIST dataset."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6bde13ee-c838-4a13-a3c2-c530b27bce5d",
   "metadata": {},
   "source": [
    "Let's load the MNIST dataset from Keras into two separate sets:\n",
    "* A training set\n",
    "* A validation set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "63ccd0ab-d5b6-469f-a9a3-f8b0ac5dd542",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load MNIST database\n",
    "(X_train,y_train), (X_valid,y_valid) = mnist.load_data()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "01adf800-c667-485e-9c46-add0e5cacac2",
   "metadata": {},
   "source": [
    "Let's print a randomly selected sample from the training set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "59682b6d-63b8-42a8-8642-c74af40cf82f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "i = random.randint(0, y_train.shape[0]-1)\n",
    "plt.title(f\"Image from class {y_train[i]}\")\n",
    "plt.imshow(X_train[i], cmap='Greys')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9be560ff-2f1c-4777-a16e-cd20b5ead50a",
   "metadata": {},
   "source": [
    "Let's calculate the distinct classes as well as the number of samples from the dataset that belong to each class."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "9ae7dad1-4c4e-4b8b-91d1-1968654809b9",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Collect all labels\n",
    "labels = np.concatenate((y_train, y_valid))\n",
    "\n",
    "# Find classes from unique labels and their respective data counts\n",
    "classes, counts = np.unique(labels, return_counts=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "76f800e9-5489-465b-b11a-21e68e302c83",
   "metadata": {},
   "source": [
    "Let's print a histogram of the counts of the distinct digits in MNIST"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "93c98fc8-16e9-4661-bb21-468df6a72423",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.bar(classes, counts)\n",
    "plt.xticks(classes)\n",
    "plt.xlabel(\"Class\")\n",
    "plt.ylabel(\"Quantity\")\n",
    "plt.title(\"Classes vs Counts in MNIST dataset\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "23d207a0-92a8-4b02-81f0-0f36382cc895",
   "metadata": {},
   "source": [
    "Let's alter the data so that it fits our problem, by:\n",
    "* Rescaling the pixel values from [0,255] to [0,1]\n",
    "* Changing the labels of the samples from single-digit integer numbers to one-hot encoding vectors"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "4591c9eb-16b3-4acb-bfa3-5cece398765c",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Rescale pixel values\n",
    "X_train = X_train / 255.0\n",
    "X_valid = X_valid / 255.0\n",
    "\n",
    "# Labels --> one-hot encodings\n",
    "y_train = to_categorical(y_train.reshape([-1, 1]))\n",
    "y_valid = to_categorical(y_valid.reshape([-1, 1]))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "73d1fa0c-832a-4e67-beec-275ca149f3fe",
   "metadata": {},
   "source": [
    "We initialize 5 distinct dictionaries to store the measurements which will be used for the model evaluation and comparisons."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "e7cb77b7-9e09-498d-ae03-4d1cb5700828",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Initialize lists for data storage\n",
    "times = {'dense': [], 'conv': []}\n",
    "accuracies = {'dense': [], 'conv': []}\n",
    "losses = {'dense': [], 'conv': []}\n",
    "precisions = {'dense': [], 'conv': []}\n",
    "recalls = {'dense': [], 'conv': []}\n",
    "f1s = {'dense': [], 'conv': []}"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "44618203-701e-4445-8ad5-cf603c3d1fbd",
   "metadata": {},
   "source": [
    "We are going to use the following set of parameters for the fully-connected model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "ef906250-5cd3-43d2-8281-68003af334fb",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Number of intermediate layers\n",
    "intermediate_layers = 5\n",
    "\n",
    "# Number of units per intermediate layer\n",
    "units = [32] * intermediate_layers\n",
    "\n",
    "# Dropout rate\n",
    "dropout_rate = 0.1\n",
    "\n",
    "# Learning rate\n",
    "learning_rate = 0.01\n",
    "\n",
    "# Epochs\n",
    "epochs = 10\n",
    "\n",
    "# Epochs for patience\n",
    "patience = 2\n",
    "\n",
    "# Activation function\n",
    "activation = tanh"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "efc00042-b819-4ff1-8535-32ec79c07dfc",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "flatten (Flatten)            (None, 784)               0         \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 784)               0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 32)                25120     \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 32)                0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 32)                1056      \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 32)                0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 32)                1056      \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 32)                0         \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 32)                1056      \n",
      "_________________________________________________________________\n",
      "dropout_4 (Dropout)          (None, 32)                0         \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 32)                1056      \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 10)                330       \n",
      "=================================================================\n",
      "Total params: 29,674\n",
      "Trainable params: 29,674\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Epoch 1/10\n",
      "1875/1875 [==============================] - 25s 14ms/step - loss: 0.7912 - accuracy: 0.7911 - val_loss: 0.3865 - val_accuracy: 0.8974\n",
      "Epoch 2/10\n",
      "1875/1875 [==============================] - 17s 9ms/step - loss: 0.3379 - accuracy: 0.9054 - val_loss: 0.2707 - val_accuracy: 0.9241\n",
      "Epoch 3/10\n",
      "1875/1875 [==============================] - 21s 11ms/step - loss: 0.2618 - accuracy: 0.9248 - val_loss: 0.2234 - val_accuracy: 0.9363\n",
      "Epoch 4/10\n",
      "1875/1875 [==============================] - 20s 11ms/step - loss: 0.2218 - accuracy: 0.9363 - val_loss: 0.2025 - val_accuracy: 0.9405\n",
      "Epoch 5/10\n",
      "1875/1875 [==============================] - 16s 9ms/step - loss: 0.1956 - accuracy: 0.9427 - val_loss: 0.1762 - val_accuracy: 0.9491\n",
      "Epoch 6/10\n",
      "1875/1875 [==============================] - 16s 9ms/step - loss: 0.1769 - accuracy: 0.9481 - val_loss: 0.1652 - val_accuracy: 0.9509\n",
      "Epoch 7/10\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.1635 - accuracy: 0.9523 - val_loss: 0.1532 - val_accuracy: 0.9549\n",
      "Epoch 8/10\n",
      "1875/1875 [==============================] - 15s 8ms/step - loss: 0.1522 - accuracy: 0.9544 - val_loss: 0.1456 - val_accuracy: 0.9584\n",
      "Epoch 9/10\n",
      "1875/1875 [==============================] - 15s 8ms/step - loss: 0.1414 - accuracy: 0.9586 - val_loss: 0.1394 - val_accuracy: 0.9583\n",
      "Epoch 10/10\n",
      "1875/1875 [==============================] - 11s 6ms/step - loss: 0.1347 - accuracy: 0.9592 - val_loss: 0.1327 - val_accuracy: 0.9597\n",
      "Time (sec): 168.36666321754456\n",
      "Accuracy: 0.9558\n",
      "Log Loss: 1.213922861433029\n",
      "Precision: 0.9600630124351694\n",
      "Recall: 0.9648711650551218\n",
      "F1 score: 0.9554217875744142\n"
     ]
    }
   ],
   "source": [
    "model = build_dense_model(X_train, y_train, activation, units, learning_rate)\n",
    "early_stopping = EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=patience, restore_best_weights=True)\n",
    "\n",
    "t = time()\n",
    "dense_history = model.fit(X_train, y_train, epochs=epochs, validation_data=(X_valid,y_valid), callbacks=[early_stopping])\n",
    "t = time() - t\n",
    "\n",
    "# Evaluate the model\n",
    "y_val_pred = model.predict(X_valid)\n",
    "y_val_pred_cat = (np.asarray(y_val_pred)).round()\n",
    "acc, loss, prec, rec, f1 = accuracy_score(y_valid, y_val_pred_cat), \\\n",
    "                            log_loss(y_valid, y_val_pred_cat), \\\n",
    "                            f1_score(y_valid, y_val_pred_cat, average='macro', zero_division=0), \\\n",
    "                            precision_score(y_valid, y_val_pred_cat, average='macro', zero_division=0), \\\n",
    "                            recall_score(y_valid, y_val_pred_cat, average='macro', zero_division=0)\n",
    "\n",
    "times['dense'].append(t)\n",
    "accuracies['dense'].append(acc)\n",
    "losses['dense'].append(loss)\n",
    "precisions['dense'].append(prec)\n",
    "recalls['dense'].append(rec)\n",
    "f1s['dense'].append(f1)\n",
    "\n",
    "print('Time (sec):', t)\n",
    "print('Accuracy:', acc)\n",
    "print('Log Loss:', loss)\n",
    "print('Precision:', prec)\n",
    "print('Recall:', rec)\n",
    "print('F1 score:', f1)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1416ffdb-4ea9-41fe-8e82-9f066754b092",
   "metadata": {},
   "source": [
    "Change the dimensions of both training and validation sets so that they are proper for the convolutional layers."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "bb9d4a5a-78c4-4fa6-aa7a-dea50a437888",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = np.reshape(X_train, X_train.shape+(1,))\n",
    "X_valid = np.reshape(X_valid, X_valid.shape+(1,))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "367535fe-225f-45c2-80ec-9a667aa82931",
   "metadata": {},
   "source": [
    "We are going to use the following set of parameters for the convolutional model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "a4b7c6b5-ed81-43fd-ab7d-b81cd8693931",
   "metadata": {},
   "outputs": [],
   "source": [
    "activation = relu\n",
    "kernel = (3, 3)\n",
    "filters = [9, 36, 49]\n",
    "units = [32]\n",
    "dropout_rate = 0.1\n",
    "patience = 2\n",
    "learning_rate = 0.001\n",
    "epochs = 10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "cc5496f9-fa19-41e5-81f0-807172ae27a1",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d (Conv2D)              (None, 28, 28, 9)         90        \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 9, 9, 9)           0         \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 9, 9, 36)          2952      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 3, 3, 36)          0         \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 3, 3, 49)          15925     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 1, 1, 49)          0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 49)                0         \n",
      "_________________________________________________________________\n",
      "dropout_5 (Dropout)          (None, 49)                0         \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 32)                1600      \n",
      "_________________________________________________________________\n",
      "dense_7 (Dense)              (None, 10)                330       \n",
      "=================================================================\n",
      "Total params: 20,897\n",
      "Trainable params: 20,897\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Epoch 1/10\n",
      "1875/1875 [==============================] - 184s 98ms/step - loss: 0.6924 - accuracy: 0.7650 - val_loss: 0.2289 - val_accuracy: 0.9289\n",
      "Epoch 2/10\n",
      "1875/1875 [==============================] - 170s 91ms/step - loss: 0.2799 - accuracy: 0.9115 - val_loss: 0.1517 - val_accuracy: 0.9528\n",
      "Epoch 3/10\n",
      "1875/1875 [==============================] - 194s 103ms/step - loss: 0.2185 - accuracy: 0.9297 - val_loss: 0.1711 - val_accuracy: 0.9439\n",
      "Epoch 4/10\n",
      "1875/1875 [==============================] - 194s 103ms/step - loss: 0.1956 - accuracy: 0.9381 - val_loss: 0.1367 - val_accuracy: 0.9563\n",
      "Epoch 5/10\n",
      "1875/1875 [==============================] - 194s 104ms/step - loss: 0.1758 - accuracy: 0.9455 - val_loss: 0.1057 - val_accuracy: 0.9669\n",
      "Epoch 6/10\n",
      "1875/1875 [==============================] - 199s 106ms/step - loss: 0.1696 - accuracy: 0.9462 - val_loss: 0.1189 - val_accuracy: 0.9625\n",
      "Epoch 7/10\n",
      "1875/1875 [==============================] - ETA: 0s - loss: 0.1640 - accuracy: 0.9487Restoring model weights from the end of the best epoch.\n",
      "1875/1875 [==============================] - 230s 123ms/step - loss: 0.1640 - accuracy: 0.9487 - val_loss: 0.1297 - val_accuracy: 0.9581\n",
      "Epoch 00007: early stopping\n",
      "Time (sec): 1367.2298517227173\n",
      "Accuracy: 0.9617\n",
      "Log loss: 0.9843551276683807\n",
      "Precision: 0.9663948320406203\n",
      "Recall: 0.9721763285652696\n",
      "F1 score: 0.9610170330171025\n"
     ]
    }
   ],
   "source": [
    "model = build_cnn_model(X_train, y_train, relu, kernel, filters, units, dropout_rate, learning_rate)\n",
    "early_stopping = EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=patience, restore_best_weights=True)\n",
    "\n",
    "t = time()\n",
    "conv_history = model.fit(X_train, y_train, epochs=epochs, validation_data=(X_valid, y_valid), callbacks=[early_stopping])\n",
    "t = time() - t\n",
    "\n",
    "# Evaluate the model\n",
    "y_val_pred = model.predict(X_valid)\n",
    "y_val_pred_cat = (np.asarray(y_val_pred)).round()\n",
    "acc, loss, prec, rec, f1 = accuracy_score(y_valid, y_val_pred_cat), \\\n",
    "                            log_loss(y_valid, y_val_pred_cat), \\\n",
    "                            f1_score(y_valid, y_val_pred_cat, average='macro', zero_division=0), \\\n",
    "                            precision_score(y_valid, y_val_pred_cat, average='macro', zero_division=0), \\\n",
    "                            recall_score(y_valid, y_val_pred_cat, average='macro', zero_division=0)\n",
    "\n",
    "times['conv'].append(t)\n",
    "accuracies['conv'].append(acc)\n",
    "losses['conv'].append(loss)\n",
    "precisions['conv'].append(prec)\n",
    "recalls['conv'].append(rec)\n",
    "f1s['conv'].append(f1)\n",
    "\n",
    "print('Time (sec):', t)\n",
    "print('Accuracy:', acc)\n",
    "print('Log loss:', loss)\n",
    "print('Precision:', prec)\n",
    "print('Recall:', rec)\n",
    "print('F1 score:', f1)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b446f44c-ea1b-4a34-94ba-f32bf6e52ffe",
   "metadata": {},
   "source": [
    "Let's plot the accuracies and losses to see what we achieved and compare the models."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "665091d9-d8f5-49d3-871b-4afdf95794a8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(16,6))\n",
    "\n",
    "xr = min(len(dense_history.history['accuracy']), len(conv_history.history['accuracy']))\n",
    "\n",
    "# Plot training & validation accuracy values\n",
    "plt.subplot(1, 2, 1)\n",
    "plt.plot(range(1,xr+1),dense_history.history['accuracy'][:xr])\n",
    "plt.plot(range(1,xr+1),dense_history.history['val_accuracy'][:xr])\n",
    "plt.plot(range(1,xr+1),conv_history.history['accuracy'][:xr])\n",
    "plt.plot(range(1,xr+1),conv_history.history['val_accuracy'][:xr])\n",
    "plt.title(\"Models' accuracy on MNIST\")\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Dense-Train', 'Dense-Validation', 'Convolutional-Train', 'Convolutional-Validation'], loc='best')\n",
    "plt.ylim(0., 1.)\n",
    "plt.xlim(1, epochs)\n",
    "plt.xticks(range(1,epochs+1))\n",
    "\n",
    "ylim = ceil(max([max(dense_history.history['loss']), max(dense_history.history['val_loss']), max(conv_history.history['loss']), max(conv_history.history['val_loss'])]))\n",
    "\n",
    "# Plot training & validation loss values\n",
    "plt.subplot(1, 2, 2)\n",
    "plt.plot(range(1,xr+1),dense_history.history['loss'][:xr])\n",
    "plt.plot(range(1,xr+1),dense_history.history['val_loss'][:xr])\n",
    "plt.plot(range(1,xr+1),conv_history.history['loss'][:xr])\n",
    "plt.plot(range(1,xr+1),conv_history.history['val_loss'][:xr])\n",
    "plt.title(\"Models' loss on MNIST\")\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Dense-Train', 'Dense-Validation', 'Convolutional-Train', 'Convolutional-Validation'], loc='best')\n",
    "plt.ylim(0., ylim)\n",
    "plt.xlim(1, epochs)\n",
    "plt.xticks(range(1,epochs+1))\n",
    "\n",
    "plt.show()\n",
    "# fig.savefig('./acc-loss-1.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7b89d520-1344-4a0a-8a42-29956d229d90",
   "metadata": {},
   "source": [
    "Now, we should compare the two models, to see which performs better in total. For the comparison, the following metrics will be used:\n",
    "* Training time\n",
    "* Accuracy\n",
    "* Log loss\n",
    "* Precision\n",
    "* Recall\n",
    "* F1 score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "7b7fad74-91b5-4078-839c-f1c50bdcddeb",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1800x1080 with 6 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(25,15))\n",
    "\n",
    "plt.subplot(2, 3, 1)\n",
    "plt.bar('dense', times['dense'][0], color='r', width=0.3,\n",
    "        edgecolor='grey', label='dense')\n",
    "plt.bar('convolutional', times['conv'][0], color='b', width=0.3,\n",
    "        edgecolor='grey', label='convolutional')\n",
    "plt.title(\"Models' training times\")\n",
    "plt.ylabel('Time (sec)')\n",
    "\n",
    "plt.subplot(2, 3, 2)\n",
    "plt.bar('dense', accuracies['dense'][0], color='r', width=0.3,\n",
    "        edgecolor='grey', label='dense')\n",
    "plt.bar('convolutional', accuracies['conv'][0], color='b', width=0.3,\n",
    "        edgecolor='grey', label='convolutional')\n",
    "plt.title(\"Models' accuracies\")\n",
    "plt.ylabel('Accuracy')\n",
    "\n",
    "plt.subplot(2, 3, 3)\n",
    "plt.bar('dense', losses['dense'][0], color='r', width=0.3,\n",
    "        edgecolor='grey', label='dense')\n",
    "plt.bar('convolutional', losses['conv'][0], color='b', width=0.3,\n",
    "        edgecolor='grey', label='convolutional')\n",
    "plt.title(\"Models' log losses\")\n",
    "plt.ylabel('Log Loss')\n",
    "\n",
    "plt.subplot(2, 3, 4)\n",
    "plt.bar('dense', precisions['dense'][0], color='r', width=0.3,\n",
    "        edgecolor='grey', label='dense')\n",
    "plt.bar('convolutional', precisions['conv'][0], color='b', width=0.3,\n",
    "        edgecolor='grey', label='convolutional')\n",
    "plt.title(\"Models' precisions\")\n",
    "plt.ylabel('Precision')\n",
    "\n",
    "plt.subplot(2, 3, 5)\n",
    "plt.bar('dense', recalls['dense'][0], color='r', width=0.3,\n",
    "        edgecolor='grey', label='dense')\n",
    "plt.bar('convolutional', recalls['conv'][0], color='b', width=0.3,\n",
    "        edgecolor='grey', label='convolutional')\n",
    "plt.title(\"Models' recalls\")\n",
    "plt.ylabel('Recall')\n",
    "\n",
    "plt.subplot(2, 3, 6)\n",
    "plt.bar('dense', f1s['dense'][0], color='r', width=0.3,\n",
    "        edgecolor='grey', label='dense')\n",
    "plt.bar('convolutional', f1s['conv'][0], color='b', width=0.3,\n",
    "        edgecolor='grey', label='convolutional')\n",
    "plt.title(\"Models' F1 scores\")\n",
    "plt.ylabel('F1 score')\n",
    "\n",
    "plt.show()\n",
    "# fig.savefig(f'./measurements-1.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "599cd3ef-0434-4605-aa46-b7681435da1b",
   "metadata": {},
   "source": [
    "#### Experiment #2\n",
    "For the second experiment we are going to use the SVHN dataset."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "12adbabd-8071-41ad-8491-2bdb3fb7db41",
   "metadata": {},
   "source": [
    "Let's load the SVHN dataset into two separate sets:\n",
    "* A training set\n",
    "* A validation set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "31f0891f-b320-418c-bd63-6c158c132adb",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load SVHN dataset\n",
    "X_train, y_train, X_valid, y_valid, data, labels = load_svhn_train(svhn_path)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "59c2cff2-8f31-4b9f-97a8-b10298f5ded8",
   "metadata": {},
   "source": [
    "Let's calculate the distinct classes as well as the number of samples from the dataset that belong to each class."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "c0b454d3-7642-4c2e-a2cb-e95ef26a99b2",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Find classes from unique labels and their respective data counts\n",
    "classes, counts = np.unique(labels, return_counts=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "61b7dd85-c53a-495d-af66-6f816190fbb6",
   "metadata": {},
   "source": [
    "Let's print a randomly selected training sample from the dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "cf402450-eebc-45ac-82eb-188bece4fde1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "i = random.randint(0, data.shape[0]-1)\n",
    "plt.title(f\"Image from class {labels[i][0] % classes.size}\")\n",
    "plt.imshow(data[i], cmap='Greys')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3540a0de-98b4-4b0c-8fa3-36bbc7aa44ba",
   "metadata": {},
   "source": [
    "Let's print a histogram of the counts of the distinct classes in SVHN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "cccfe10e-f639-4fad-a7b0-d9d77ca568ca",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.bar(classes % classes.size, counts)\n",
    "plt.xticks(classes % classes.size)\n",
    "plt.xlabel(\"Class\")\n",
    "plt.ylabel(\"Quantity\")\n",
    "plt.title(\"Classes vs Counts in SVHN dataset\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b76d1390-4ce4-49ef-ad0e-2daed73172c2",
   "metadata": {},
   "source": [
    "We initialize 5 distinct dictionaries to store the measurements which will be used for the model evaluation and comparisons."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "678ffa97-76c4-401b-a389-092ce75508dd",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Initialize lists for data storage\n",
    "times = {'dense': [], 'conv': []}\n",
    "accuracies = {'dense': [], 'conv': []}\n",
    "losses = {'dense': [], 'conv': []}\n",
    "precisions = {'dense': [], 'conv': []}\n",
    "recalls = {'dense': [], 'conv': []}\n",
    "f1s = {'dense': [], 'conv': []}"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "51cb408a-c46f-4491-887d-9d5066afbd15",
   "metadata": {},
   "source": [
    "We are going to use the following set of parameters for the fully-connected model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "7ccd8687-8713-4d67-87c3-2d135fd394b9",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Number of intermediate layers\n",
    "intermediate_layers = 3\n",
    "\n",
    "# Number of units per intermediate layer\n",
    "units = [512] * intermediate_layers\n",
    "\n",
    "# Dropout rate\n",
    "dropout_rate = 0.1\n",
    "\n",
    "# Learning rate\n",
    "learning_rate = 0.01\n",
    "\n",
    "# Epochs\n",
    "epochs = 10\n",
    "\n",
    "# Epochs for patience\n",
    "patience = 2\n",
    "\n",
    "# Activation function\n",
    "activation = tanh"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "e2398381-3a57-435c-a8d2-9ac7c17e34da",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "flatten_2 (Flatten)          (None, 3072)              0         \n",
      "_________________________________________________________________\n",
      "dropout_6 (Dropout)          (None, 3072)              0         \n",
      "_________________________________________________________________\n",
      "dense_8 (Dense)              (None, 512)               1573376   \n",
      "_________________________________________________________________\n",
      "dropout_7 (Dropout)          (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dense_9 (Dense)              (None, 512)               262656    \n",
      "_________________________________________________________________\n",
      "dropout_8 (Dropout)          (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dense_10 (Dense)             (None, 512)               262656    \n",
      "_________________________________________________________________\n",
      "dense_11 (Dense)             (None, 11)                5643      \n",
      "=================================================================\n",
      "Total params: 2,104,331\n",
      "Trainable params: 2,104,331\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Epoch 1/10\n",
      "2061/2061 [==============================] - 39s 19ms/step - loss: 2.1985 - accuracy: 0.2081 - val_loss: 2.2255 - val_accuracy: 0.2133\n",
      "Epoch 2/10\n",
      "2061/2061 [==============================] - 79s 38ms/step - loss: 1.6776 - accuracy: 0.4289 - val_loss: 1.6289 - val_accuracy: 0.4436\n",
      "Epoch 3/10\n",
      "2061/2061 [==============================] - 76s 37ms/step - loss: 1.2972 - accuracy: 0.5802 - val_loss: 1.2595 - val_accuracy: 0.5906\n",
      "Epoch 4/10\n",
      "2061/2061 [==============================] - 81s 39ms/step - loss: 1.1245 - accuracy: 0.6423 - val_loss: 1.1698 - val_accuracy: 0.6290\n",
      "Epoch 5/10\n",
      "2061/2061 [==============================] - 66s 32ms/step - loss: 1.0204 - accuracy: 0.6778 - val_loss: 1.0780 - val_accuracy: 0.6607\n",
      "Epoch 6/10\n",
      "2061/2061 [==============================] - 63s 31ms/step - loss: 0.9440 - accuracy: 0.7028 - val_loss: 1.0551 - val_accuracy: 0.6556\n",
      "Epoch 7/10\n",
      "2061/2061 [==============================] - 70s 34ms/step - loss: 0.8847 - accuracy: 0.7209 - val_loss: 0.8911 - val_accuracy: 0.7187\n",
      "Epoch 8/10\n",
      "2061/2061 [==============================] - 77s 37ms/step - loss: 0.8366 - accuracy: 0.7359 - val_loss: 1.0678 - val_accuracy: 0.6577\n",
      "Epoch 9/10\n",
      "2061/2061 [==============================] - 65s 31ms/step - loss: 0.7983 - accuracy: 0.7470 - val_loss: 0.8573 - val_accuracy: 0.7295\n",
      "Epoch 10/10\n",
      "2061/2061 [==============================] - 44s 21ms/step - loss: 0.7672 - accuracy: 0.7579 - val_loss: 0.8880 - val_accuracy: 0.7106\n",
      "Time (sec): 663.0601871013641\n",
      "Accuracy: 0.6345891345891346\n",
      "Log Loss: 5.2327410195194695\n",
      "Precision: 0.6329802631983983\n",
      "Recall: 0.7493357611247032\n",
      "F1 score: 0.5644807873232834\n"
     ]
    }
   ],
   "source": [
    "model = build_dense_model(X_train, y_train, activation, units, learning_rate)\n",
    "early_stopping = EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=patience, restore_best_weights=True)\n",
    "\n",
    "t = time()\n",
    "dense_history = model.fit(X_train, y_train, epochs=epochs, validation_data=(X_valid,y_valid), callbacks=[early_stopping])\n",
    "t = time() - t\n",
    "\n",
    "# Evaluate the model\n",
    "y_val_pred = model.predict(X_valid)\n",
    "y_val_pred_cat = (np.asarray(y_val_pred)).round()\n",
    "acc, loss, prec, rec, f1 = accuracy_score(y_valid, y_val_pred_cat), \\\n",
    "                            log_loss(y_valid, y_val_pred_cat), \\\n",
    "                            f1_score(y_valid, y_val_pred_cat, average='macro', zero_division=0), \\\n",
    "                            precision_score(y_valid, y_val_pred_cat, average='macro', zero_division=0), \\\n",
    "                            recall_score(y_valid, y_val_pred_cat, average='macro', zero_division=0)\n",
    "\n",
    "times['dense'].append(t)\n",
    "accuracies['dense'].append(acc)\n",
    "losses['dense'].append(loss)\n",
    "precisions['dense'].append(prec)\n",
    "recalls['dense'].append(rec)\n",
    "f1s['dense'].append(f1)\n",
    "\n",
    "print('Time (sec):', t)\n",
    "print('Accuracy:', acc)\n",
    "print('Log Loss:', loss)\n",
    "print('Precision:', prec)\n",
    "print('Recall:', rec)\n",
    "print('F1 score:', f1)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "26c225ed-bc58-4395-8511-4a066df36c27",
   "metadata": {},
   "source": [
    "We are going to use the following set of parameters for the convolutional model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "767fee85-a588-418e-923e-6c69f9d25007",
   "metadata": {},
   "outputs": [],
   "source": [
    "activation = relu\n",
    "kernel = (3, 3)\n",
    "filters = [9, 36, 49]\n",
    "units = [32]\n",
    "dropout_rate = 0.1\n",
    "patience = 2\n",
    "learning_rate = 0.001\n",
    "epochs = 10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "7f58b83d-d75f-4a67-ae17-d5644b4572b5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_3\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_3 (Conv2D)            (None, 32, 32, 9)         252       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_3 (MaxPooling2 (None, 10, 10, 9)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 10, 10, 36)        2952      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_4 (MaxPooling2 (None, 3, 3, 36)          0         \n",
      "_________________________________________________________________\n",
      "conv2d_5 (Conv2D)            (None, 3, 3, 49)          15925     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_5 (MaxPooling2 (None, 1, 1, 49)          0         \n",
      "_________________________________________________________________\n",
      "flatten_3 (Flatten)          (None, 49)                0         \n",
      "_________________________________________________________________\n",
      "dropout_9 (Dropout)          (None, 49)                0         \n",
      "_________________________________________________________________\n",
      "dense_12 (Dense)             (None, 32)                1600      \n",
      "_________________________________________________________________\n",
      "dense_13 (Dense)             (None, 11)                363       \n",
      "=================================================================\n",
      "Total params: 21,092\n",
      "Trainable params: 21,092\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Epoch 1/10\n",
      "2061/2061 [==============================] - 181s 88ms/step - loss: 2.0231 - accuracy: 0.2782 - val_loss: 1.5645 - val_accuracy: 0.4559\n",
      "Epoch 2/10\n",
      "2061/2061 [==============================] - 172s 83ms/step - loss: 1.4376 - accuracy: 0.5028 - val_loss: 1.2089 - val_accuracy: 0.5995\n",
      "Epoch 3/10\n",
      "2061/2061 [==============================] - 176s 85ms/step - loss: 1.2064 - accuracy: 0.5936 - val_loss: 1.0922 - val_accuracy: 0.6493\n",
      "Epoch 4/10\n",
      "2061/2061 [==============================] - 175s 85ms/step - loss: 1.0667 - accuracy: 0.6464 - val_loss: 0.9895 - val_accuracy: 0.6821\n",
      "Epoch 5/10\n",
      "2061/2061 [==============================] - 179s 87ms/step - loss: 0.9800 - accuracy: 0.6779 - val_loss: 0.9044 - val_accuracy: 0.7052\n",
      "Epoch 6/10\n",
      "2061/2061 [==============================] - 160s 78ms/step - loss: 0.9059 - accuracy: 0.7051 - val_loss: 0.9021 - val_accuracy: 0.7095\n",
      "Epoch 7/10\n",
      "2061/2061 [==============================] - 68s 33ms/step - loss: 0.8560 - accuracy: 0.7240 - val_loss: 0.8077 - val_accuracy: 0.7450\n",
      "Epoch 8/10\n",
      "2061/2061 [==============================] - 47s 23ms/step - loss: 0.8193 - accuracy: 0.7349 - val_loss: 0.7598 - val_accuracy: 0.7576\n",
      "Epoch 9/10\n",
      "2061/2061 [==============================] - 162s 79ms/step - loss: 0.7890 - accuracy: 0.7455 - val_loss: 0.7211 - val_accuracy: 0.7715\n",
      "Epoch 10/10\n",
      "2061/2061 [==============================] - 186s 90ms/step - loss: 0.7566 - accuracy: 0.7563 - val_loss: 0.7226 - val_accuracy: 0.7697\n",
      "Time (sec): 1508.9998018741608\n",
      "Accuracy: 0.700928200928201\n",
      "Log Loss: 4.433130622528434\n",
      "Precision: 0.6811265894158107\n",
      "Recall: 0.7753097680358817\n",
      "F1 score: 0.6107658027074837\n"
     ]
    }
   ],
   "source": [
    "model = build_cnn_model(X_train, y_train, relu, kernel, filters, units, dropout_rate, learning_rate)\n",
    "early_stopping = EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=patience, restore_best_weights=True)\n",
    "\n",
    "t = time()\n",
    "conv_history = model.fit(X_train, y_train, epochs=epochs, validation_data=(X_valid, y_valid), callbacks=[early_stopping])\n",
    "t = time() - t\n",
    "\n",
    "# Evaluate the model\n",
    "y_val_pred = model.predict(X_valid)\n",
    "y_val_pred_cat = (np.asarray(y_val_pred)).round()\n",
    "acc, loss, prec, rec, f1 = accuracy_score(y_valid, y_val_pred_cat), \\\n",
    "                            log_loss(y_valid, y_val_pred_cat), \\\n",
    "                            f1_score(y_valid, y_val_pred_cat, average='macro', zero_division=0), \\\n",
    "                            precision_score(y_valid, y_val_pred_cat, average='macro', zero_division=0), \\\n",
    "                            recall_score(y_valid, y_val_pred_cat, average='macro', zero_division=0)\n",
    "\n",
    "times['conv'].append(t)\n",
    "accuracies['conv'].append(acc)\n",
    "losses['conv'].append(loss)\n",
    "precisions['conv'].append(prec)\n",
    "recalls['conv'].append(rec)\n",
    "f1s['conv'].append(f1)\n",
    "\n",
    "print('Time (sec):', t)\n",
    "print('Accuracy:', acc)\n",
    "print('Log Loss:', loss)\n",
    "print('Precision:', prec)\n",
    "print('Recall:', rec)\n",
    "print('F1 score:', f1)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c7588128-dcf5-4cde-a590-78b43c568322",
   "metadata": {},
   "source": [
    "Let's plot the accuracies and losses to see what we achieved and compare the models."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "9749eb1c-43ea-48cf-b448-fc6db0f36d36",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA7YAAAGDCAYAAAAf0oyvAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuNCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8QVMy6AAAACXBIWXMAAAsTAAALEwEAmpwYAAC1TElEQVR4nOzdd3zN1//A8dfJvdl7WrH3DpKITSm+atau0qWqpfPXVqtL995aWqo6FC2ltLq01B5B7E0QKwmRPe89vz8+V0QECUluwvv5eNyH3Pv5nPN5f27VyftzltJaI4QQQgghhBBClFcO9g5ACCGEEEIIIYS4HpLYCiGEEEIIIYQo1ySxFUIIIYQQQghRrkliK4QQQgghhBCiXJPEVgghhBBCCCFEuSaJrRBCCCGEEEKIck0SW3HTUUrVUEpppZS5EOferZRaVRpxCSGEEOLalHbbrpSapJT6/nrqEEIUL0lsRZmmlIpWSmUppQLyfR5la8Bq2Cm0iyilZiql7rZ3HDcKpVSwUmq+UipeKZWolNpu+0XERSl1Til1SwFlPlRKzbP9HK2U6pbv+EW/yNjOOa2Ucs/z2Wil1PISvDUhhLjpSdtuP0opH6XUDKXUKaVUslJqn1Jqgu3YHqXUvQWUeVQpFWn7eblSanS+452VUjF53i9XSmUoparm+aybUiq6xG5MCCSxFeXDYWD4+TdKqaaAq/3CKV8K8/S6DPoOOAZUB/yBUcBprXUGMNf2PpdSyoTxd+SbIl7HDDx63dEKIYQoKmnb7eNDwANoCHgDfYGDtmPfkK99tRlJ0dvXVOCFa4xRiGsiia0oD77j4n9o7wK+zXuCUspbKfWtUipOKXVEKfW8UsrBdsyklHrP1vt3CLitgLJfKaVOKqWOK6VesyVK5DtP2XoFY229iNuUUk2uFrxSKlwptdbW03hSKTVZKeWU53hjpdTfSqmzth7EiXninqiUOmh7qrpJKVW1oOFWeZ+g2nomV9tiPQtMUkrVVkr9q5Q6Y/seZimlfPKUr6qU+tn2/Z2xxehsi6lpnvOClFLpSqnAAu7Twfa9H7F9R98qpbxtx87HfJdS6qgthueu8LWFATO11qla6xyt9Rat9e+2Y98AA5VSbnnO74Hx79nv+Su6ineBJ/N+F0IIIUpFuW7bC6inr1Jqp62tX66Uapjn2ARbDMlKqb1Kqa62z8OVUpFKqSRb+//BFeq/Xyl1wNYuL1JKVc5zTCulxiql9iulEpRSnyml1GWqCgN+0FonaK2tWus9Wut5tmPfAe2VUtXz1N0QaAbMLuJX8gkwXClVp4jlhLhmktiK8mAd4KWUamhrlIYC+ee1fIrx5LEW0AmjsbzHdux+oDfQAggFBuUr+w2QA9SxndMdGM2lugMdgXqAjy2OMwBa67u11jMvE78FeBwIANoAXYGHAJRSnsBS4A+gsi2Gf2zlnsB4mt0L8ALuBdIuc438WgOHgCDgdUABb9qu0RCoCkyyxWACfgWOADWAKsAcrXUmMAe4M0+9w4GlWuu4Aq55t+3VBeO/gwcwOd857YH6tu/gxbwNfz7rgM+UUsOUUtXyHtBarwFOArfn+XgkRkOdc5n6LicSWA48WcRyQgghrk95b9tzKaXqYSR+jwGBwBJgsVLKSSlVHxgPhGmtPTEexEbbin4MfKy19gJqAz9epv5bMNrwIUAljPZ6Tr7TemMkrc1t5/W4TLjrgNeVUvcopermPaC1jgGWYbSp540Clmit46/wFRTkODAN2+8aQpQGSWxFeXH+ye6twB6MfzCB3MRsKPCs1jpZax0NvM+Ff5iHAB9prY9prc9iNA7ny1YA/gc8ZusdjMUYpjOsgBiyAU+gAaC01ru11ievFrjWepPWep2t5zEa+AKjgQajITqltX5fa51hi3+97dho4Hmt9V5t2Kq1PnPVb8pwQmv9qe2a6VrrA1rrv7XWmbak9IM8MYRjJLxP2b6DDK31+bmo3wB3nH9CjvGdfneZa44APtBaH9JapwDPAsPUxUOhX7bFsxXYitEAF2QwsBJjGNNhZcy7Cstz/FtsT/qVUl5APy4dJrXQ9uT8nFLqHPD5Za71IvBwQb3QQgghSlS5bdvzGQr8Zmtns4H3MIZVt8V4uO0MNFJKOWqto7XW54f+ZgN1lFIBWusUrfW6y9Q/Apihtd5se+j8LNBGXTwX+S2t9Tmt9VGM5DTkMnU9DMzCSLZ32XqB/5fn+DfYvmNb2z+CS9vXT/K1r79e5lpvAn2UUo0vc1yIYiWJrSgvvgPuwOgR/DbfsQDACeMJ5nlHMHoewUjajuU7dl51wBE4mecf6C8wejovorX+F6MH8jPgtFLqS1tSdUVKqXpKqV+VsVBDEvCGLWYwek4PXqbolY5dTd77PT+EeI5tKFQSxlPxvDEcKai305ZkpwKdlFINMJ58L7rMNStz6X8DM1Ahz2en8vychtGrewnbEKlntNaNbeWjMBLV80OrvgW6KKWqYDylP6C13pKvmv5aa5/zL2y95AVcawdGo/zMZe5LCCFEySi3bXs+F7V/WmurLbYqWusDGD25k4BYW1t8fhjxfRg9xXuUUhuVUr0LWX8KRq9ylTznFLZ9Tddav6G1boWxhsWPwE9KKT/bKT8DlZRSEUBnwA34LV81j+RrXwuM2/YgfTLwymXuS4hiJYmtKBe01kcwFprohfGPbl7xGE89q+f5rBoXnvyexEje8h477xiQCQTk+Ufay5ZQFRTHJ7bGoDFGY/RUIcKfgvEkuq5tuNFEjKHB569f+zLlLncs1fZn3jmmFfOHmu/9m7bPmtliuDNfDNXU5ReZ+sZ2/khgnm0Bp4Kc4NL/BjnA6cucXyi24U/vYTTsfrbPjmL06I6wxZX/F6KiegljWFuVq50ohBCieJTztj2vi9o/20PYqudj1Vr/oLVubztHA2/bPt+vtR6OkXC/DcxTeVbqv0L97hhJ6fECzi00rfX5h+3uQE3bZ2nAPIye9JEYU5OyruMy72JMUWp1PbEKURiS2Iry5D7gFq11at4PtdYWjCeOryulPG2LHjzBhbk6PwKPKGMLGV/y9MzZhhv9BbyvlPJSxgJItZVSnchHKRWmlGqtlHLESC4zMIYYXY0nkASk2Ho9H8xz7FegolLqMWUs1uSplGptOzYdeFUpVddY20I1U0r5256AHgfuVMbiGfdy+eQ4bwwpwDlbL2feRnsDxi8Ibyml3JWxpU67PMe/AwZgJLdXSiBnA48rpWoqpTwwGsu51zDvFaXU20qpJkops20e8oMYvbJ5h2J/gzGUqh3GsKprZnuiPhd45HrqEUIIUWTltW3P60fgNqVUV1s9/4eRWK9RStVXSt2ilHK21Z1+vn6l1J1KqUBbD+85W10FXfsH4B6lVIitnjeA9bbh2UWilHrBds9OSikXjJ0BzgF785z2Dcbw6oEUfTXki2itz2EMIX/6euoRojAksRXlhtb6oNY68jKHH8ZokA4BqzAagRm2Y9OAPzHmdG7m0qfCozCGO+0CEjCeVFYq4BpetroSMIYEncHoSbyaJzGGWiXbys/Nc0/JGHOL+mAMI9qP8WQTjHmwP2I0zknAV1zYCuF+jOT0DMYT5jVXieFloCWQiDGkKPc7sP3y0AdjmPFRIAajQTt/PAbje9MYvaSXMwMjCV6B8QQ+A+O/y7VwAxZgNLaHMJ5U9813zjzAF/jnGuZDFeQVjKfWQgghSkk5btvz3sNejIe/n2L0NPcB+th6Op2Bt2yfn8LonZ1oK9oT2KmUSsFYSGpYQaOitNb/YKw5MR/jQXRtCp4vXKhwga9t8ZzA+B3kNtvw5vNWYPy+cFxrvfEar5PXxxT9YYEQRaa0zj9iUQghLqaUmoGxINXz9o5FCCGEEEKI/C43p04IIQBjD1qMrXVa2DkUIYQQQgghClRiQ5GVUjOUsdn1jsscV0qpT2zLjG9TSrUsqViEENdGKfUqsAN4V2t92N7xCCGuj20O/Qal1Fal1E6l1MsFnCPtsxBCiHKnJOfYzsSYO3A5/wPq2l5jMFaOFUKUIVrrF7TWHlrr1+0dixCiWGRiLNTTHGOfy562bT3ykvZZCCFEuVNiia3WegVw9gqn9AO+1YZ1gI9SqqBJ/UIIIYQoBrY29/wiMY62V/7FNqR9FkIIUe7Yc1XkKly8sXYMsoekEEIIUaJs24RFAbHA31rr9flOkfZZCCFEuWPPxaNUAZ8VuESzUmoMxnAo3N3dWzVo0KAk4xJCCHET2bRpU7zWOtDecZQW2xZfIUopH2CBUqqJ1jrvehiFap+lbRZCCFFSrqVttmdiGwNUzfM+GGM/rUtorb8EvgQIDQ3VkZGX2+5MCCGEKBql1BF7x2APWutzSqnlGOth5E1sC9U+S9sshBCipFxL22zPociLgFG21RcjgESt9Uk7xiOEEELc0JRSgbaeWpRSrkA3YE++06R9FkIIUe6UWI+tUmo20BkIUErFAC9hLFKB1noqsAToBRwA0oB7SioWIYQQQgBQCfhGKWXCeLj9o9b6V6XUWJD2WQghRPlVYomt1nr4VY5rYFxJXV8IIYQQF9NabwNaFPD51Dw/S/sshBCi3LHnHFshxE0uOzubmJgYMjIy7B2KuAm4uLgQHByMo6OjvUMRQogyS9pmUZqKs22WxFYIYTcxMTF4enpSo0YNlCpoIVYhiofWmjNnzhATE0PNmjXtHY4QQpRZ0jaL0lLcbbM9F48SQtzkMjIy8Pf3l4ZTlDilFP7+/tIDIYQQVyFtsygtxd02S2IrhLAraThFaZG/a0IIUTjy76UoLcX5d00SWyHETc1kMhESEkLjxo1p3rw5H3zwAVartVRjGDduHCEhITRq1AhXV1dCQkIICQlh3rx5hSrfq1cvzp07V7JBCiGEEKVE2mZxLWSOrRDipubq6kpUVBQAsbGx3HHHHSQmJvLyyy+XWgyfffYZANHR0fTu3Ts3nvMsFgsmk+my5ZcsWVKS4QkhhBClStpmcS2kx1YIIWyCgoL48ssvmTx5MlprLBYLTz31FGFhYTRr1owvvvgCgOXLl9O5c2cGDRpEgwYNGDFiBMYOKfDMM8/QqFEjmjVrxpNPPglAXFwcAwcOJCwsjLCwMFavXn3VWJYvX06XLl244447aNq0KQD9+/enVatWNG7cmC+//DL33Bo1ahAfH090dDQNGzbk/vvvp3HjxnTv3p309PTi/pqEEEKIUiNtsygs6bEVQpQJLy/eya4TScVaZ6PKXrzUp3GRytSqVQur1UpsbCy//PIL3t7ebNy4kczMTNq1a0f37t0B2LJlCzt37qRy5cq0a9eO1atX06hRIxYsWMCePXtQSuUOQXr00Ud5/PHHad++PUePHqVHjx7s3r37qrFs2LCBHTt25K4UOGPGDPz8/EhPTycsLIyBAwfi7+9/UZn9+/cze/Zspk2bxpAhQ5g/fz533nlnkb4DIYQQAqRtLoi0zWWXJLZCCJHP+Se8f/31F9u2bcudT5OYmMj+/ftxcnIiPDyc4OBgAEJCQoiOjiYiIgIXFxdGjx7NbbfdRu/evQFYunQpu3btyq0/KSmJ5ORkPD09rxhHeHj4Rcvff/LJJyxYsACAY8eOsX///ksaz5o1axISEgJAq1atiI6OvvYvQgghhCgjpG0WVyOJrRCiTCjq09uScujQIUwmE0FBQWit+fTTT+nRo8dF5yxfvhxnZ+fc9yaTiZycHMxmMxs2bOCff/5hzpw5TJ48mX///Rer1cratWtxdXW9qJ4ePXpw+vRpQkNDmT59+iWxuLu7X3TNpUuXsnbtWtzc3OjcuXOBy+Pnj0uGOwkhhLhW0jZL21yeyBxbIYSwiYuLY+zYsYwfPx6lFD169GDKlClkZ2cDsG/fPlJTUy9bPiUlhcTERHr16sVHH32Uu9BE9+7dmTx5cu555z//888/iYqKKrDhzC8xMRFfX1/c3NzYs2cP69atu/YbFUIIIcoJaZtFYUmPrRDippaenk5ISAjZ2dmYzWZGjhzJE088AcDo0aOJjo6mZcuWaK0JDAxk4cKFl60rOTmZfv36kZGRgdaaDz/8EDCGKY0bN45mzZqRk5NDx44dmTp1apHi7NmzJ1OnTqVZs2bUr1+fiIiIa75nIYQQoiyTtllcC3V+vHp5ERoaqiMjI+0dhhCiGOzevZuGDRvaOwxxEyno75xSapPWOtROId0QpG0W4sYhbbMobcXVNstQZCGEEEIIIYQQ5ZoktkIIIYQQQgghyjVJbIUQQgghhBBClGuS2AohhBBCCCGEKNcksRVCCCGEEEIIUa5JYiuEEEIIIYQQolyTxFYIcVMzmUyEhITQuHFjmjdvzgcffIDVai3VGKKjowkODr7kuiEhIWzYsOGyZZo0aQJAZGQkjzzySIHn1ahRg/j4+Cte/4033rjofdu2bQsbuhBCCFHspG2WtvlaSGIrhLipubq6EhUVxc6dO/n7779ZsmQJL7/8cqnGUKNGDapWrcrKlStzP9uzZw/JycmEh4dftXxoaCiffPLJNV8/f+O5Zs2aa65LCCGEuF7SNkvbfC0ksRVCCJugoCC+/PJLJk+ejNYai8XCU089RVhYGM2aNeOLL74AYPny5XTu3JlBgwbRoEEDRowYgdYagGeeeYZGjRrRrFkznnzySQDi4uIYOHAgYWFhhIWFsXr16kuuPXz4cObMmZP7fs6cOQwfPpzo6Gg6dOhAy5YtadmyZYEN2/Lly+nduzcAZ86coXv37rRo0YIHHnggNy6A/v3706pVKxo3bsyXX36ZG296ejohISGMGDECAA8PDwC01jz11FM0adKEpk2bMnfu3KvevxBCCFGcpG2WtrmwzPYOQAghAPj9GTi1vXjrrNgU/vdWkYrUqlULq9VKbGwsv/zyC97e3mzcuJHMzEzatWtH9+7dAdiyZQs7d+6kcuXKtGvXjtWrV9OoUSMWLFjAnj17UEpx7tw5AB599FEef/xx2rdvz9GjR+nRowe7d+++6LpDhgyhRYsWfPrpp5jNZubOnctPP/1EUFAQf//9Ny4uLuzfv5/hw4cTGRl52fhffvll2rdvz4svvshvv/2W20gCzJgxAz8/P9LT0wkLC2PgwIG89dZbTJ48maioqEvq+vnnn4mKimLr1q3Ex8cTFhZGx44dL3v/7du3L9J3LYQQooyTtlna5nJEElshhMjn/BPOv/76i23btjFv3jwAEhMT2b9/P05OToSHhxMcHAwY822io6OJiIjAxcWF0aNHc9ttt+U+qV26dCm7du3KrT8pKYnk5GQ8PT1zP6tYsSKNGzfmn3/+oUKFCjg6OtKkSRMSExMZP348UVFRmEwm9u3bd8XYV6xYwc8//wzAbbfdhq+vb+6xTz75hAULFgBw7Ngx9u/fj7+//2XrWrVqFcOHD8dkMlGhQgU6derExo0b8fLyKvD+b6bGUwghROmSttkgbfPlSWIrhCgbivj0tqQcOnQIk8lEUFAQWms+/fRTevTocdE5y5cvx9nZOfe9yWQiJycHs9nMhg0b+Oeff5gzZw6TJ0/m33//xWq1snbtWlxdXS+qp0ePHpw+fZrQ0FCmT5+eO+SpQoUKDB8+HIAPP/yQChUqsHXrVqxWKy4uLle9B6XUJZ8tX76cpUuXsnbtWtzc3OjcuTMZGRlXrOdKQ5gKun8hhBA3GGmbpW0uR2SOrRBC2MTFxTF27FjGjx+PUooePXowZcoUsrOzAdi3bx+pqamXLZ+SkkJiYiK9evXio48+yh1C1L17dyZPnpx73vnP//zzT6Kiopg+fToAAwcOZMmSJcydO5dhw4YBxpPoSpUq4eDgwHfffYfFYrniPXTs2JFZs2YB8Pvvv5OQkJBbj6+vL25ubuzZs4d169bllnF0dMy9x/x1zZ07F4vFQlxcHCtWrCjUghlCCCFEcZG2+dK6pG0umPTYCiFuaucXZ8jOzsZsNjNy5EieeOIJAEaPHk10dDQtW7ZEa01gYCALFy68bF3Jycn069ePjIwMtNZ8+OGHgDHMaNy4cTRr1oycnBw6duzI1KlTLynv4+NDREQEp0+fpmbNmgA89NBDDBw4kJ9++okuXbrg7u5+xft56aWXGD58OC1btqRTp05Uq1YNgJ49ezJ16lSaNWtG/fr1iYiIyC0zZswYmjVrRsuWLXMbXoABAwawdu1amjdvjlKKd955h4oVK7Jnz57CfblCCCHENZC2Wdrma6HK22pZoaGh+kqTs4UQ5cfu3btp2LChvcMQN5GC/s4ppTZprUPtFNINQdpmIW4c0jaL0lZcbbMMRRZCCCGEEEIIUa5JYiuEEEIIIYQQolyTxFYIIYQQQgghRLkmia0QQgghhBBCiHJNElshhBBCCCGEEOWaJLZCCCGEEEIIIco1SWyFEDe1U6dOMWzYMGrXrk2jRo3o1asX+/btK/Hr1qhRg/j4+CueM3PmTE6cOJH7fvTo0ezatatY45g5cybjx4+/6LMBAwYQEhJCnTp18Pb2JiQkhJCQENasWVOoOtu2bVusMQohhLi5SNssbfO1MNs7ACGEsBetNQMGDOCuu+5izpw5AERFRXH69Gnq1atn5+iMhq1JkyZUrlwZgOnTp5fKdRcsWADA8uXLee+99/j1118vOp6Tk4PZfPnmo7CNrBBCCJGftM0Fk7b56qTHVghx01q2bBmOjo6MHTs297OQkBDat2/PU089RZMmTWjatClz584FjMakc+fODBo0iAYNGjBixAi01vz+++8MGTIkt47ly5fTp08fAGbPnk3Tpk1p0qQJEyZMuCSG6OhomjRpkvv+vffeY9KkScybN4/IyEhGjBhBSEgI6enpdO7cmcjIyCvW6+HhwXPPPUfz5s2JiIjg9OnTACxevJjWrVvTokULunXrlvt5Yc2cOZPBgwfTp08funfvTkpKCl27dqVly5Y0bdqUX3755aIYrvR9CSGEEJcjbXPhSdt8MemxFUKUCW9veJs9Z/cUa50N/BowIfzSBuu8HTt20KpVq0s+//nnn4mKimLr1q3Ex8cTFhZGx44dAdiyZQs7d+6kcuXKtGvXjtWrV3PrrbfywAMPkJqairu7O3PnzmXo0KGcOHGCCRMmsGnTJnx9fenevTsLFy6kf//+V4190KBBTJ48mffee4/Q0NCLjl2p3tTUVCIiInj99dd5+umnmTZtGs8//zzt27dn3bp1KKWYPn0677zzDu+//36Rvs+1a9eybds2/Pz8yMnJYcGCBXh5eREfH09ERAR9+/ZFKXVRmYK+r/bt2xfpukIIIexD2uaLSdtctkmPrRBC5LNq1SqGDx+OyWSiQoUKdOrUiY0bNwIQHh5OcHAwDg4OhISEEB0djdlspmfPnixevJicnBx+++03+vXrx8aNG+ncuTOBgYGYzWZGjBjBihUrrju+K9Xr5ORE7969AWjVqhXR0dEAxMTE0KNHD5o2bcq7777Lzp07i3zdW2+9FT8/P8AYKjZx4kSaNWtGt27dOH78eIFPmgv6voQQQoiikra5YNI2XyA9tkKIMuFKT29LSuPGjZk3b94ln19pSI6zs3PuzyaTiZycHACGDh3KZ599hp+fH2FhYXh6ehZqaI/ZbMZqtea+z8jIuGqZK9Xr6OiY+2Q2b3wPP/wwTzzxBH379mX58uVMmjTponIWiyX3CXnfvn155ZVXLqnb3d099+dZs2YRFxfHpk2bcHR0pEaNGgXGfrnvSwghRNknbbNB2ubyQXpshRA3rVtuuYXMzEymTZuW+9nGjRvx9fVl7ty5WCwW4uLiWLFiBeHh4Vesq3PnzmzevJlp06YxdOhQAFq3bs1///1HfHw8FouF2bNn06lTp4vKVahQgdjYWM6cOUNmZuZFi0F4enqSnJx8ybUKU29+iYmJVKlSBYBvvvnmkuMmk4moqCiioqIKbDgLqi8oKAhHR0eWLVvGkSNHrlpGCCGEuBppmy+QtrlopMdWCHHTUkqxYMECHnvsMd566y1cXFyoUaMGH330ESkpKTRv3hylFO+88w4VK1Zkz57LzzMymUz07t2bmTNn5jZOlSpV4s0336RLly5orenVqxf9+vW7qJyjoyMvvvgirVu3pmbNmjRo0CD32N13383YsWNxdXVl7dq1uZ8Xpt78Jk2axODBg6lSpQoREREcPnz4Wr6yXCNGjKBPnz6EhoYSEhJyUdxCCCHEtZK2+drd7G2zKm+rYIWGhurzK48JIcq33bt307BhQ3uHIW4iBf2dU0pt0lqHXqaIKARpm4W4cUjbLEpbcbXNMhRZCCGEEEIIIUS5JomtEEIIIYQQQohyTRJbIYQQQgghhBDlmiS2QgghxE1CKVVVKbVMKbVbKbVTKfVoAed0VkolKqWibK8X7RGrEEIIURSyKrIQQghx88gB/k9rvVkp5QlsUkr9rbXele+8lVrr3naITwghhLgm0mMrhBBC3CS01ie11pttPycDu4Eq9o1KCCGEuH6S2AohbmqnTp1i2LBh1K5dm0aNGtGrVy/27dtX4tetUaMG8fHxVzxn5syZnDhxIvf96NGj2bUrf8fa9Zk5cybjx4+/6LPly5fTpk2biz7LycmhQoUKnDx5ssB6li9fTu/eRgffokWLeOuttwo8z8PD44rxnDt3js8//zz3/YkTJxg0aNBV70MUnVKqBtACWF/A4TZKqa1Kqd+VUo0vU36MUipSKRUZFxdXkqEKIW4y0jZL23wtJLEVQty0tNYMGDCAzp07c/DgQXbt2sUbb7zB6dOn7R0acGnjOX36dBo1alTi1+3YsSMxMTFER0fnfrZ06VKaNGlCpUqVrlq+b9++PPPMM9d07fyNZ+XKlZk3b9411SUuTynlAcwHHtNaJ+U7vBmorrVuDnwKLCyoDq31l1rrUK11aGBgYInGK4S4eUjbXDBpm69OElshxE1r2bJlODo6Mnbs2NzPQkJCaN++PU899RRNmjShadOmzJ07FzCefHbu3JlBgwbRoEEDRowYgdaa33//nSFDhuTWsXz5cvr06QPA7Nmzadq0KU2aNGHChAmXxBAdHU2TJk1y37/33ntMmjSJefPmERkZyYgRIwgJCSE9PZ3OnTsTGRl5xXo9PDx47rnnaN68OREREbm/CCxevJjWrVvTokULunXrdsVfEBwcHBg8eHDufQPMmTOH4cOHs2HDBtq2bUuLFi1o27Yte/fuvaR83ifNhw8fpk2bNoSFhfHCCy/knpOSkkLXrl1p2bIlTZs25ZdffgHgmWee4eDBg4SEhPDUU09d9P1kZGRwzz330LRpU1q0aMGyZctyr3f77bfTs2dP6taty9NPP33ZexOglHLESGpnaa1/zn9ca52ktU6x/bwEcFRKBZRymEKIm5S0zQWTtvnqZPEoIUSZcOqNN8jcvadY63Ru2ICKEyde9viOHTto1arVJZ///PPPREVFsXXrVuLj4wkLC6Njx44AbNmyhZ07d1K5cmXatWvH6tWrufXWW3nggQdITU3F3d2duXPnMnToUE6cOMGECRPYtGkTvr6+dO/enYULF9K/f/+rxj5o0CAmT57Me++9R2ho6EXHrlRvamoqERERvP766zz99NNMmzaN559/nvbt27Nu3TqUUkyfPp133nmH999//7LXHz58OGPGjGHChAlkZmayZMkSPvzwQ0wmEytWrMBsNrN06VImTpzI/PnzL1vPo48+yoMPPsioUaP47LPPcj93cXFhwYIFeHl5ER8fT0REBH379uWtt95ix44dREVFAVz0ZPp8+e3bt7Nnzx66d++eOzQtKiqKLVu24OzsTP369Xn44YepWrXqVb/nm41SSgFfAbu11h9c5pyKwGmttVZKhWM8BD9TimEKIcoIaZsvJm1zFFB222bpsRVCiHxWrVrF8OHDMZlMVKhQgU6dOrFx40YAwsPDCQ4OxsHBgZCQEKKjozGbzfTs2ZPFixeTk5PDb7/9Rr9+/di4cSOdO3cmMDAQs9nMiBEjWLFixXXHd6V6nZyccufTtGrVKrfxiYmJoUePHjRt2pR3332XnTt3XvEaYWFhpKSksHfvXn7//XciIiLw9fUlMTGRwYMH06RJEx5//PGr1rN69WqGDx8OwMiRI3M/11ozceJEmjVrRrdu3Th+/PhVh5mtWrUqt44GDRpQvXr13Maza9eueHt74+LiQqNGjThy5MgV67qJtQNGArfk2c6nl1JqrFLqfPfIIGCHUmor8AkwTGut7RWwEEKAtM0gbfPVSI+tEKJMuNLT25LSuHHjAueIXOl3eGdn59yfTSYTOTk5AAwdOpTPPvsMPz8/wsLC8PT0vGI955nNZqxWa+77jIyMq5a5Ur2Ojo4YnXIXx/fwww/zxBNP0LdvX5YvX86kSZMuKmexWHKfkPft25dXXnmFYcOGMWfOHHbv3p3bAL7wwgt06dKFBQsWEB0dTefOna8a7/l48po1axZxcXFs2rQJR0dHatSocdV7v5b/LuJiWutVwKX/QS4+ZzIwuXQiEkKUZdI2G6Rtvrb7Lu22WXpshRA3rVtuuYXMzEymTZuW+9nGjRvx9fVl7ty5WCwW4uLiWLFiBeHh4Vesq3PnzmzevJlp06YxdOhQAFq3bs1///1HfHw8FouF2bNn06lTp4vKVahQgdjYWM6cOUNmZia//vpr7jFPT0+Sk5MvuVZh6s0vMTGRKlWMXV2++eabS46bTCaioqKIiorilVdeAYwhT99//z3//vsvffv2vaSemTNnXvGaAO3atWPOnDmA0WDmjScoKAhHR0eWLVuW+xT3cvcMxsIZ5+vYt28fR48epX79+leNQQghRPkhbfMF0jYXTYkmtkqpnkqpvUqpA0qpS5bhUkp5K6UW27YU2KmUuqck4xFCiLyUUixYsIC///6b2rVr07hxYyZNmsQdd9xBs2bNaN68ObfccgvvvPMOFStWvGJdJpOJ3r178/vvv+cON6pUqRJvvvkmXbp0oXnz5rRs2ZJ+/fpdVM7R0ZEXX3yR1q1b07t3bxo0aJB77O6772bs2LG5C1ScV5h685s0aRKDBw+mQ4cOBAQUbh2gRo0a4ebmxi233IK7uzsATz/9NM8++yzt2rXDYrFctY6PP/6Yzz77jLCwMBITE3M/HzFiBJGRkYSGhjJr1qzc+/b396ddu3Y0adKEp5566qK6HnroISwWC02bNmXo0KHMnDnzoqfBQgghyj9pm69M2ubLUyU1bUYpZQL2AbcCMcBGYLjWeleecyYC3lrrCUqpQGAvUFFrnXW5ekNDQ/X5lceEEOXb7t27adiwob3DEDeRgv7OKaU2aa1DL1NEFIK0zULcOKRtFqWtuNrmkuyxDQcOaK0P2RLVOUD+xxYa8LSt0ugBnAVkYpQQQgghhBBCiEIrycS2CnAsz/sY22d5TQYaAieA7cCjWmsrQgghhBBCCCFEIZVkYlvQqov5xz33AKKAykAIMFkp5XVJRUqNUUpFKqUi4+LiijtOIYQQQgghhBDlWEkmtjFA3h14gzF6ZvO6B/hZGw4Ah4EG+c5Ba/2l1jpUax0aGBhYYgELIUqfbI8pSov8XRNCiMKRfy9FaSnOv2slmdhuBOoqpWoqpZyAYcCifOccBboCKKUqAPWBQyUYkxCiDHFxceHMmTPSgIoSp7XmzJkzuLi42DsUIYQo06RtFqWluNtmc7HUUgCtdY5SajzwJ2ACZmitdyqlxtqOTwVeBWYqpbZjDF2eoLWOL6mYhBBlS3BwMDExMcgUA1EaXFxcCA4OtncYQghRpknbLEpTcbbNJZbYAmitlwBL8n02Nc/PJ4DuJRmDEKLscnR0pGbNmvYOQwghhBA20jaL8qokhyILIYQQQgghhBAlThJbIYQQQgghhBDlmiS2QgghhBBCCCHKtRKdYyuEEOLGZE1LIyc2lpy4OLJjY8mJjSMnNhadmQkmE8pkApMDysH2p8mc+16ZTeBgQpkcjD/zvjeZL3xucrhQl4MDymw2/szzXjnkOcdkusx784W68r83mez9VQohhBCiGEhiK4QQIpc1PZ2cuLjcpDUnNvZC4nr+89hYrCkpl5RVzs44uLigrVawWNBWK9piAYsFrFY73I0QQgghbhaS2AohxE3Ampl5UWJqJKqxFxLXuDhyYuOwJiVdUlY5OWEODMQcFIRz3bq4t2uHOSgQc2AgjkFBmG0vB09PlFIFXl9rnZvsYrGgLVaw5OR5b8l3/MJ7nZNjJMb53ueeY0ucL7y3gtWCzrEYfxbwXltywGJFWy0wblxJf/1CCCGEKGGS2AohRDlmzcrCkm848EUJbJzxuSUx8dLCjo44BhoJqnPNWri3jjCSVFsSaw4yElcHb+/LJqyFpZQCs5nrq6WESGIrhBBClHuS2AohRBmks7LIiY/PM4f10uHAObGxWM6du7Sw2Wwkp4GBOFavjltYmC1hNZLV8z2sJm9vY87pTSrbYiUhLcveYQghhBCiGEhiK4QQpURnZZFz9iw58WfIiY/DcuaM8fOZeCzx8bafz5ATH4+1oB5WkwlzQICRsAYH49qyBeagIGM4cGCehNXX96ZLWC1WTWJ6NmdTs0hIyzL+TM3ibFoW8SkpnE5N4Ez6Wc5mnCMp6xyp2Ylk6mSUKdXeoQshhBCiGEhiK4QQ1+G6k1XAwc0NU2AAZv8AnGvVwi08DLN/wIXeVdtcVpOf302xiq/WmuTMHCMxzU1Us0lIzSI+NY3YlATi0s6SkJnAucxzpOQkkmFJAodUlDkNZUpFmc7/mYoy2XplFeBqewHOgIuDh53uUgghhBDFSRJbIYTI57LJanw8ljOFTFbd3TEF+Ocmq+6twzH5G+/NAf6YAwIwBQRg9vfHwdW1lO+wdKVnWTiblpUvUc3ibEoGp1MTiEtL4Ex6AolZCSRnJZJmSUI7pOZJTNNQ5vM/Z1yo2Nn2ApwAJwc33M3eeDt54+NclQA3X4Lc/Al098PH2QdfZ198XC786eXkhdnBjBpVJmf+CiGEEKIIJLEVQtwUcpPVuHijN/V6k9XatS9OVgONJPVGT1YtVk2CLUk9k3rhzzMpmZxOOUdc+lnOpCVwLiuB5KwkUnPOkYMtKTXn7UlNM15KGxU72V6AI2BWzribvfF09MbXpQL+Ln4EuvsR5O53SYLq6+yLj7MPjiZHe30tQgghhLAzSWyFEDccrTXZR4+Sun49aevWk7ZxIzlxcQWem5usBgRenKwGBBo9qzd4snq+N/VsijEf9WxqJvHJRk/q6dQEzqSdJSHjHIlZ50jNTiJDJ4NDWr6e1PNJqm2vWrPt5QYmwBEzbrYk1cfZDz/X2gS6GUmqn4vvRQmqr4sv3s7euJpvvO/6hqZln2IhhBD2JYmtEOKGkH3yJKnr1pO2fj2p69eTc/IkAObAQNxat8apVs0Lw4D9/THZElcHFxc7R158rFZNUkb2RT2p8SnpnExO4HTqWc6kneVMupGkpmQnkm5JIkelXEhMC+pJhYuG/LpgxtXkiYejN95OPvi6VL+QpLr64OdiG/br4pv7p5vZ7bq3CxJlXPw+yM4Axxvn/ychhBDliyS2QohyKSc+3uiRXb+B1PXryD5yFACTjw9urVvjPuZ+I6GtWbPcJlVZOVZjLqrtFZeazsnEeE6lnDWG/KbbFk/KTiItJ5FMnWIsoJQnScWUfnGSaiJ3ASUzZjxMXniYvfF08sHPpQb+rr5UcPejgru/0YtqS1DPv9wd3cvt9ylKUHY6/Psq9Hjd3pEIIYS4SUliK4QoFyznzpG6caMxtHjDejL3HwDAwcMDt7Aw/O64A7eICJzr1i3zW91orTmXlsWBM3EcOnuaI+fiOJEUT2zqWc5knCUxM4HUnCRySMk33Df90socjZcDjng5eOFm9sLLyRcf51r4u/oQ5O5PRXe/3AWUfFx8chdScjW7SpIqioV2C4C1k6Fud6jVyd7hCCGEuAlJYiuEKJMsKamkb4okdd16UtevI3P3HtAa5eqKW8uWePXti3tEBC4NG6LM9v+nLD0nnYSMBM6mJ3AkMZYjCbGcSD7D6dQzxoq/mQmk5CSRaU0ih2TIP9z3PAdQro64KiNJ9XT0x9u5Fn4uvgS6+VLRw5+KHv74u/nmLprk4+Ijc1KFXR3I8MLiVwXTwofgwdXg6mPvkIQQQtxk7P/boBBCANaMDNK3bDHmya5bR/qOHWCxoBwdcW3RgoDx43CPiMC1aVOUk1OJxpJjzeFc5jnOZZwjITOBhAxjyO+Z9LMcT47ndMoZ2zDgBFJzEsmwJmMlq8C6tFZgccekPXB28MTXMRhvJx/8XH0Jcvejilcg1XwCqeETRJCbf/lNUo+uhxXvQGoc+NYA35rgV/PCn15VwOHG34P3ZpVpgdedHuOFhMdQvz8Nt39p75CEEELcZCSxFULYhc7KIn37dlLXrSNt3XrSo6LQ2dlgMuHatCn+o0fjHtEa1xYtrnuBp7TsNOLT40nITLgoWc19n5HAmYwLc1bTcpIvH7fFGW1xN1457jjoGribffBy8sbfxZdAd3+qeAVQzTuQWn4VqOnvT5CHKw4ON+iQ35Nb4d/XYP9f4B4EFZvCqR2wZwlYsy+cZ3ICn2p5Et4aF//sWA6TeZGrko8rM6L96Vb3Xtpumwb1ekKT2+0dlhBCiJuIJLZCiFKhc3LI2LXrwhY8mzej09NBKVwaNsR35EjcW4fj2ioUk4d70erWmvj0eI4lH7voFZMcQ0xKDGczzhZYTmkzSntgzXEjJ9sNneOHtgTnJq4uDp65+6dW9gigqk8AwT6eVPB2oZK3C5W8XPFyNd+c81Tj9sKy12HXL+DiA90mQfgYcLL9t7NaIDEGEg7D2cOQEH3h52PrITPp4vo8K13ay3s++XXzg5vxOy5H/N2duCU0mFGRHdhUaR3evz4O1SLAq7K9QxNCCHGTkMRWCFEitNVK5r59xvY7tr1krSkpADjXrYPPwIG4R7TGLTQUk4/PVevLtmRzPOU4MSkxlySvx5JjyLRk5J6rUDjhj8rxJzO9DkHpOZzLqcC5nAq5Saufiy8Vvbyo5OVGpUAXKnq7UNHLSFgr2l5uTvJP5CXOHob/3oZtc8HRDTpNgDbjwMX74vMcTOBb3XjV6nzxMa0h7WyepNeW+J49DAf/heSTF5/v7GUkuRclvTLEuax5pV8T9p5KZkTcfSxyfAaHX8bBnT/LQwkhhBClQn5rE0IUC601WYcPX0hk16/Hcu4cAI7Vq+HVq5eRyIaHYw4IKLCO5KzkfAnrhT9PpZ7CijX3XAeccLQGYMnyIz0tFEuWP9YsP6xZ/pitvlTycWaY0yr6pP5MQFYMVmUioUYvcsIexLdeG5zMZXvl5DIn6QT89w5s+Q4czEYy2+5xcPcvel1KGeXc/SE49NLjWWlw7siFpPd8j+/pnZcOcXZwNIY4F5T0yhDnUuXiaGLKna3o82k6n5ju4rGDX8DG6RB+v71DE0IIcROQxFYIcc2yYmIuJLLr1pETFweAuVIlPDp3xi2iNe6tW+NYqRIAVm0lLi2OY6ciLySwKTHEJMdwNOkYiVnnLqrfjKet17UC2RkNsGb7o7P8sWb7EeAaQDU/d6r5uVG1pitV/dyo5udGddd0gvZ8h8PGaZBwBiq3hNYv4HBqO/6bv4UfF0PV1hDxEDToDSb5Z/CKUuNh5QdGgqKt0Opu6PAkeFUquWs6uUFQQ+OVn9UCScfzJb22xPfYRshMvPh8z0oFL2YlQ5xLRGUfVz4b0ZIR07Po5r2Zxn+9gKrZCQLr2Ts0IYQQNzildQHbTZRhoaGhOjIy0t5hCHHD0FpDTg7WzCx0ViY6MxNrRgY6KwudaXuf55glJYX0qCjS1q0n+/hxAEwBAbiHh+MW0RqnsFac9lW5Q4bP97geTT7G8eTjZFkzc6+tcMBk9cOS5Udmui/WLD90ttHr6kogVX39chPWqr6uVPM3fg72dcPFMd/w07OHYO1nsGUW5KQbi9e0fQSqt72QvGQmw5bvYd0Uo0fQpxqEPwAtR146lPZml34O1nxqfFc56dB8uDHs2Le6vSO7PK0hPaHgpPfsYUg+cfH5tiHO6sFVm7TWBXQdi8LK3zZ/vfowUxav5j+PibgG1YL7/gaTox0jFEIIUZ4opYrcNktiK0QZoC0WdFbWFRNKa2Ym+qLk0/hTZ+U5lpmJNevCz7nvMzIvc8z4Gav16kHmoby8sLZoSELjKkTX8WKfdyrHkmM4knSUuPRYNBf+XXHQzqgcf7IyfLHYhgpbs/1QOf5UdK9EdT9Pqvq6UdXPLTeJrebnhq+bY+EWZYqJhNUfw+7Fxi/OzYZAm4chqMHly1gtsHcJrP0cjq4BJ09ocSe0fsDozbuZZabA+qmw5hPISITGA6DzxBujxy07HRKOXJz0nj2MGjlfEtvrlL9t1lrzfz9uJWPbz3zu+LHxUKTLRDtGKIQQojyRxFaIMkprTc6pU2QeOEDmgYNkHthP5oEDZEUfwZqWBtnZV6/kSpRCubjg4OSEcnFBOTvj4OyEcnJGOTujnJ1wyP3Z9t7ZJc/PzrZzbT+ff+/kRLw1iejM4xxKi2FvajR7Ug4R7Z6MzpN0Oli9sGT5kpPplzvP1Zrtj4+5IsE+QVT3czd6XPMkr5W8XTCbrnGeq9UK+/+E1Z8YiamLN4TeZySmnhWLVteJLUaCu/NnY6ht/V7G/NFqbW6uYarZGRA5A1Z9YOxFW68ndHkOKjWzd2Ql7loaT3GxgtrmjGwLA6es4YEz79DHYTXq3j+hapidIhRCCFGeSGIrhJ1dSGAP2pLY/WQdOEjmwYO5KwIDmPz9ca5TB6eaNTB5eqFcnPMkl5cmmA7OTrbPXfL8fOEY5uvfcsaqrRxJOsKuM7vYfWY32+J2sOfsHtItqbabM2HNrEhORiWsmUHobD/MlkCqeART3c/XNkTYSF6r+RvDhT2ci3n+anaGsRrv2skQvw+8qxpzZVuOBGfP66s76QRsmAabvjaGs1ZuYdTdeMCNPYTSkm0Mz17xrjF3tWZHuOUFqBpu78hKjSS21+9ybXNMQhrDP/mTeTxFoI8nDg+uurAllBBCCHEZktgKUUq01uScPk3m/gNGAnvwAFn7DxScwNaujXOdOjjXrWMks3XqYPb1tWP0kGPNIToxml1nd7EzfhdRp3dwIHEfWdZ04wRtxpJRCUtGZawZVfCgBg0C6tC4oj/1K3pSK9Cdqr5uBHo6l84erukJsPErWP8FpMZCxWbQ7lFo1K/4k86sNNg625hbema/sfhQ+P3Q6h5jsaEbhdUC2+fB8jeNIbnBYUZCW6uTvSMrdZLYXr8rtc2rD8QzecbXzHJ6HdXqHlSfD0s5OiGEEOXNtbTNshyoEFdQ6ATWzw/nOnXw7tsX57p1cKpdG+e6de2ewAJkW7M5dO4Qu87sYsvpHWyN3cnRlP3k6CwAtNURa0YlLBkhkFmFYPd6NAmsS6OavjSo5EXDSp4EepRSAptfwhFY9zls/g6yU6F2V2j3CNTsVHLDhJ3cIOw+I5E9sBTWfQb/vAL/vQshw41e3IC6JXPt0qC1MR952esQtwcqNIXhc6Fej5tr6LUoNe3qBLCr50Cm/7WFMZtmQP3/Qb3u9g5LCCHEDUYSWyHIl8AeNJLYKyewfXCqY/TAOtepg9mvbPTkZVmy2H9uPzvidrH++DZ2ndnFyfTDWDHm8GqLE5bMyljTw3CjOnW8G9C8Yj0aVfKmYSUvagd6lI39XU9EGYsX7VxoJFtNB0Ob8VCxSenF4OBg/PJdr7uxf+q6z40VlyNnQN3uRoJbq3P5SQa1hgP/wL+vwsko8K8Lg76GRv2NexWiBI3uUJMnjj3Cnr3bqPnzQzg/vP7a9kAWQgghLkOGIoubSm4Cm3cBp/NzYJOTc88z+fkZQ4jr1imTCSxARk4G+xL2sfHEdjae2Ma+c3uIzzqCxgKAtrhgyagMmcEEOdemgV9DWlWqS6PKPtSv6Emgp7Od7yAfrY0e0jWfwOEVxkrFoXdD6wfBu4q9ozOkxEHkV8aerqlxENQYIh40Em9HF3tHd3nRq42E9uhaY4ujTs9As6Gyh6+NDEW+foVpm9Oycvi/T2fxSfLjZNXugfudP5SfB0NCCCFKlcyxFcJGa01ObKxtCPF+sg4etPXG5ktgfX1z57861a6Nc526ONctWwksQFp2Gjvj97Dy6Ba2nN7J4aS9JObEgDK26dE5blgyquCiq1LVrR5NAhsRHlyXhpW8qRXojuO1rj5cGnKyYMc8Y8/U2F3gWdlIFlvdVXb3ls3OMGJe+znE7gT3QGNV5rD7wCPI3tFdcHwT/PsaHPwXPCpCp6egxSgwO9k7sjJFEtvrV9i2+eiZNOZ/+hSP8z2ZfT7DudWdpRCdEEKI8kYSW3FT0tnZZOzaRfrWrWTu33/FBNapjm0hpzKawAIkZyaz/sR2Vh6JYmf8LmLS9pNqPQnK+H/VmuOBzqyCr7kmtTzr07JiU8KDa9Gwsjd+7uUoYclIhMivjT1Tk08avZ9tH4YmA8tP4qU1HP7PSHD3/wkmJ2g6BNo8BBUa2y+u0zth2Ruw51dw9YMOT0DYaHB0tV9MZZgkttevKG3zij2ncPmhH03NR3F5eC3Kt0bJBieEEKLckcWjxE3BmppK+tatpEVuIm3TJtK3bkVnZAAXEliv3rddSGDr1MbsXzbncsWnJbD04BY2nNjG3rO7OZV5kCx1Ove4NdsLR0tVKjmHUt+7AeFVmtO6ag1qBXpc+x6w9pYYY6w4vOkbyEo2FoLqOxnqdC1/wxKVMubZ1uoM8fuN+9o6G6K+N+6rzTioc2vpzWE9c9BIaHfMN7Y/6vKc0ft9vVshCVGMOjaoyHdt3qXhuiGc/OYeKj+yFBxM9g5LCCFEOSeJrSjzcs6cMRLYTZtJ27SJjN27wWIBBwecG9THZ/Bg3Fq1xLVFSxwrlKFhoPlordmXcIhf96xl9fFIjqTsIsvh5IXj2T64U50abh1oFNCIdlWbE1G9Bj5u5aT38mpO7TCGG++YZ/R0Nh5g9NBWDrF3ZMUjoC70/gBueR42zTT2xP1hiLFIU8RYaD685PbvPHcM/nsbon4AszO0fwzaPnJjbU8kyjSrthbp/Dt7duC76IcZdfodDi9+h5r9ni2hyIQQQtwsZCiyKFO01mTHxNh6YyNJ37SZrMOHAVBOTrg2a4ZraCvcWoXi2iIEk4eHnSO+vPScdLbF7uCPg+vYeGIzMel7sChjhWVtccVN16auVxNaVWpG5xohNK8cjMmhnPVYXo3WcGi5sSDUwX/B0R1ajjJ6EX2r2zu6kmXJNlZ1XvcZnNgCLj4Qeg+EjwGvysVzjeTTsPJ92PS18T70Xmj/BHhWKJ76bxI301BkpVRV4FugImAFvtRaf5zvHAV8DPQC0oC7tdabr1Svbx1ffWrvKZxNhV+ULjUjmy3v9SY8J5Kzd/xJxXo3xX8CIYQQhSBzbEW5oy0WMvftI23T5txENic2FgAHLy/cWrbEtVVL3FqF4tKkMQ5OZbf38nTqabbERvHP4fVsOR1FbOYhtDJWKLZkBuJJbRr5NadbzXB6NWyOr1sZW5W4OJ1P6tZ8DKe2g0cFaP2AkXi52n9v31KlNRxdZyS4e34D5WD0Vkc8BFVaXludaWdh9cew4UvIyYQWI6Dj0+BTtXhjv0ncZIltJaCS1nqzUsoT2AT011rvynNOL+BhjMS2NfCx1rr1lep1remqJ82dxITwCUWK58ixo7h/1Z5kky8V/28trm5uRb0lIYQQNyCZYyvKPGtmJhnbtxs9sps3kb55S+4+seaKFXELC8MttBWuLVvhXLcOqozur5ljzWFfwj6iYqNYeXQjW+O3kpwTB4C2mrGkV8VT3UKzgOb0qNOaW+rWxN/jBk5kz8tMhs3fGnNNE49BQH1j/myzIcYQ2ZuRUlC9jfFKiIb1X8Dm72D7T1CtjZHgNritcHMMM5KM73btZOO7bjoIOj8L/rVL/DbEjUFrfRI4afs5WSm1G6gC7MpzWj/gW208+V6nlPJRSlWylS2Qv4s/3+/+nvZV2tOuSrtCx1O9ajW2dnyH5ise4J/pj3PLw1NR5W2uvRBCiDJBEltRoixJSaRv2ZK70FPG9u3o7GwAnOrUxqtXL9xCW+HWqhXmypXL7C80iZmJbIvbRlRsFOtObGb32R1ka2PBKmu2F5b06njSnhZBIXSv05IOdStSwasM72ta3JJOGqsbR34NmYlQvR30eg/qdi+9hZPKA98a0PNNIxnd8p3xnf04EnyqQ+ux0OJOcPG6tFx2ujFnd9WHkH4WGvSGLhPtu/KyKPeUUjWAFsD6fIeqAMfyvI+xfXZRYquUGgOMAahWrRp1fOrw/Ornmd93Pn4uhZ/f3fyWYWzf9yddTs7l91+70avP4KLfjBBCiJueDEUWxSr79GnSIiNzF3rK3LfPGIppNuPSuBFurUKNhZ5atsTsWzaHpGqtOZJ0hKi4KKJio4g8tYUjyYdsBx2wZFTCkl4Nd12H8Eot6FK7Pm3rBFDVrwhD6HKyjCG6p7YCytiGxewCjm7g6AJmV+Oz8y+zy4U/y1LyH7vHWBBq21zQFmjYB9o+CsGt7B1Z+WC1GFvyrP0cjq0DZy9oMdIYtu1b3fh7svkbWPEepJyC2rcYi1NVke+3ON1MQ5HPU0p5AP8Br2utf8537DfgTa31Ktv7f4CntdabLldfaGionvXXLIb/Npz2VdrzcZePi/Sg0pqRTNx7rcnJzuT4sKWEN6x5TfclhBDixiBDkUWp0lqTdfiwLZHdRNqmzWTHxACg3NxwCwnBc/w4Y6GnZk1xKKNzpzJyMth5ZidRsVFExUWx5XQUiVnnAFBWV7LTqmFJuxVXXZuIyiG0DwmmbW1/agW4F/4Xt7SzcGwDHFtvvI5vgpyMawvY7Gokv45uFxLevMmvo+tVzikgeb7c5wUNj9UajqyG1Z8Ye7eaXaHV3cberX61ru2eblYOJmjUz3gd32QMM97wBayfAvX+B6e3w7mjxpDlQTOgRuGHeApxOUopR2A+MCt/UmsTA+SdsB0MnLhavfX96vNoy0d5L/I95u+fz6B6gwodk4OLJx7Dv8Llu15s/vExTjw6h8o+su+yEEKIwpMeW1FoOjubjN27L1royZKQAIDJzw+3Vq1y58e6NGyAMpfN5yaxabG5SezW2K3sOrOLHJ0DgMkSREZyVSzp1XHJqUV41Qa0rR1E29r+1K/giUNhVi3W2thP9Ng6Y9GgYxsgfq9xzMEMFZtBtQio2tpYPMjBEXLSjeGm2Rl5fk43kt/stEs/z063vc/I93OarUze8ulgzb62L8vBMU/Ca0t+LVlw9iC4BRgr/IaNBveyuU9wuZR4HDZOM+Yq+1SDLs+Xzz1+y5GbqcfWtuLxN8BZrfVjlznnNmA8FxaP+kRrHX6les+3zVZt5YG/H2Br3FZ+7P0jNbxrFCm+s7++hF/kR7zpOZHHH3kSF0fZ31YIIW5GsiqyKFbWtDTSt269sNBT1FZ0ejoAjtWq4daypZHItmqFU40aZXJ+bI41h/0J+3OHFW+N28rxlOMAmHDEnFOd5MQq5KRVxym7JqHVqtK2dgBta/vTuLIXZlMh5odmZxjbuRxbd6FXNu2McczFx0hgq4YbyWzlluBkh55rS07hEuZLkuK0SxNmSzbU627sy+ooPSqi/LvJEtv2wEpgO8Z2PwATgWoAWuuptuR3MtATY7ufe7TWV2x487bNsWmx3L7odoI9gvmu13c4OjgWPkBLNomTO2E5e4RP63/Li8NvKZNtixBCiJIlQ5HFdcs+cYLEX38jeelSMnbuBIsFlMK5QQN8Bg60zY9thWOFIHuHWqCkrKTcRZ6i4qLYHredtJw0AFwdfDFl1STrbAuyU6thzg6mRbUA2jYLoG0df5oH++BkLkQimxJrJK9H1xl/noi60CPqVxvq9bQls60hoF7ZWDzJZAaTJzh72jsSIYQd2ebNXjFTtK2GPO5arxHkFsSkNpN4fPnjTImawiMtHyl8YZMj3nfMJGdKezrufpnv19ZiZFuZbyuEEOLqJLEVWJKSSP7rLxIXLSZtwwYAXJo3w3/0aKNHNiQEk2fZTYhOpZ5i0cFF/BH9BwcSDqDRKBzwc6yOS2Y4ibGVyEytSprFj2bBPtzezJ+2tQNoVd336sPcrFaI23OhN/boOkg4bBwzOUPlFsbc0vOJrHtAyd+wEEKUcd2qd+P2urczfft02lZuS2jFIjx0D6yHQ/dX6fLH07yw5BM2Vn6BsBqFX2VZCCHEzUmGIt+kdFYWKatWkbhoMSn//ovOysKpRg28+vbBu08fnKpWvXoldpRlyWLZsWUsOLCAtSfWYtVWKjs3wppel+MnK5CWUgWlnWlUyYu2tY1ENqymHx7OV3mWk5VqLOJzdL0tmd1obF8DxpzS83Njq0VApeY3796sQtxAbqahyCWloLY5LTuNwYsHk23NZl7feXg5FbCV1eVYrWR/ezs50WsYYXqPzx8ZQkXvm2gLNSGEuMnJUGRxRVpr0qOiSFy0iOQlv2NJTMTk54fPkCF49+uLS5MmZX4u056ze1iwfwG/Hf6NxMxE/F2CqGHqw+79Ddib4Uu9Ch4MbuxPm9oBRNTyw8fN6coVJh6/uDf21HZj2xqAwIbQZMCF3li/WrKAjxBCFJKboxtvdniTUb+P4vV1r/N2x7cLX9jBAcfbP8fhswgmZXzMuO9r8MMD7XA2y2JSQgghCiaJ7U0gKzqaxEWLSVy8mOxjx1DOznh27Yp3v764t22LcizCwh52kJiZyG+HfmPhgYXsPrsbRwdHmvm2JyEthK1bAnF1dGRYaDD3tqtJjQD3y1dkyYHYnXl6YzdA4jHjmKObsTdo+8eN3tjgUHAtm/vsCiFEedEssBkPNn+QyVGT6RDcgd61ehe+sFdlTH0+otm8e2h34hsmLfLlzdubllywQgghyjVJbG9QOWfPkrTkdxIXLyJj6zZQCvc2EQQ89BCet3bD5OFh7xCvyGK1sO7kOhYcWMC/R/8l25pNfd8G9Kz4IFG7a7J8p5VAT2ee6lGDEa2rFdwzm5EIMRsv9MYe3wRZKcYxz8pQrTW0GW+sWFyxKZjKdoIvhBDl0eimo1l9YjWvr3udFkEtqOJRpfCFm9wOe3/nse3z6b+xOT9U8eaO1tVKLlghhBDlliS2NxBrejrJ//5L0qLFpKxaBRYLzg0aEPTUU3j1vg3HChXsHeJVHUs6xsKDC/nlwC+cTjuNt7M3fWsNhOQwftvkQGRyJg0quvPuoJr0Dal88bC0hCO2lYptvbGndwIalANUaGJsT3N+jqx3sAwrFkKIUmByMPFG+zcYtHgQE1dOZEaPGZgcijCkuNe7qCOr+SLtC7ovqkr9ip60qi4jaoQQQlxMEttyTlsspG3YQOKixST/9RfW1FTMFSrgf8/dePXpi0v9evYO8arSstNYenQpC/YvIPJ0JA7KgTaV23BXg0fYfaAqc/44RUZ2Nh3rBfLBkJq0rxNgzAXOTIYDK+DAP3DwH0iINip08oSqYdCwr9EbGxwq29wIIYQdBXsG81zr55i4aiJf7fiKMc3GFL6wqw+q/xQqfduXl11/5MHvPfn14fYEecliUkIIIS6QxLacyti7l8RFi0j69TdyTp/GwcMDz5498O7TF7ewUJSpbC+wobVmW/w2FuxfwB/Rf5CanUpVz6o83OJhqjl1ZP6GFF785zSODqfoF1KZ0R1qUT/IHU5thZXfwMF/jT1krTng6A41O0LEOKjeFoIaQlF6A4QQQpS43rV6szJmJVOiptC2cluaBDQpfOFanSBiHLev+4zfLc15aJYbP9wfUbi9x4UQQtwUZLufciT71CmSfvuNxF8WkblvH5jNeHTogHffPnh06YKDS9l/eh2fHs/ig4tZcGABhxMP42p25dbqt9K3Vn9Ox1bmq1WH2RqTiI+bI3e2rs7dzV0JOLXa6JE9uAzS4o2KKjaDOl2hdldjaLH5KqsfCyHEZch2P9evsG1zUlYSAxcNxNnkzI+9f8TN0a3wF8nOgC87kZF8hohzr9Enogmv9i9CciyEEKLckO1+bkCWlBSS//yLxMWLSVu/HrTGtXlzKrzwPF69emH2LfvzjLKt2ayMWcmCAwtYGbMSi7YQEhjCy21fpm3FW/g16ixPfBvN8XNR1PN3YlqHVDqbtuN46BVYu92oxD3wQiJbuwt4BNn3poQQQhSZl5MXb7R/g/v+vI93Nr7DpLaTCl/Y0QVu/xKXaV2ZXWku/1vnQdNgb4aElu1914UQQpQOSWzLIJ2dTcqqVSQtXkzyP/+iMzNxrFaNgHHj8O7TG6fq1e0dYqEcPHeQBfsXsPjQYs5mnCXANYBRjUfRv05/XHQlZq6J5qXv1xGQFcN9AQfoU3M3AfEbUBvTwMHRWOip60tGQluhKTjIkDMhhCjvwiqGcW+Te/lqx1d0qNKBrtW7Fr5wpebQZSIN/3mZpyuF8vxCE/UreNK8qk+JxSuEEKJ8kMS2jNBak7FtG4mLFpO0ZAmWhARMPj74DById98+uDRvbiyYVMYlZyXzR/QfLNy/kG3x2zArM52qdmJAnQG0q9KO3SdSmbp4G8m7v6ODw1aWOe8kQJ2CZMCxFoSMMBLZGu1lwSchhLhBjQsZx5oTa5i0dhJNA5sS5FaEUTjtHoV9fzL29BSWur/P2O83sfjh9gR4OJdcwEIIIco8mWNrZ1lHj5K4aDGJixeRfeQoytkZj1u64N2nLx7t26Gcyv7cUau2sun0JhbsX8DfR/4mw5JBHZ869K/Tn961euPr5EPk2n85vG4RNZPW01Ltx6ysWB3dcajVGercYgwx9qtp71sRQtyEZI7t9buWtvlQ4iGGLh5KywotmdJtCg6qCKNyzh6Gqe1J8W9K6LGHaV7Vj+9Ht8bRJCN7hBDiRlDm5tgqpXoCHwMmYLrW+q0CzukMfAQ4AvFa604lGVNZkJOQQNLvv5O0aDHpUVGgFG7h4QSMeQDP7rdi8iwfPZUnU07yy8FfWHhgIcdTjuPh6EHf2n0ZUHcAjR39yN7/Dye+eQjzqdWEk0w4EOfdEEuTRzDX745D1XAwOdr7NoQQQthBLe9aPBX2FK+ue5VZu2cxstHIwhf2qwk938Jj0Xh+at6ePpta8MaS3bzUp3HJBSyEEKJMK7HEVillAj4DbgVigI1KqUVa6115zvEBPgd6aq2PKqVu2BWBrBkZpCxfTuKixaSsWAE5OTjXrUvQk/+H12234Vipkr1DLJRMSybLji5jwYEFrD2xFo2mdcXWjG/6AF1xwzV6JTlzRqPid+MEuGtvNjmH4d20B806DSDQq4K9b0EIIUQZMbjeYFYeX8lHmz6idaXW1PMtwt7rLe6Evb/TdM/HTGg5g7dXR9Ms2JsBLYJLLmAhhBBlVokNRVZKtQEmaa172N4/C6C1fjPPOQ8BlbXWzxe23vI2FDn7xAniPvuM5D//wpqSgjkoCK/evY15sw0a2Du8QtFas/vsbhbsX8CSw0tIykqiknsl+lfqQF+LE8FHN0L0KshJJ0eZ2Wipz3/WpmRW60LPrl0JrxVQLuYHCyFuTjIU+fpdT9t8NuMst/9yO74uvszpPQdnUxHmyqbEwZQ2aI8K3KneIDImjfkPtqVJFe9rikUIIUTZUNaGIlcBjuV5HwO0zndOPcBRKbUc8AQ+1lp/m78ipdQYYAxAtWrVSiTYkpB98iRHRo4i5+xZvHr0wLtfX9zCw1Emk71DK5RzGef47fBvLNi/gL0Je3FycKKrTwMGZClaH43CYcd7AKR71mC1a3d+OFOXLQ5N6NWqDve1r0mtQA8734EQQoiyzs/Fj9fav8aDSx/ko00fMSF8QuELewRC309Rs4cxLfxvuiZ04YHvjMWk/NzL/hoVQgghik9JJrYFddHl7x42A62AroArsFYptU5rve+iQlp/CXwJxlPhEoi12GXHxnL07nuwJCZS/dtvcW1afjaRP3juIJ9Hfc6yY8vItmbT2MmP5y1e9Dy6B2/LAXDyxFqzI5ur3cMnR6qxPNaNAA8nRnWtwXsR1eWXCSGEEEXSvkp77mhwB9/v/p4OVTrQtkrbwheu/z9oeRduGybz3W0d6fWLhYdnb+abe8Ixy2JSQghx0yjJxDYGyLtrejBwooBz4rXWqUCqUmoF0BzYRzmWc/YsR++9l+y4OKpNn15uktpsazZf7/iaqVun4GLVDE1Np3/CGepnH4PKIdD2MVKCO/H98UC+Xnec00mZ1A3y4O2BNekXUgUXx/LREy2EEKLsebzV42w4tYHnVj/Hz31/xtfFt/CFe7wBh/+jzur/4+3bfuTxXw7xzp97mdirYckFLIQQokwpyUeZG4G6SqmaSiknYBiwKN85vwAdlFJmpZQbxlDl3SUYU4mzJCZy9L7RZB+Loernn+PWsoW9QyqUPWf3MOK3EXy65VNuychh8amzTKh0C/X7ToWnDnJk4G+8lDKAsO/TeOuvQ9QN8mTmPWH89XhHhoZVk6RWCCHEdXExu/BWh7dIzEzkpTUvUaQ1QJw9YMCXkBjDgNjJjGpTnS9XHGLR1vzP04UQQtyoSqzHVmudo5QaD/yJsd3PDK31TqXUWNvxqVrr3UqpP4BtgBVjS6AdJRVTSbOkpHD0/jFkHThA8Oef4R6Rf0px2ZNlyeLLbV/y1fav8Hby5MN0J7qdPQ13LYIqLdl05CzTfj7MX7tOYXJQ9GlemdHta9Gospe9QxdCCHGDqe9Xn0dbPsp7ke8xf/98BtUbVPjC1VpD+ydg5Xu8MLgnu0/68/S8rdQN8qBhJWmzhBDiRldiqyKXlLK6KrI1LY2jY8aQviWK4E8+xrNrV3uHdFU74nfwwuoXOHDuAH2q92DC3nV4x+3Hesc8/kitw7SVh9hy9Bzero6MaF2Nu9rWoIKXi73DFkKIYiWrIl+/4mybrdrKA38/wNa4rfzY+0dqeNcofOGcLPiqGyTGEDdqOb1n7MPZbGLR+Hb4uMn6D0IIUV5cS9ssqyoUA2tmJjHjx5O+eQtV3n2nzCe1GTkZfLDpA0YsGUFSVhKfdXyfNw7txPv0bvSQb3lhqw8PzdrMmZQsXu7bmDXP3MLTPRtIUiuEEKLEOSgHXmv3Gk4mJ55Z+QzZ1uzCFzY7GUOSs1IJ/PcppoxoycnEdB6ZE4XFWr4e5AshhCgaSWyvk87K4vgjj5K6Zi2VXn8dr1697B3SFW2J3cLgxYP5esfXDKgzgIW3/UjH1V/A0bVw+5d8dKQms9YfZUzHWix7sjN3ta2Bu3NJrjEmhBBCXKyCewUmtZnEzjM7mRI1pWiFgxpAt5dh3x+0jF/EK/2asGJfHO//tbdkghVCCFEmSGJ7HXRODseffIqU//6j4qSX8BnQ394hXVZadhpvb3ibu36/i2xrNl/e+iWTIl7A89cn4MBS6PMx3ya34uN/9jMkNJhn/9cAk0NBOzYJIYQQJa9b9W4MqDOA6dunE3mqiMOcw8dArc7wx0SG185meHg1Pl9+kCXbT5ZIrEIIIexPEttrpC0WTjw7keS//iLomQn4Dhtm75Aua8PJDdy+6Ha+3/09wxoM4+e+P9OmUgQsfhR2LYTur7PYfCsvLdpJt4YVeGNAU5SSpFYIIYR9PRP+DFU9qzJx1USSspIKX9DBAfp9DiYzLBjLpN71aFHNhyd/2sq+08klF7AQQgi7kcT2GmirlVOTJpG0eDGBjz2G/9132zukAqVkpfDq2le576/7MCkTX/f4momtJ+JmdoU/n4Mt30HHp1kZOJQnfowirLofk+9oIRvaCyGEKBPcHN14s8ObxKbF8vq614tW2LsK3PYBxGzAed0nTL2zFe7OZsZ8G0liehHm7QohhCgXJIMpIq01p994k3M/zcN/7AMEjH3A3iEVaNXxVQxYNIB5++dxV6O7mNd3HqEVbQuL/fc2rPsMWo9la52HeOC7TdQO9GDaXaGyH60QQogypVlgM8Y2H8uSw0v49dCvRSvcdBA0GQTL36JCym4+H9GSmIR0HpuzBassJiWEEDeUqya2SqneSilJgDGS2rj33yfh++/xu/tuAh991N4hXSIxM5HnVz3Pg0sfxM3sxrf/+5Ynw57E1exqnLD2c1j+JoSM4EDL57h75kb8PZz49t5wvF0d7Ru8EEIIUYDRTUfTIqgFr697neMpx4tW+Lb3wD0Ifh5DWBVXXurTiGV74/ho6b6SCVYIIYRdFCZhHQbsV0q9o5RqWNIBlWXxn33Omelf4TN8GEETni5z81CXHV3GgF8G8OuhX7m/6f381Ocnmgc2v3DC5u/gz2ehYV9Odnqbu76OxOSg+O7e1gTJVj5CCCHKKLODmTfav4FGM3HlRCxWS+ELu/pC/88hfh8sncSdEdUZ3CqYT/49wKf/7Edr6bkVQogbwVUTW631nUAL4CDwtVJqrVJqjFLKs8SjK0POTJ9O/OTJeA8YQMUXXihTSW1CRgJPr3iaR5Y9gq+LLz/c9gOPtHwEJ1Oezeh3LoDFj0DtWzj3v88Z9fVmEtOzmXlPODUC3O0XvBBCCFEIwZ7BPNf6OTbHbmbGjhlFK1y7C7R+ENZPRR1axhu3N2VAiyq8//c+Xl68S4YlCyHEDaBQQ4y11knAfGAOUAkYAGxWSj1cgrGVGWe/+57Y997Hq1cvKr32KsqhbIzM1lrzR/Qf9P+lP38f+ZtxIeOYc9scGvk3uvjE/Uth/v0QHE7agJnc+/02jpxJY9qoUJpU8bZP8EIIIUQR9a7Vm541evJ51OfsiN9RtMLdXoKA+rDwIRwzz/H+4Obc264mM9dE88SPUWRbrCUTtBBCiFJRmDm2fZRSC4B/AUcgXGv9P6A58GQJx2d3CT/+yOnXX8ejW1cqv/0WylQ2FleKT4/nieVP8NR/T1HJvRJze89lbPOxOJryzZM9sgbm3glBDckeNoeHftpL1LFzfDI8hDa1/e0TvBBCCHENlFI8H/E8AW4BPLPyGdKy0wpf2NEVbv8SUuNgyZM4OChe6N2Qp3rUZ2HUCe7/NpK0rJySC14IIUSJKkzX42DgQ611M631u1rrWACtdRpwb4lGZ2eJixZx6qVJuHfoQJUPPkA52n9xJa01iw8upt/CfqyIWcHjrR7n+17fU8+33qUnn4iCH4aCT1WsI37m6V+PsHxvHK/1b0rPJpVKPXYhhBDienk7e/NG+zc4mnSUdza+U7TClUOg87OwYz5sn4dSinFd6vDm7U1ZsS+OO6ev51xaVonELYQQomQVJrF9Cdhw/o1SylUpVQNAa/1PCcVld0l//MmJZ57FLTyc4E8/wcHJ6eqFStip1FOM/3c8E1dNpJZ3Leb1nce9Te7F7GC+9OS4vfD97eDigx65gNf/i2PBluM82b0ed7SuVvrBCyGEEMUkrGIY9zS5h/n75/PPkSL+KtLuMajaGhY/BtGrARgeXo3PR7Rkx/EkhnyxllOJGcUesxBCiJJVmMT2JyDvxBOL7bMbVvKyZRx/8klcQ0Ko+vlnOLjYd8VgrTXz981nwC8D2HByAxPCJjCz50xqetcsuEBCNHzbDxzMMGohU7dk8tWqw9zdtgbjutQp1diFEEKIkjA+ZDwN/Royae0kYtNiC1/QZIZBM8CrEnzXH3YuBKBnk0rMvCeM4wnpDJyyhkNxKSUStxBCiJJRmMTWrLXOHZdj+9n+3ZclJGX1ao4/8iguDRpQ9YupOLjbd8Xg4ynHGfP3GCatnURD/4b83Pdn7mx0JyaHy8z1TT5lJLXZ6TByAXMPOfL2H3vo27wyL/ZuVKZWcxZCCCGulaPJkbc6vkVGTgYvrH4Bqy7C4k/ewXDvn1C5Bfx0N6z/AoC2dQKYM6YNGdkWBk1dy/aYxJIJXgghRLErTGIbp5Tqe/6NUqofEF9yIdlP2saNxIwbj1OtWlSbPg2Tp/12NLJqK7P3zGbALwPYFreNFyJeYHr36VT1qnr5Qmln4dv+kBoPd/7Mn/H+PPvzdjrWC+S9wc1xcJCkVgghxI2jlnctngp7ijUn1jBr96yiFXbzg1G/QIPb4Pen4a8XwGqlabA3P41tg6ujiWFfrmXNgRvyVx4hhLjhFCaxHQtMVEodVUodAyYAD5RsWKUvfetWjj0wFsfKlak24ytMPj52i+VI0hHu+eMe3lj/Bi2DWrKw30KG1B+Cg7rCf67MZPh+IJw9BMNnsz6rBg/P3kKzYB+m3tkSJ3PZ2KJICCGEKE6D6w2mc3BnPtr0EfsS9hWtsKMrDPkWwkbDmk9gwQOQk0WtQA/mP9iWKr6u3P31Rn7ffrJkghdCCFFsrprtaK0Paq0jgEZAI611W631gZIPrfRk7NrF0fvHYAoIoNrXX2P2t882OBarhW92fsPARQPZn7CfV9u9ypRuU6jkcZUVjLPT4YdhcGobDPmGXc4hjP4mkmp+bnx9dxhuTgUsLiWEEELcAJRSvNzuZTydPJmwYgKZlsyiVeBggl7vQdcXYfuPMGsQZCRR0duFHx9oQ5MqXoz7YTOzNxwtmRsQQghRLArVjaeUug14CHhcKfWiUurFkg2r9GTu38/Re+/DwcOd6l/PwLFCkF3iOHjuIKN+H8V7ke/RplIbFvZfSP86/a8+J9aSbcwPOrIaBnzB0YBOjJqxAQ8XM9/eG46v+w07HVoIIW5qSil3pYyhPEqpekqpvkopu+xLp7Ptu/+rn4sfr7Z7lQPnDvDRpo+KXoFS0OH/oP8Uoz39uhckncTHzYnvR7emY71Anv15O58tO4DWutjjF0IIcf2umtgqpaYCQ4GHAYWxr231Eo6rVGQePsyRe+5FOTpS/euvcaxSpdRjyLZmM23bNAYvHszR5KO83eFtPrnlE4LcCpFgWy3GsKl9f0DvD4it0ZuRM9aTY7Xy3X3hVPZxLfkbEEIIYS8rABelVBXgH+AeYKY9Ask6dJD0HTvtcelcHYI7cEeDO/h+9/esOb7m2ioJuQPumGtM6/mqO8Ttw83JzLRRofQPqcy7f+7l1V93Y7VKciuEEGVNYXps22qtRwEJWuuXgTbAFVYwKh+yYmI4es+9YLVSbebXOFUv/Vx979m9jPhtBJ9s+YQuVbuwsN9CetXqVbiVi7WGXx83Npnv9jJJTUZy94yNxCZl8vXdYdQJst/CV0IIIUqF0lqnAbcDn2qtB2BMG7JDJIojI0eS/I99t7d/vNXj1PGpw3OrnyMhI+HaKqnTDe75DXLSYUZ3OLoOR5MDHwwJ4Z52NZix+jD/99NWsi1FWIVZCCFEiStMYnt+l/I0pVRlIBu4zAaq5UP2qVMcvfserOnpVJvxFc61a5fu9S3ZfBb1GcN+HcbptNN80PkD3u/8Pv6uhZzbqzX8/QJs/gY6/B8ZrR/m/m8i2Xc6makjW9Gimm/J3oAQQoiyQCml2gAjgN9sn9llUQWnWrVwrlOHmPEPc/abb+w2XNfF7MJbHd4iMTORl9a8dO1xVG4B9/0Nrn7GFnq7f8XBQfFi70Y82b0eC7YcZ8y3kaRnWYr3BoQQQlyzwiS2i5VSPsC7wGYgGphdgjGVqJy4OI7efQ+Wc+eoNn06Lg0alOr1d8bvZMivQ5i6dSr/q/k/fun3C7dWv7Volax8D9Z8CuFjyOn0HI/M3sL6w2d5f0hzOtULLJnAhRBClDWPAc8CC7TWO5VStYBl9ghEmc1U//YbPLt14/Sbb3H61VfROfaZd1vfrz6PtnyUZceWMX///GuvyK+mkdxWaAI/joQN01BKMf6Wurw+oAnL98Vx51frOZeWVXzBCyGEuGZXTGxti1L8o7U+p7WejzG3toHWulwuHpWTkMDRe+8j+/Rpqn75Ba5Nm5TatTMtmXy46UPuWHIHSZlJTL5lMm90eAMfF5+iVbT+C/j3NWg2DN3zLZ7/ZSd/7TrNS30a0S+k9OcICyGEsA+t9X9a675a67dt7XW81vqRK5VRSs1QSsUqpXZc5nhnpVSiUirK9ip0e+/g6kqVjz/C7757SfhhNsceeghLSmoR76p4jGw0kohKEbyz8R2iE6OvvSJ3f7hrMdTtAUuehKUvg9aMaF2dz+5oyfaYRIZ+sY5TiRlXr0sIIUSJumJiq7W2Au/neZ+ptU4s8ahKgCUpiaP33UfW0aNUnfI5bi1bltq1o2KjGLRoEDN2zGBAnQEs6L+ATlU7XUNFPxibyDfoDf0+4/2/DzBn4zHGd6nDPe3K9ehwIYQQRaSU+kEp5aWUcgd2AXuVUk9dpdhMoOdVzlmptQ6xvV4pUkwODlR46ikqvvIyqavXcGTECLJPlv4esA7KgdfavYaTyYlnVj5DtjX72itzcoOh30PLu2DVB7DwQbBk06tpJb6+J4yYhDQGTlnD4Xj7JPFCCCEMhRmK/JdSaqAq1IpGZZMlJZVj948hc/8Bgj/9BPeIiFK79uKDixn1+yiyLFl8cesXTGo7CS8nr6JXtGsR/DIOanWGQTOYsfYYk5cdYHh4Vf6ve71ij1sIIUSZ10hrnQT0B5YA1YCRVyqgtV4BnC3pwHyHDKHql1+Qffw40UOG2mXF5AruFXipzUvsPLOTKVFTrq8ykxn6fAxdnoOts+GHoZCZTLs6AcweE0F6toVBU9aw43i5fPYvhBA3hMIktk8APwGZSqkkpVSyUiqphOMqNtb0dGLGjiV9xw6qfPA+Hh07ltq1z2ac5a0NbxESFMLP/X6mbeW211bRwX9h/n1QJRSG/cDC7fG88usuejauyGv9mxZuFWUhhBA3GkfbvrX9gV+01tlAcaza1EYptVUp9btSqvG1VuLRrh3Vf5iFcnQ0Vkz+999iCK1obq1+KwPqDGD69ulEnoq8vsqUgk5PQ99P4dBymHkbJJ+mWbAPP41tg4ujiWFfrmPNwfhiiV0IIUTRXDWx1Vp7aq0dtNZOWmsv2/tr6HIsfdbMTGLGjSdt0yYqv/M2XrcWcZGm6/R+5Puk5aQxqc0k3B3dr62So+tgzggIqA8jfmL54VSe/GkrEbX8+GhYCCYHSWqFEOIm9QXGgo7uwAqlVHXgeh88bwaqa62bA58CCy93olJqjFIqUikVGRcXV+A5LvXqUWPuHGPF5HHj7bJi8jPhzxDsGczEVRNJyiqG5/ItR8HwORC/H766FeIPUDvQg/kPtqWStwt3z9jIHztOXf91hBBCFMlVE1ulVMeCXqUR3PXQWVkcf+xxUtesodLrr+N9222lev2Npzay6OAi7m58N7V8al1bJSe3wqwh4FUZRv7M5jjNg99vpn5FT6aNCsXF0VS8QQshhCg3tNafaK2raK17acMRoMt11pmktU6x/bwEo1c44DLnfqm1DtVahwYGXn5FfnNgYL4Vk18r1RWT3RzdeKvDW8SmxfL6uteLp9J63eHuXyEr1Uhuj22korcLP41tQ+MqXjw0axNzNhwtnmsJIYQolMIMRX4qz+sFYDEwqQRjum46J4fjT08gZdkyKr70Ij63DyjV62dbsnlt3WtU8ajCmGZjrq2S+P3w3e3g7AkjF7I/1ZV7Z24kyMuZmfeE4+niWLxBCyGEKFeUUt5KqQ/O95oqpd7H6L29njornl9TQykVjvF7wpnrjfXiFZN/KPUVk5sFNmNs87EsObyEXw/9WjyVVmkF9/0FLt7wTR/Y+zs+bk7MGt2aDnUDeebn7Xy+/IDd9vQVQoibTWGGIvfJ87oVaAKcLvnQro22Wjn53HMk//EHQRMm4Dt8eKnH8M2ubziUeIiJrSfianYtegXnjhobwisFo37hOAGMmrEBR5MD393bmkBP5+IPWgghRHkzA0gGhtheScDXVyqglJoNrAXqK6VilFL3KaXGKqXG2k4ZBOxQSm0FPgGG6WLKzHJXTH7ZPismj246mpDAEF5f9zrHU44XT6X+tY29boMawJw7IPJr3JzMTBsVSt/mlXnnj7289tturFZJboUQoqQVpsc2vxiM5LbM0Vpz6qVJJP6yiMDHHsX/nrtLPYaY5Bimbp1Kt2rd6Bh8DSO2k08bSW1WCoxcyFnXaoz6aj0pmTl8e2841fzdij9oIYQQ5VFtrfVLWutDttfLwBXnvmith2utK2mtHbXWwVrrr7TWU7XWU23HJ2utG2utm2utI7TWa4o7aN+hQ6j6RZ4Vk3eWzorJZgczb3Z4E41m4sqJWKyW4qnYIxDu+hVqd4VfH4Nlb+BkUnw0NIS729bgq1WHefKnrWRbrMVzPSGEEAUqzBzbT5VSn9hek4GVwNaSD61otNacfuNNzv30E/4PPEDA2LFXL1QCMbyx/g1MysSE8AlFryA9Ab4bYCS3I+aR6tuAe2ZuJCYhna/uCqNhpXKxZpcQQojSka6Uan/+jVKqHZBux3gKzaO9sWIyjmaO3Fl6KyYHewbzXOvn2By7ma92fFV8FTt7wPDZ0OJO+O9tWDQeB53DS30a8cSt9fh5y3Ee+G4T6VnFlEwLIYS4RGF6bCOBTbbXWmCC1vrOEo2qiLTWxH3wAQnffYffXXcR+Nijdonjn6P/sPL4SsaFjKOie8WiFc5MgVmD4cx+GDaLrEqhjP1+EzuOJzL5jpaE1/QrmaCFEEKUV2OBz5RS0UqpaGAy8IB9Qyo8l3r1qDl37oUVk7/9tlTmo/au1Zv/1fgfn275lCeWP8GxpGPFU7HJEfpOho5Pw5bvYc4dqOw0Hulal1f7N2HZ3lhGfrWexLTs4rmeEEKIixQmsZ0HfK+1/kZrPQtYp5QqU+Nh46dM4cy06fgMG0rQMxPssq9ranYqb254k/q+9bmj4R1FK5ydAXOGw/HNMHgm1pqdefKnrazcH8+btzfl1kYVSiRmIYQQ5ZfWeqttW55mQDOtdQvgFjuHVSQXVkzuyuk33iyVFZOVUrzS7hUeCnmIVcdX0feXvryz8R0SMxOLo3K45Tno/REcWAoze0NKHCMjqvPp8BZsjTnH0C/Xcjop4/qvJYQQ4iKFSWz/AfKugOQKLC2ZcIruzFcziP/kU7z796fiiy/aJakF+Dzqc+LS4nihzQuYHcyFL2jJhnn3wOEV0H8Kun4vXvl1F4u2nmBCzwYMCa1ackELIYQo92xb9JzfoPUJuwZzDYwVkz/G797SWzHZxezCg80f5LcBv9Gvdj9m7Z5Fr5978e3Ob8m2FEOPaug9MHQWxO42tgM6c5DezSrz9d3hHD2bxsApa4iOL71VoYUQ4mZQmMTW5fyedgC2n8tEj+3ZWbOIffddvHr9j0qvv4ZyuJa1sK7f3rN7mbV7FoPqDaJ5YPPCF7RaYeFDsHcJ9HoPmg/ls2UHmLkmmtHtazK20zXufyuEEOJmZZ+nu9dJOThQ4el8KyafOlXi1w10C2RS20n82PtHmgQ04d3Id+n3Sz/+PvL39Q+LbtAL7loEGYnwVXc4von2dQOYfX8EaVkWBk1dw47jxdBLLIQQAihcYpuqlGp5/o1SqhVlYHGKc/PmcfrV1/Do2pXKb7+NMpnsEodVW3ll3St4O3vzaMsizO3VGpY8Cdt/hK4vQvj9/LD+KO/9tY/bW1RhYq+Gdut9FkIIUW6V631lcldMjokhevCQUlsxub5ffb649QumdJuCs8mZJ5Y/wV1/3MW2uG3XV3HVcGOvWyc3Y1jy/r9pXtWHHx9og5PJgeFfrmPdoeveJlgIIQSFS2wfA35SSq1USq0E5gLjSzSqq0hc/CsnX3gR9w4dqPLhByhHR7vFMn//fLbFbePJ0CfxdvYufMF/XobIr6DdY9Dh//h9+0meX7idLvUDeXtQMxwcJKkVQghxKaVUslIqqYBXMlDZ3vFdL4/27ag++4c8KyYvK7Vrt6/Snp/6/MRLbV7iaNJRRiwZwdP/PX19+94G1IX7loJ/HfhhKGz5njpBHsx/qC0VvF0YNWMDf+4s+d5pIYS40anCDLVRSjkC9TGGOO3RWtttSb8W9erpHxydcAsNpeoXU3FwcbFXKJxJP0OfhX1o4NeAr7p/Vfge1pUfGIlt6L1w2wesOXSGu2dspGmwN9/f1xpXJ/v0PgshxM1IKbVJax1q7zjKs9DQUB0ZGVmsdebExXHswYfI2LmTCs8+g+/IkaU6kik1O5Wvd3zNNzu/waIt3NnwTkY3G42X0zVuvZeZDD+OgoP/QpfnoONTJKRlc8/MjWyLOcebtzdlaFi14r0JIYQop66lbS7MPrbjAHet9Q6t9XbAQyn10LUGeb2yj8Xg2qwZVT//zK5JLcAHmz4gPSed51s/X/jGduN0I6ltOhh6vc+OE0mM+XYTNQLc+OquUElqhRBCCGwrJn/3bamumJyXu6M741uMZ/GAxfyv5v+YuXMmt/18Gz/s/oFs6zU833f2hOFzodkwWPY6/PoYvi4OzBrdmnZ1ApgwfztTlh8slS2PhBDiRlSYocj3a63PnX+jtU4A7i+xiK5CubhQ9csvcHB3t1cIAGw4uYFFBxdxT+N7qOVTyEWets6F356E+r2g/xQOn03nrhkb8HZ15Nt7W+Pj5lSyQQshhBDlyCUrJo8bV+IrJudX0b0ir7d/nbm951LPtx5vbniT23+5nX+P/lv0JNTsBAOmQvsnYNNMmHsn7iqLr+4Ko0/zyrz9xx7eWLIbq1WSWyGEKKrCJLYOKk93pFLKBNgtA3OuVROTp6e9Lg9AtiWb19a/RhWPKoxpNqZwhfb8BgsfhJodYNDXnE61MPKr9Wjg2/vCqeht395nIYQQoizKXTF50iRSV63myJ13lsqKyfk19G/I9O7TmXzLZJRSPLrsUe798152niniAldKQbeXjN0Q9v0B3/bFKTOBj4eGMKpNdaatPMxT87aRbbGWzI0IIcQNqjCJ7Z/Aj0qprkqpW4DZwO8lG9YVlIGVgmfunMnhxMNMbD0RF3MhEtJDy+Gnu6FyCxj2A4k5Ju6asYGE1Cxm3hNG7UCPkg5ZCCGEKNd8hw2l6tSpZB87RvSQoaW2YnJeSik6Ve3E/L7zeb718xxKPMSwX4fx7MpnOZVaxGQ7/H4Y+h2c2g4zuuOQeISX+zbm8W71mL85hrHfbSIj21IyNyKEEDegwiS2E4B/gAeBccA2wLUkgyrLjiUf44ttX3Br9VvpGNzx6gVORMHsO8C/Loz4iQwHN0Z/s5GDcSl8MTKUZsE+JR2yEEIIcUPw6NDeWDHZbCr1FZPzcnRwZGiDofw64Ffua3Iff0X/Re8Fvflk8yekZhdhqHTDPjDqF0iNh+m3ok5u5dFudXm1fxP+3RvLyK/Wk5hut/U6hRCiXLlqYqu1tgLrgENAKNAV2F3CcZVJWmveWP8GJmViQtiEqxew5MCi8eDiDSMXkOPsw/gfNhN5JIEPh4bQvm5AyQcthBBC3EBc6tWj5ty5ONeuTcy4cZz99lu7Lbjk6eTJY60eY/GAxXSr3o1p26fR6+de/Lj3R3KshVzoqlqEsdet2Rlm3gYH/mFkRHU+Hd6CqGPnGPrFWmKTMkr2RoQQ4gZw2cRWKVVPKfWiUmo3MBk4BqC17qK1nlxaAZYlS48uZdXxVYxvMZ4K7hWuXiByhjHEqOebaI8gnv15O0t3x/JKvyb0blbutxoUQggh7OKSFZNfe71UV0zOr7JHZd7q8Bazb5tNDa8avLruVQYtGsSKmBWFS7oD68N9f4NvDfhhCETNpnezysy4O4yjZ9MYNHUtR86U7qJZQghR3lypx3YPRu9sH611e631p8BNO9kjNTuVt9a/RQO/BgxvMPzqBVJi4d/XoFYXaNSPt//Yy0+bYnisW11GRlQv+YCFEEKIG9hFKybPmmWXFZPzaxLQhJk9Z/JR54/ItmYz7p9xjPl7DHvP7r16Ya9KcM8SqN4WFo6FlR/QoU4AP9wfQXJGNgOnrGXnicSSvwkhhCinrpTYDgROAcuUUtOUUl0B+6/cZCeTt0wmLj2OFyJewOxgvnqBpZMgOw16vcu0lYeZ+t9BRkZU59GudUs8ViGEEOJmUFZWTL4oJqXoWr0rC/st5JnwZ9h9djeDFw/mhdUvEJsWe+XCLt4wYr6x1/0/L8OSpwip4slPY9vgaFIMnrqWl37Zwf7TyaVzM0IIUY5cNrHVWi/QWg8FGgDLgceBCkqpKUqp7qUUX5mw+8xuftjzA4PrDaZZYLOrFzi6DqJmQdvx/H7Sg9eX7Oa2ppWY1Lcxqgys6iyEEELcSMrCisn5OZocGdFwBL8N+I27Gt/Fb4d+o/eC3nwW9Rlp2WmXL2h2ggFfQttHYOM0+HEUdXzN/PxQW3o0rsjsDce49cMVDP9yHUu2n5RtgYQQwkYVZcEFpZQfMBgYqrW+pcSiuoLQ0FAdGRlZatezaisjl4wkJiWGRf0X4e3sfeUClhz4sjOkJ5B472q6Tt5IZR9XfhrbBmezqVRiFkIIUXhKqU1a61B7x1GelXbbfDkZe/dx7MGxWBLOUeX99/G8pYu9Q8p1LPkYH2/+mD+j/yTQNZDxLcbTr3Y/TA5X+N1g3RT441mo2hqGzwY3P+JTMvkx8hiz1h3l+Ll0Kng5c0d4dYaHVyXIqxBbEAohRDlwLW1zYbb7yaW1Pqu1/sJeSa09zNs3j23x23gy9MmrJ7VgLBh1ejv0fIO3/j1GQlo2b97eVJJaIYQQooS51K9HjTlz8qyY/J29Q8pV1bMq73V6j+/+9x2VPSrz0pqXGPLrENacWHP5QhEPwuCv4cRmmNEDzh0lwMOZhzrXYcXTXZg2KpT6Fb34cOk+2r71L+N+2Mz6Q2fstkq0EELYU5F6bMuC0nwqHJ8eT9+FfWno15Dp3adffRhxSix8GgrBrYhsP51BX6zj/g41ee62RqUSrxBCiKKTHtvrV1Z6bM+zpqVxYsIEkv9eiu+IEVR49hmUuRDrY5QSrTV/HfmLDzd9yPGU47Sr0o4nWz1JHd86BReIXgWz7wAHB2NRyuBQCA6Dis3A0YXD8anMWneEHyOPkZSRQ/0KntzZpjoDWlTBw7ns3LcQQhTWtbTNkthewbMrn+XP6D+Z33c+Nb1rXr3Aggdh+09kPbCG22adJC3Lwl+Pd8RdGhUhhCizJLG9fmUtsQXQFgux773P2a+/xr1TR6q8/wEmD3d7h3WRLEsWs/fM5ottX5CancrtdW9nXMg4AlwL2Oc+djf89zYc2whJMcZnDo5QsYmR5FYJJaNCCxYddeHb9UfYcTwJD2czt7eswsiI6tSt4Fm6NyeEENdBEttitOHkBu776z7GNBvDwy0evnqBo+uMYULtn+Az0wje/XMvX90VSteGhdjvVgghhN1IYnv9ymJie17CnDmcevU1nOvWperUKThWrGjvkC5xLuMcX2z7gjl75uBkcuLeJvcyqvEoXM2uBRdIOgnHIyEmEo5vguObIdu21ZGrH7pKK056NmFxfGWmHfIl3uJGRC0/RrWpwa2NKuBoKtJMNCGEKHWS2BaTLEsWAxcNJMeaw4J+C3AxX2UxhjwLRh0Zvpzun0XStWEQn49oVaJxCiGEuH6S2F6/spzYAqSsXMXxxx7Dwd2d4Cmf49q4sb1DKtCRpCN8tOkjlh5dSpBbEI+0eIQ+tfvgoK6SiFpyIG6PLdndCDGbjPcYv+MluNZgXWYNVmfW5KhrI0LD2zM0ohYVZLEpIUQZVeYSW6VUT+BjwARM11q/dZnzwoB1GKstz7tSnaXReH6x9QsmR01mSrcptK/S/uoF1n8Bvz+NHvwNo9ZVYsvRc/zzf52kwRBCiHJAEtvrV9YTW7CtmDx2LJbERKq8916ZWjE5v02nN/HexvfYcWYHDf0a8mTok4RXCi9aJRlJxqJTMUbPrj4eiUqNAyBdO7FD1yTRrzlVm3WkXsvOKO9gkC0JhRBlRJlKbJVSJmAfcCsQA2wEhmutdxVw3t9ABjDD3ontsaRjDFg0gE7BnXi/8/tXL5BnwaiFjT/lsR+38kq/xoxqU6PEYhRCCFF8JLG9fuUhsQXIjo0l5qFxZOzaReBjj+EzaCBmPz97h1Ugq7by++Hf+Xjzx5xMPUnn4M483upxavnUurYKtYZzRyFmI4kH1pF8YC2BqXtxJhuANOdAHKuH41gtzJizWykEnD2K74aEEKIIylpi2waYpLXuYXv/LIDW+s185z0GZANhwK/2TGy11jz4z4NExUbxS79fqOBeiPmxtgWjku5dSZevY6jq58b8B9ticpCnnkIIUR5IYnv9yktiCxevmIxSuDRpgkeH9ri374Brs6ZlavVkgExLJt/v+p7p26eTkp1Ci6AW9KjRg1ur30qQW9B11Z2ens6q1cs4tHk5Qck7aOlwkOrqlHFQOUBQI2MF5iq2VZgD6hkrMwshRAkra4ntIKCn1nq07f1IoLXWenyec6oAPwC3AF9xmcRWKTUGGANQrVq1VkeOHCmRmP+M/pMn/3uSCWETuLPRnVcvcGQtfN0T2j/BM4kD+GlTDIvHt6dRZa8SiU8IIUTxk8T2+pWnxBaMB9kZO3aQsnIlqStXkb51K1itOHh54d62rS3RbY9jhbKzAOTZjLPM2zePP6P/ZF/CPhSKlhVa5ia5Ba6kXEhaa7bGJPLt2mhWb9tLQ+t+evsdp6PbEQKTdqAyEo0Tnb2gcgsjyT2f8HoEFtMdCiHEBWUtsR0M9MiX2IZrrR/Oc85PwPta63VKqZnYscc2JSuFfgv74e/qzw+3/YDZ4SpPbC058GUnSD9HZO8/GTRjKw90rMWzvRoWe2xCCCFKjiS216+8Jbb5WRITSV27NjfRzYmNBcC5Xj3cO7THo0MHXFu2xMHJyc6RGg4lHuLP6D/5K/ovDpw7gEIRWjGUnjV60rVaV/xd/a+57rOpWfwYeYzv1x0hJiGdCh6OjG0KAwJP4nN2q7E41emdoC1GAZ/qF/bVrRIKlZqB2bmY7lQIcbMqa4ntVYciK6UOA+fH7AYAacAYrfXCy9VbUo3n2xveZtbuWczqNYumgU2vXsC2YFT2oG/5318+pGdZ+PuJjrg5la0hTEIIIa5MEtvrV94T27y01mTu20/qqpWkrFxF2qZNkJ2NcnPDvXXr3ETXqWpVe4cKwIGEA/x15C/+iP6Dw4mHcVAOhFUMo0eNHnSr1g1fF99rqtdi1fy3L5bv1h5h+b44HJSiR+MK3BlRnTZVXVEnt9q2G7JtO5R03Cjo4Ggkt+eHLwe3At+asjCVEKJIylpia8ZYPKorcBxj8ag7tNY7L3P+TOzUY7vrzC6G/zacwfUG83zE81cvkBILn7aC4FA+rfQ27y/dz9d3h9Glwf+3d+fxbV513vc/R5ZkLd7iJYntLM6+Oc3uNmmTNEDpNnSBYWmBYYChlK0wMEBh7uF+7udmoDDMPMwMhbZsBYa2w9JCBtqGUqC0JVub1UmapdlrJ/G+yrIlneePS5btxKntOJYs+/t+vfSSdEm6/LOb5uTr37nOGd61LiIiknwKtsM3loLt+WJtbbRt3eYE3T8/T9fp0wB4p08nuHYtWevWEli1Cpf/InvOJom1lsONh9l0fBObjm/iRPMJMkwGFZMruGHGDbxh6hvI8+Vd0rlP1LXxyNaT/PdLp2hs72L2xCzee9V03rq8lGyfx3lTc1WvoPuysyJzV7vzWqDQCblT4wtTlSzXwlQi8rpGVbAFMMbcBHwTZ7ufH1hr/9kYczeAtfaB8977MCkIttFYlPc8+R6q2qr4n9v/hxzvIK6PfeJu2PsLTt3xB974o9e4buEk7r9z+WWrSUREkkfBdvjGcrDtzVpL14kTtD7/Aq0vPE/71m3Yjg6M10tg1aqebu7MmZgUdiittRxqOMSm45t4+vjTnGo5hdu4ubLkSq6ffj1vmPYGcjNzh3zejq4o/7O7iv/acoLdp5sIeDO4fVkpf7O6jHmTs/u+ORqBmgPxfXVfglPboO6w85pxwaRF8Y5uhXNfMEtdXRFJGHXBdiRc7sHzv1/5b7689cvct/Y+bp5588AfiC8YZa/5DO85/mb2nGri2c+sZ6L2rBURSUsKtsM3XoLt+WLhMO0vvURbPOh2HnkVAHdJMVnXrCW49hqCq1eTkZW67qS1lgP1BxKd3NdaX8PtcrOmZA3Xl13PhqkbyPZmD3yi8+w+1chPtpxg4+4qOiMxKsryee/q6Vy/aDJe90VWTm6vh9dedkLu6e3O43Cz85o/Px50453d0hWQOfS6RGRsULAdotpQLbc8cQsLCxfy3eu+O/BvV3stGLVx7RPc88vD/N/bynnvVdMvSz0iIpJ8CrbDN16D7fm6qqpofeEF2p5/gbbNm4m1toLbTWDpUmfa8tpryJw/H5OiLXOsteyv258IuVVtVXhcHq4uuZo3l72ZDVM3kOUdWghvaOvk5y+f4r+2nORkfTtF2ZncsWoqd1w5jeLcAaZnx6JQczDe1d0Gp7ZD7cH4i8bZbqh7+vKUCiiYre2GRMYJBdshuvf5e/nd8d/xy1t+yYzcGQN/YMsD8PTnab31h6z7TTbTCwL88u41uLRnrYhI2lKwHT4F2wvZri5Cu3Ylpi2H9x8AIKOwkKyrrya4di3Bq9fgnnBpizsNuz5r2Vu7NxFyz7afxevyck3pNVxfdj3rp64n6AkO+nyxmOW5wzX8ZPMJ/njwHC5juG6Bs9hUxYz8i3dxzxdq7FmQ6tQ25z4c327Ilxdfgbmip6vrG/qUahEZ/RRsh2BL9RY+9LsPcfeSu/nY0o8N/IHEglGr+JzvSzy+s4rf3HMN8ydrz1oRkXSmYDt8CrYDi9TU0Prii04398UXiTY2gjH4Fi8m65prCK69Bv8VV2AyMpJeW8zG2FOzx9lC6MTvONd+jsyMTNaWruX6sutZN2UdAU9g0Oc7WdfOT7ed4GfbT9HQ3oXfk8HKsglcNbOAq2YWcMWUXDwZgwy6sZhzbW739OXT2+HcAcACBormO2F3aoUTeAvnqqsrMgYo2A5SZ7STt218G1Eb5YlbnyAzYxD7rT1xN1T+kl1veYrbHjvL3etnce+N84dVh4iIpJ6C7fAp2A6NjUbp2LcvsW9uaM8eiMVw5eYSXLM6cX2uZ2Lyd1uI2Ri7zu1KhNzaUC2+DB/rpqzj+rLrWTtlLX734FaA7uiK8qeD59j8ah1bjtZz8GwLAAFvBivL8rlqZj6rZxZQXjqEoAvQ0QSv7XBCbnfg7Wh0XsvMdbYY6l6UasoK8KemKy4il07BdpAe2P0A9++6nwfe9ABXl1498AdO/AV+eCORqz/N9XvWE47EeObv1+P3Jv+3qiIicnkp2A6fgu3wRBsbadu8mdbnX6Dt+eeJ1NQAkDl/PllrryF4zVoCy5ZivN7k1hWLsvPcTp4+/jTPnHiG+o56/G4/66es54ayG7i69Gp87sEvnlnXGmbbsXo2H61jy9E6Dp1tBSCYCLoFrJ5VQHlJDu6hBF1roe5I36B7bj/YmPN64dye6ctTVjldXpf+DScyminYDsLJ5pPc/uvb2TBtA99Y/42BPxCNwIPrINzM/Qsf4V/+eIqH37+Ka+dpz1oRkbFAwXb4FGwvH2st4UOHaHv+eVqff4H2HTugqwtXIEDgqqsIVKwiWFFB5rx5SZ22HI1Fefnsy2w6vonfn/w99R31BNwBrp16LdeXXc/VpVcPbgZcL7WtYbYerWdLPOgePtcTdFfNcLq5V80sYNFQgy5AuKWnq9sdeEP1zmve7HhXt3u7oZUQyB/a+UVkRCnYDsBay92/v5vdNbvZeNtGJgYGEU7jC0adufF7rNuYxfXlk/nPO5Zd0tcXEZHRR8F2+BRsR060tY32bVudacsv/oWukycBcGVnE1ixgkBFBYFVq/AtmI9xu5NSUyQWYfuZ7Ww6volnTz5LY7iRLE8WG6Zu4Pqy61ldshpvxtC7yzUtYbYeq4sH3XqOxINuVqabVWUTWD2rO+jmkjHUhTuthfqjfbu6Z/eBjTqvF8zuCbmlKyBrImTmgDeo/XVFUkDBdgBPH3+azz73We6tuJd3L3j3wB9oOQvfWomdsoo72z9LZXWzs2dttvasFREZKxRsh0/BNnm6zp6lfdt22rdto337djqPHwfAlZWFf8VygqtWEaiowLdwYVKCblesi+3V29l0YhO/P/F7mjubyfZks2FaPOQWr8aT4bmkc59r6ejT0X21pg2A7Ew3FTPyE4tRLSzJGXrQBehsg6qdPUH31DZor+37HuNyOryZ8Zsvp+dxZrYTfvvcn3e8+/2eoBa1EhkCBdvX0drZyi2/uoVCfyGP3vwoGYO5tuLxD8O+x9m07ld8+KlG/vn2ct59pfasFREZSxRsh0/BNnW6zp2jfft257ZtO51HjwLgCgTwr1hBYNUqAqtW4i8vx3guLWAOupZoF1uqt7Dp+Cb+cPIPtHS1kO3JZnHRYhYVLGJx4WLKC8spChRd0vnPNXew5VhP0D3aHXR9bq7sFXQXFF9i0LUWGo5D9S4INTjTmXvfOpouPBZuga62QZzcXCQQXyQIXyw4e7MUkFOh/hhs/x7segTyZ8C1X4TZb1Q3fwQp2L6O+7bdxyMHHuGRmx+hvLB84A/EF4wKXfX3rNm2hhmFQX6hPWtFRMYcBdvhU7AdPSI1NbS/9FIi7IYPHwHABAIEli51pi5XrHKC7gguRtUV7WJz9Wb+eOqPVNZWcrjhMNH4tN9JgUmUF5YnbosKFpHtzR7y1zjb3JGYtrzlaB3Hap2AmeNzUzGjwFl1eVYBCybnjOy/36IR6Own8F4sCIcvcryzdXBfr78Osi8XZl8Hi24H7+C3ZpLXYS28+gfY9hAc2uQsODb3BqjeA00nnanrG74IM69VwB0BCrYXsa9uH3f+9k7eMfcd/ONV/zjwB3otGPWFku/z8911/PaetcybPPS/dEVEZHRTsB0+BdvRK1JfT/v2lxJTl8OHDgFgfD78y5YSWOUsRuW74gpcIxh0Q5EQB+sPsrd2L5W1lVTWVnKy5WTi9bKcMhYXLmZRodPZnZc/b8iLUZ1p6khco7v51TqO17UDkOv3UNFrMar5k7NHZ6MiFr1IEG6O384/1gId8fuWM9B82unsLn47rHgfFC9J9XeUnsItsPsxJ9DWHoJgEax4P6x8P+SUQKQTdv0X/Pkb0PwaTFvjBNwZa1Nd+ZiiYNuPaCzKu598N2fazrDx9o3keHMG/tCW78DT93Jw/Xe4flMuH712Fp+7QXvWioiMRQq2w6dgmz4iDQ09Hd1t2wkfPAjWYjIz8S9ZkliMyr90Ca7MoQXLoWoKN7Gvdh+VdZWJwFsbcq5xdbvczJ0wl/KCns7uzNyZg7uULK66KcTWo/XOPrrH6jgRD7p5AQ8VZfmJxajmTRqlQXcorHVmG+74Eez7FUTDULIMlr8PFv+109WV11f3qhNmdz3i/OKgdAVUfBgW3Qbufv5fiIRhx4/h+X+FlmooWwsb/hGmr0566WORgm0/HnvlMf556z/ztbVf46aZNw38gfiCUbEpq3jTmU8QsbDpU+u0Z62IyBilYDt8CrbpK9rURPvLL9O+1enodrzyCsRiGI8nHnSdxaj8S5bg8vtHtBZrLWfbz7Kvdp8TdOsq2Ve7j9YuZ4qu3+1nYcHCxLW65YXllARLMIOcBlrVGGLrMaebu+VoPSfre4Luld0d3VkFzJ2Y5kG3vR72/MwJuef2OwtXlb8VVvytE9Y0bbZHLAZHfg/bHnTuXR5nOveVH3ZWyB6MrhC8/DA8/2/Qdg5mbnA6uFMrRrT0sU7B9jy1oVre8sRbKC8s56HrHhrcX3zxBaN+uORR/s9fwvzoAxWsn3tpixyIiMjoN56CrTHmB8BfAeestRcsOGGcgfLfgZuAduBvrbU7Bjqvgu3YEW1udoJufPpyx/79zj/+PR78V1xBYNVKZ0GqZctwBUb+Ws6YjXGi+URi+nJlbSWv1L9CZ6wTgHxffmJhqkWFiygvLCffN7g9aV9rDLHl1fj2QsfqOFUfAmBCwEN5aS4LinNYUJzNguIcZhVl4RnqXrqpZi2cfgl2PAyVj0NXO0wqd7q4V7wD/HmprjB1Oppg509h+3edbaCyJsPKDzjhP3vSpZ2zsx1e+j688E1nde3Z18GGLzi/TJAhU7A9z+f+/Dl+f+L3PH7L45Tllg38geMvwsM3Ub/iHq7asoYbyifzH9qzVkRkTBtnwXYd0Ar8+CLB9ibgEzjB9krg3621Vw50XgXbsSva2kpoxw7at22jbft2Oir3QTQKbjf+8nIn5FZUEFi+DFcwmJSauqJdHGo8RGVNJZV1Tth9tfFVLM6/aUuzSp2Obnwa88KChQQ8A4fw0w3tbDlaz/Zj9eyvbubg2RY6IzEAvBkuZk/MYmFJTiLwLizOIS8wctclX1YdzbD3504Xt3o3uH2w8DbnWtxpq8dPF/fcK850492POatZT70SKu6CBbeA+zL9twy3OoH5xX93Vteee6MTcHXN85Ao2PayuWozdz1zFx9Z8hE+uvSjA584vmCUDTfxN/5vsftMJ89+5lqKskf2+hIREUmt8RRsAYwxZcBvLhJsHwT+ZK19NP78IHCttbb69c6pYDt+RFvbCO3cGb9GdxuhykqIRCAjA1/5Imcf3VWr8K9YQUZWVtLqautqY3/d/sQ05n11+3it9TUAXMbFzNyZlBeWJzq7c/PmDri/biQa42htGweqm9lf3cyB6hYOVDdT0xJOvKc419cr6OayoDib6QXBS9tuKFmqdjkBd8/PndWcC+c6Xdwld0CwINXVXX6xKBx6GrY+CMeeg4xM57rjirugZOnIfd1wi/M1//Kf0NEI8/8Krv0CTB7E7iyiYNstHA3zto1vw1rL47c+PrhV9eILRr244pu8+8WJfPWti7mjYtplqlpEREYrBds+r/0GuM9a+0L8+bPA5621Fwy8xpi7gLsApk2btuLEiRMjWreMTrH2dtq7g+72lwjt2QNdXeBy4Vu40Am5VyzGt3gxntLSQV8PeznUd9T3mcJcWVtJQ7gBAK/Ly/z8+X22HZqeMx2XGXi6cU1LmAPVzb1uLRypaSUac/5N7fdkMG+yM4V5YXwq8/ziHLIy3SP6/Q5ZZ5szRXnHj+D0dsjwwoK3OCG3bG3675fbXg87/8vpnjaehJxSWPVB5/sLFiavjo4mJ2dsvt9ZlGrhrU7AnbggeTWkIQXbuO/s/g7f3vVtHrzuQdaUrBn4pC1n4Fur6CxZyZXH72bWxGx+9uHV6b1wgIiIDIqCbZ/Xfgt89bxg+zlr7cuvd051bKVbLBQitHs37dviHd09e7CdzvWwGXl5+MrL8S0ux19ejq98MZ5JE5NWm7WWqrYqp6Mb7+zur9tPKOJcW5vtyWZh4ULKC8qZnz+fORPmMD1nOm7XwIE0HIly+GxrvLPbE3ibQl2J90wvCLBgck6fa3enTPAnNexf1Nl9zgq/ux9zuosTZsDyv4Gl7770a05T5ew+p1O652cQCcH0q53FoObdDBkp/OVCqAE2f9sJuZ2tzoJe6++Formpq2kUU7AFTjSf4K2/fitvnPZGvr7+64M76eN3wb4n+ErZD/nhKy5+e89a5k7SsugiIuOBgm2f1zQVWS4r29lJx+HDdOytpGNfJaG9lYQPH3au0wXcRUX4Fi/GV74I/+LF+MrLcU+YkLT6orEoR5uO9nR16yo5VH+IiI0ATmd3Vt4s5kyYw9wJcxP3hf6BO37WWqqbOpypzFXNHDjjhN3jdW10//M72+eOd3Z7wu7cSdn4PCnajaMrBPs3Ol3cEy+Cyw3zboTlfwuzNsAQtltKqmgEDv4Wtj4EJ14Atx+ueLuzXc9om/rbXu9MT976oBO8F78d1n8eCmalurJRZdwHW2stH37mw+yt3cvG2zZSFBjEasbxBaNOlX+MtS9dzcc3zOYfrp93masWEZHRSsG2z2s3Ax+nZ/Go/7DWDrhnhYKtDEWso4OOAwf6hN3OY8foTnue0lJ85eX4F5c7Hd5Fi8jITl7DoTPaybGmYxxqOMShhkMcbjjMoYZD1IRqEu/J9+Uzd8LcPmF3Vt6sQV3+1haOcPBsS0/grW7mlTMttHc6Yd9lYGZRVjzs9ixUVZSdmdzubu1hJ+DuegTa6yB3Kix7Lyx7D+SWJq+O19NW56z6vP0H0HwacqdBxd85dQYGtzp2yrTVOgtMbfsuRDthybtg3Wchf0aqKxsVxn2wffrY03z2z5/lCxVf4M4Fdw58smhXfMGoFm6M/CshvGz61LrU/ZZMRESSbjwFW2PMo8C1QCFwFvjfgAfAWvtAfLufbwE34Gz38/7+rq89n4KtDFe0tZWOffvpqKwkVLmXjsp9dJ06lXjdW1aGb/Fi/OWLnA7v/PlJ2W6ot/qOeg43HE4E3UMNhzjSeIRw1FlMKsNkMD1neiLodt+Kg8UDBtJYzHKyvj0xjXl/fKGq1xpDifcUBL19pjEvLEnSNkSRMLzyWyfkHv0TGBfMebNzreqcN6dmem/VLmd1472/gGgYZqx3phvPvWH0dpUvpuWsE3Bf+j7EIrD0Tlj7DzBheqorS6lxHWxbOlu45Ve3MDEwkUdueoSMwfyh3vxt2PQFfj3/X/jkrlJ+8sEK1s7RnrUiIuPJeAq2I0XBVkZCpKEhHnb3EqqspGNvJZGzZ50XXS4yZ8/u09nNnDcPlze52+9EY1FOtpzsE3YPNRxKrMgMkOXJuiDszs6bTZZ34FWjm9q74lOYe67b7W8bogXFOcwsCjKjMEhZQZDpBQGCI7FYVf0x2PkTZ1Gm1rOQXexch7v8vTCh7PJ/vd6iXXBgozPd+NQW8ASdLmfFXTBx/sh+7WRoroYX/j94+YfO7IXl74W1n4HcKamuLCXGdbD9ytav8Ngrj/HozY+yqHDRwCeKLxjVNmkFS1/9EDcvLuGb79KetSIi442C7fAp2EqydJ07R0flvp7O7t5Kog3OSsd4PPjmzu1ZnGrxYjJnzcK4k99RbOtq6xN2uzu9LV0tifeUZpX2CbxzJsxhevb0AZsz/W1D9Ep1M+d6bUMEMDE7k7LCIGUFgfh9/FYYIOAd5s8k2gWHNjld3CO/d4LYrA1OF3feTZdvT1iA1nPw8sPw0g+gpdpZ2KriLqez6c+7fF9ntGh6DZ7/V2cxL2Ngxd/CNZ+GnOJUV5ZU4zbY7qvdx51P3sk7572TL175xcGd6PG7sPue4J4JD/Dnumye/cx6CrO0Z62IyHijYDt8CraSKtZaIlVVhCr39XR2K/cRa3ECpPH58C1Y0Kez6y0rw6RgKxtrLWfazvTp7B5uOMzx5uNErXN9bWZGJrPyZl1w/W6+b+DrRVvDEY7XtnGirp3jdW0cq23jRF0bx2rbqW3tP/TOKAgyvTDAjIIgZYVOp3fIobfptNPB3fET5zrXYJETOpe/b3gLIp1+GbY9CPuecK5BnfVGZ7rx7OvSfyuiwWg8CX/+Buz6KZgMWPkBuObv02+V6ks0LoNtNBblzifv5Fz7OTbetpFs7yAWF4gvGLV/9oe5qXI9X3vbYt65SnvWioiMRwq2w6dgK6OJjcXoPHEi3tnd64Te/fuxIed6VVcwiG/Roj6d3WTvsdtbOBrmaOPRPgtVHWo4RF1HXeI9hf5CJ+jmzWFuvhN6Z+bOxJsxuM5od+g9XucE32O1bfHnF4beSTmZTC8IJsJu746v3/s63eRYFI4863RxDz4FNursh7v8fc7+uB7fwIVGwrDvV06gfe1l8GY7IbniQ1A4Z1Df65jTcBz+/C+w61Fnr+FVH4SrPwVZY/vyyXEZbB858Ahf3fZVvr7u69w448aBTxBfMCra0cKa5q8yfXIhj911lfasFREZpxRsh0/BVkY7G4kQPnq077ZDr7yC7XL2mU3ssVu+CN+Chc7CVMaAwenuGuMsmjTgc4NxmfjzQXzG5QJ6fSb+HAMNnY0cbTrG0eZjHGl6lSNNRznWfIyw7cQCLlcG03PLmDVhNrPy5zBlQhnFWcWUBEso8BfgMoPrarZ0dCW6vN1htzsE17Z29nnvpJxMygqca3mnFwSZUeiE3un554XeljNOp3HHj51g5p8AS+5wQm5/18M2VztTjV/+IbTVQMGc+HTjOyBTW3ACUPcqPPd12PszcPucn8+aeyBYkOrKRsS4C7Y17TXc8qtbWFy4mAeve3Bwv2mLLxj1/Slf5r5js3jqk2uZPVH/w4iIjFcKtsOnYCvpyHZ20nHocJ+VmHvvsZtuqvLhYKnh4BTD0akeYtOKKc4uTYTdycHJlGSVJB57MjwDnrM79Pae1ux0fS8MvZNzfEwvCDgLWHV3egv8zGh+mcw9P4EDv4FYF0y90gm4i26HM3uc/VwPbHQ6vnOvdwLbzA3jY7rxpag9DM99zVkR2huEK++G1R8b/dsbDdG4C7afe+5zPHvyWR6/9XGm5wxiSeyWM/CfK6kvXM7yo3dxzxvm8Ok3a89aEZHxTMF2+BRsZayIhUKEjx7FdnY6CyJZC7EY1lqIWcA5ZmMxsICN9XpuE7fE8wE/A8RiwIWf6Tnn+e+58Hk41EJz5W5ie/bjamoFIBzwcGqanwOllp2T2jlcDGGv0wQyGIr8RUzOmkxJsCQRfouDxYnHA63a3NzRxcnzpjV3d33r2i4MvYsndHKbeZ41zb9lQvtxYi4vrlgnZOY6KwCv+iDkz7yM/zXHuHOvwHP3OdcgZ+bAVR+Fqz6SXgtqRTohVA/t9X3vQw2YtZ8e8ticgo2nLo+/VP2Fp44/xUeXfHRwoRbgd/+EjYb5RMO7KCsI8tENs0e2SBERERFJGy6/H/+iQeyuMQqV4ixQ1XXiBO07dhLauZOcXTuZ/ewR3gKQ4SI6cwrN80uonpHLoQkuXnU1sr9uP8+efJauWFef82V7sykOFvcJvt1BuCSrhHxfPuWluZSX5l5QS3NHFydq2zlW18aJ2jbnvq6dL9VsoK5tDRXmFW7O2MIhO5Xfcy0TDkygtLqO4rw2SvL8lOb5Kc71U5LnY1KOb+T36k1HE+fD2x929rx97j7ntvU7sPoTziJbvpzk1WIthJt7BdSG/gPr+a93tl7WMtKyY/vi1hd566/fijGGX97ySzIzBrGacXzBqM1TPsAdR97ET//uSq6eXTjyBYuIyKimju3wqWMrMnpFm5oI7d5N+86dhHbuIrRnD7a9HQD3pEn4ly/Dv3Qp4YUzqJkSpLqjhuq2aqpaq5z7tirOtJ7ps1URgNflZXJwck+3N6u4TxCeHOh/unPv0FvVGErcXmvsoLopRGN734DtMjAx20dJno+SPL9zy+31OM/PhIAnZYt/jRrVu+FP98HBJ51rmtfc40zrzhx4v+Q+Ip0QGmQw7dVhJRa5+Dl9ec5UaX9+P/cT+j1uMoPjYyryB773Ab6z+zs8dN1DrC5ZPfCH4gtGdYWaWV7/Za67Ygb/9s6lI16riIiMfgq2w6dgK5I+bCRC+NChRFc3tHMnXVVVgLM9kn/xYvzLluFfthT/0qW4J0wAoKWzJRF2q9uqqW51Qm/3fW2ots/X6Z7ufEHwzSpJ3Ac9wQvqawtHqG4KUdXY0RN8m/o+7ozE+nzG53FRktsddOOh97znPs/r7w88Zry2A/70VTj8OwgUwNWfdFanHmwntbPl4ufOyDwvgE54ncAav/fnwQB7M/dnXFxje8WyK6z7027eNP1NfH3d1wf3oc33w6Yv8rW8L/Fo82Ke/fR6CrRnrYiIoGB7OSjYiqS3rrNnEyG3fecuOvbvh4jTgfPOnIl/2VICy5bhX7YM74wZ/e4D3Bnt5EzbmUTY7dP1ba3iTPsZIud19fxuPwW+AvL9+c69L58Cf0GfY93Pc7w5GGOw1lLX1hkPuj2Bt7qpg9fij2taw5wfcfKDXkryfBTnOlOdux93T30uys4kYyztknJqO/zpK/DqH/p/3Zf7OoH0IoHVE18tPAnGRbAtmltk5/w/c9h4+0YK/YOYShxfMKoqdwlrTn2Er//1Et6xcurIFyoiImlBwXb4FGxFxpZYRwcde/fSvnNXIvBGGxsByMjNxb90abyruwz/4nJne6SBzmlj1IZq+3R9a0O11IXqqO+op66jjrpQHY3hRmI2dsHn3S43+ZlO8E2E3njwzffl93kcdOdS1xpJBN3eobe60XncGo6cd37DpBxfT+jN6w69PQE4x+dOvynPr+1w8lDvgOrLg4zRvdTSpYzNo/s76kdrVyv3LL9ncKEWEgtGfajmHVTMKODtK6aMbIEiIiIiImnM5fMRWLWKwKpVgLMoVefx44R27CS0ayftO3fS+txzzpszMvAtWIB/2TICy5zA6ykuvvCcxsXEwEQmBiaylKUX/drRWJTGcCN1HfHAG+q5733s1cZXqQvVXbDoVbe8zLy+wbe0gBmzerrCma4iIp1ZtIf81DTHEtOfX2sM8fLJBqr3VBOJ9W0AZmW6mZiTSWFWJkXZmRT1ui/M9lKU5aMw20tBMBOve5QseFW6PNUVJE3adWwL5xTaswfPkjGYudrHX4CHb+bpgr/hE2du5KlPrmP2xCFeQC0iImOaOrbDp46tyPgTbWoitGtX30WpQiEA3MXFTshd6nR1ffPnYTwD75s7VNZaWrta+3R960M93d/eneD6jnpau/pfhTfgDvRMgY4H3wmZ+WS6crCRbDo7A4Ta/TS2+mhszaC2rZPaljA1rWFaOvpfNGlCwJMIwIkg3OtxYZaXouxMCoJjbAr0ZTIuOrZluWWDC7XRLnjys3QEp/Cp1zbwkTfOVqgVEREREbkMMnJzyVq/nqz16wFnUaqOVw72XKu7ayfNTz4FgPH7+yxKFVi6lIy8vGHXYIwh25tNtjebstyyAd/fEemgvqO+byc4Hny7Q/HJlpPsqtlFQ0cDlgsbgG7jJicvh9xJuUzJzCPbk4vXFcRDFiYWJBYN0NXpIxTOpL3dR1O7h1On3NS2WEJd0QvO5zLO9b/nd4H7C8N5fg8uheCLSrtgO2jbHoJz+/l/vV+gpDCfj147K9UViYiIiIiMScbtxl++CH/5InjvewDoOnMmviCV09Wt+/73+yxK5Z05A09xCZ7iYjzFk/EUF+MuLsZdWIjJuPyrGPvcPkqynH14BxKJRZwp0edNgW4MN9IUbkrcn2mvoincRFO4iY5ox4Un8ju3rBIfJd5cAu5sfBnZeMjCZQPYSICuSIBw2Ed1yMuh014aWz10dvqwUT/Q83NwuwwF8U5vYVb3FOh+wnBWJjn+NLweeJjGZrBtroY/fpVX89bwyJlyHvlQ+fhZ4ltEREREZBTwTJ6M58YbybnxRgBioRChvXudqcu7dtF14gTtm7cQa2vr+0G3G8+kSbiLJzvBd/JkPCVO6PXEb67s7BENbm6Xm0J/4eDX9cHpCPcOvY3hRpo6ndDb2NHY93n4dCIQR228kxsPwd5C8MbP6c8I4s/IwWOCZNggNhokFPFzpNPH3qpMWtu9RCJ+bDQQvwUhlok3IyMx5Tk/6GVC0MuEgJcJAQ95AedYXsATP+Y8Tve8NDaD7TP/RCzayYfOvYO3LZ/KmlmD/wMpIiIiIiKXn8vvJ1hRQbCios/xaEsLXVXVdFVXETlzxnl8pppIVTWhHTtoPns20elNnCsQwF1S3Df4To4H35Ji3JMn4/J6SSaf24fP7WNScNKgP9N9nXBjuJHmcDON4cZEMO4Ox4lA3NFEY/gkrZ1NtJgWyARvXk8I7mZw4TVZRMnibMxPdSRIpM5PuMpPZ6cfGw0mQrCNOI+J+Qh4PYmQ6wTfXkE44GFCr2MTAk5YDnozRk1neOwF2+MvwN6f88vgHTTYUv7x5gWprkhERERERC4iIzubjHnZ+ObN7fd1G40Sqa0jcqaarurqnuBbXU1X9Rk69u8nWld34XkLC53Qmwi7xYlpz4kpz/3syZtMva8TJnvwn4vEIjR3Ng8ciOPPGzrO0BBuwHWRVaSdMJxNlCwaYkHqokEiTX7C5/x0hH3xANwrEEeDEPPiyXDFg2+vDnCwbye4OwR3B+LcEbpWeGwF22gX/PYfaPWX8k91b+bLb19IfjC5v6kREREREZHLx2Rk4Jk0Ec+kifiXLOn3PbFw2On2xsNuV3VVIviGjx6l9cUXse3tfT/k8eCZNCl+be/kPtf7dk97zsgeQtpMIrfLTb4vn3xf/qA/Y62lPdJOQ0cDjeHGC+/DDTR29L4/QSTchM9euOgVgAs3ma5s3GTTGgvSEglyvN1PuMFPqMNHNNLdFe4Vhq0HYyDP7+k3+HZPk76kn8klfWq02vog1Bzgi3yOpTOLedvy0lRXJCIiIiIiI8yVmYl3+nS806f3+7q1llhzczz4OrdI9ZnE49BLL9N89kmI9g1xrqysXkG3BPfEItyFRbiLCnEXFuIuKCCjsBBXZmYyvs1hMcYQ9AQJeoJMyZ4yqM/EbIyWzpZBBeHGcC0N4QZC4Wa8uf1vKes2mU4YtllEbBY10QDVnQE6W5ww3D1V+lKMnWDbXA1/uo99wat4unEZT92+eNTM9xYRERERkdQxxpCRm0tGbi6++fP7fY8z5bmWrqqebm9Xdc/1vh17K4k2NPT7WVdOjhN0u29FhWQUFjohuLAgcTwjP39EVnweKS7jIjczl9zMXKbn9P9Lg/Mlpkl3nBd++wvHHVU0hhvpzGwlI8dZP+tSjZ1g+8w/EYuE+UjLO/noG2czq0h71oqIiIiIyOA4U54n4Zk0CZYt6/c9trOTSH29c81vbQ3R2loitbVEauL3tbWE9lUSra27cLVnAJeLjPz8C0NwQUE8BBcmusGunJy0bNRdyjTprmhXny7wlVw59K875E+MRseeh70/50fud+AunMlHtGetiIiIiIhcZsbrdRakmjx5wPfG2tuJ1NXFQ28NkdpaJwh3h+C6OsJHXyVaU4vtunBRJ+PxkFFU2BN4Cwp6dYIL+0yJdvmH0+tMPU+Gh6JAEUWBoks+R/oH22gXPPlZGjOLua/pJn5052Iy3enT3hcRERERkbHHFQjgDQTwTp36uu/rvv63b+e3uxtcl5geHdqzx1n92V54/aorGCSjsFfXt1fn12T6ANv9xcBarLXOofhz6H2812sXO37Ba0M9fuFrieOXKP2DbXzBqM93fYZbVszkqpkFqa5IRERERERkUHpf/5s56/VnntpIhGhDQ2Lac58p0fFQHD58mLbNm4k1NyfpO7hMjOm5XYL0DrbN1dg/fZWXvavY7rqKZ2/SnrUiIiIiIjI2Gbcbd1ER7qKBp+zGwmGitbXEwp1gnACdCI19QqSJ35kLjvf53HnBs//jFzmXMZjzv2738f6C7CWE2/QOtr/7X8QinXw6dCf/6+0LmaA9a0VERERERHBlZuIqHT/bn7pSXcAlO/Y8VP6Ch2K3MGXmQm5fNn7+o4mIiIiIiEiP9OzYRrvgyX+g1j2Zb3fcwq9vK0/LpbBFRERERERk+NKzY7v1Aah5hc+3v5sPbVjITO1ZKyIiIiIiMm6lX7CNdmH/9FVedK3geME6Prx+ZqorEhERERERkRRKv2Db/BrRri6+EHoPX7lde9aKiIiIiIiMd+kXbEMN3B95C6tXrORK7VkrIiIiIiIy7qXd4lFdxsN/e97KkzfNT3UpIiIiIiIiMgqMaMfWGHODMeagMeaIMebefl5/tzFmT/z2F2PMkoHOeThWymffspS8gPasFRERERERkREMtsaYDOB+4EZgIXCHMWbheW87Bqy31l4B/F/goYHOm+XP5Lal2rNWREREREREHCPZsa0Ajlhrj1prO4HHgFt7v8Fa+xdrbUP86RZgykAnnZYf0J61IiIiIiIikjCSwbYUONXr+en4sYv5IPDUCNYjIiIiIiIiY9BILh7VX1vV9vtGYzbgBNtrLvL6XcBdANOmTbtc9YmIiIiIiMgYMJId29PA1F7PpwBV57/JGHMF8D3gVmttXX8nstY+ZK1daa1dWVRUNCLFioiIiIiISHoayWC7HZhjjJlhjPEC7wI29n6DMWYa8DjwXmvtoRGsRURERERERMaoEZuKbK2NGGM+DmwCMoAfWGv3GWPujr/+APAloAD4dnxBqIi1duVI1SQiIiIiIiJjz0heY4u19kngyfOOPdDr8d8BfzeSNYiIiIiIiMjYNpJTkUVERERERERGnIKtiIiIiIiIpDUFWxERkXHEGHODMeagMeaIMebefl6/1hjTZIzZFb99KRV1ioiIDMWIXmMrIiIio4cxJgO4H7gOZ1u+7caYjdba/ee99Xlr7V8lvUAREZFLpI6tiIjI+FEBHLHWHrXWdgKPAbemuCYREZFhU7AVEREZP0qBU72en44fO99qY8xuY8xTxphF/Z3IGHOXMeYlY8xLNTU1I1GriIjIoCnYioiIjB+mn2P2vOc7gOnW2iXAfwK/6u9E1tqHrLUrrbUri4qKLm+VIiIiQ6RgKyIiMn6cBqb2ej4FqOr9Bmtts7W2Nf74ScBjjClMXokiIiJDp2ArIiIyfmwH5hhjZhhjvMC7gI2932CMmWyMMfHHFTj/VqhLeqUiIiJDoFWRRURExglrbcQY83FgE5AB/MBau88Yc3f89QeAvwY+YoyJACHgXdba86cri4iIjCoKtiIiIuNIfHrxk+cde6DX428B30p2XSIiIsOhqcgiIiIiIiKS1hRsRUREREREJK0p2IqIiIiIiEhaU7AVERERERGRtKZgKyIiIiIiImlNwVZERERERETSmoKtiIiIiIiIpDUFWxEREREREUlrCrYiIiIiIiKS1hRsRUREREREJK0p2IqIiIiIiEhaU7AVERERERGRtKZgKyIiIiIiImlNwVZERERERETSmoKtiIiIiIiIpDUFWxEREREREUlrCrYiIiIiIiKS1hRsRUREREREJK0p2IqIiIiIiEhaU7AVERERERGRtKZgKyIiIiIiImlNwVZERERERETSmoKtiIiIiIiIpDUFWxEREREREUlrCrYiIiIiIiKS1hRsRUREREREJK0p2IqIiIiIiEhaU7AVERERERGRtKZgKyIiIiIiImlNwVZERERERETSmoKtiIiIiIiIpDUFWxEREREREUlrCrYiIiIiIiKS1hRsRUREREREJK0p2IqIiIiIiEhaU7AVERERERGRtKZgKyIiIiIiImlNwVZERERERETSmoKtiIiIiIiIpDUFWxEREREREUlrCrYiIiIiIiKS1hRsRUREREREJK0p2IqIiIiIiEhaU7AVERERERGRtKZgKyIiIiIiImltRIOtMeYGY8xBY8wRY8y9/bxujDH/EX99jzFm+UjWIyIiMt5pbBYRkbFoxIKtMSYDuB+4EVgI3GGMWXje224E5sRvdwHfGal6RERExjuNzSIiMlaNZMe2AjhirT1qre0EHgNuPe89twI/to4tQJ4xpngEaxIRERnPNDaLiMiYNJLBthQ41ev56fixob5HRERELg+NzSIiMia5R/Dcpp9j9hLegzHmLpzpUABhY0zlMGtLtkKgNtVFDJFqTp50rFs1J4dqTo55qS4giTQ290jHP6uqOXnSsW7VnByqOTmGPDaPZLA9DUzt9XwKUHUJ78Fa+xDwEIAx5iVr7crLW+rIUs3JkY41Q3rWrZqTQzUnhzHmpVTXkEQam+NUc3KkY82QnnWr5uRQzclxKWPzSE5F3g7MMcbMMMZ4gXcBG897z0bgb+IrMF4FNFlrq0ewJhERkfFMY7OIiIxJI9axtdZGjDEfBzYBGcAPrLX7jDF3x19/AHgSuAk4ArQD7x+pekRERMY7jc0iIjJWjeRUZKy1T+IMkL2PPdDrsQU+NsTTPnQZSks21Zwc6VgzpGfdqjk5VHNypGPNl0xjc4JqTo50rBnSs27VnByqOTmGXLNxxi8RERERERGR9DSS19iKiIiIiIiIjLi0CbbGmB8YY86l03YCxpipxpg/GmMOGGP2GWM+meqaBmKM8Rljthljdsdr/j+prmmwjDEZxpidxpjfpLqWwTDGHDfG7DXG7EqXVVmNMXnGmF8YY16J/7leneqaXo8xZl7859t9azbGfCrVdQ3EGPP38f//Ko0xjxpjfKmuaSDGmE/G6903mn/G/Y0lxph8Y8wzxpjD8fsJqawxnWhsTg6NzcmjsXnkaWxOnvE2NqdNsAUeBm5IdRFDFAE+Y61dAFwFfMwYszDFNQ0kDLzBWrsEWArcEF8VMx18EjiQ6iKGaIO1dmkaLcH+78DT1tr5wBJG+c/bWnsw/vNdCqzAWQjnidRW9fqMMaXAPcBKa205zgI/70ptVa/PGFMOfAiowPlz8VfGmDmpreqiHubCseRe4Flr7Rzg2fhzGZyH0dicDBqbk0tj8wjS2Jwc43FsTptga639M1Cf6jqGwlpbba3dEX/cgvMXTWlqq3p91tEaf+qJ30b9hdjGmCnAzcD3Ul3LWGWMyQHWAd8HsNZ2WmsbU1rU0LwReNVaeyLVhQyCG/AbY9xAgH72EB1lFgBbrLXt1toI8Bxwe4pr6tdFxpJbgR/FH/8IuC2ZNaUzjc3JobFZLkZjc1JpbB4hl2tsTptgm+6MMWXAMmBriksZUHza0C7gHPCMtXbU1wx8E/gcEEtxHUNhgd8ZY142xtyV6mIGYSZQA/wwPq3se8aYYKqLGoJ3AY+muoiBWGtfA74BnASqcfYQ/V1qqxpQJbDOGFNgjAngbBUzNcU1DcWk7n1a4/cTU1yPJInG5hH3TTQ2jzSNzUmgsTklhjw2K9gmgTEmC/gl8ClrbXOq6xmItTYanx4yBaiIT2UYtYwxfwWcs9a+nOpahuhqa+1y4EacqXDrUl3QANzAcuA71tplQBtpMmXTGOMFbgF+nupaBhK/huRWYAZQAgSNMe9JbVWvz1p7APga8AzwNLAbZ7qnyKilsXlkaWxOGo3NSaCxOT0o2I4wY4wHZ+D8qbX28VTXMxTxqSx/YvRfP3U1cIsx5jjwGPAGY8x/pbakgVlrq+L353CuLalIbUUDOg2c7tUl+AXOYJoObgR2WGvPprqQQXgTcMxaW2Ot7QIeB9akuKYBWWu/b61dbq1dhzOd6HCqaxqCs8aYYoD4/bkU1yMjTGNzUmhsTg6NzcmhsTn5hjw2K9iOIGOMwbnm4YC19t9SXc9gGGOKjDF58cd+nP+RX0lpUQOw1n7BWjvFWluGM6XlD9baUf1bNGNM0BiT3f0YeDPOlJFRy1p7BjhljJkXP/RGYH8KSxqKO0iDqU5xJ4GrjDGB+N8hb2SULwQCYIyZGL+fBryV9Pl5A2wE3hd//D7g1ymsRUaYxubk0NicHBqbk0Zjc/INeWx2j2g5l5Ex5lHgWqDQGHMa+N/W2u+ntqoBXQ28F9gbvy4G4IvW2idTV9KAioEfGWMycH7x8TNrbVos0Z9mJgFPOH834gYesdY+ndqSBuUTwE/j04eOAu9PcT0Dil9Xch3w4VTXMhjW2q3GmF8AO3CmDO0EHkptVYPyS2NMAdAFfMxa25DqgvrT31gC3Af8zBjzQZx/vLw9dRWmF43NSaOxOTk0NieJxuakGVdjs7F21C+qJyIiIiIiInJRmoosIiIiIiIiaU3BVkRERERERNKagq2IiIiIiIikNQVbERERERERSWsKtiIiIiIiIpLWFGxFRjljTNQYs6vX7d7LeO4yY8yo3qNPRERktNHYLDL6pM0+tiLjWMhauzTVRYiIiEiCxmaRUUYdW5E0ZYw5boz5mjFmW/w2O358ujHmWWPMnvj9tPjxScaYJ4wxu+O3NfFTZRhjvmuM2WeM+Z0xxp+yb0pERCSNaWwWSR0FW5HRz3/edKd39nqt2VpbAXwL+Gb82LeAH1trrwB+CvxH/Ph/AM9Za5cAy4F98eNzgPuttYuARuBtI/rdiIiIpD+NzSKjjLHWproGEXkdxphWa21WP8ePA2+w1h41xniAM9baAmNMLVBsre2KH6+21hYaY2qAKdbacK9zlAHPWGvnxJ9/HvBYa7+chG9NREQkLWlsFhl91LEVSW/2Io8v9p7+hHs9jqJr70VERIZDY7NICijYiqS3d/a63xx//BfgXfHH7wZeiD9+FvgIgDEmwxiTk6wiRURExhGNzSIpoN/+iIx+fmPMrl7Pn7bWdm8rkGmM2YrzS6o74sfuAX5gjPksUAO8P378k8BDxpgP4vz29yNA9UgXLyIiMgZpbBYZZXSNrUiail/Hs9JaW5vqWkRERERjs0gqaSqyiIiIiIiIpDV1bEVERERERCStqWMrIiIiIiIiaU3BVkRERERERNKagq2IiIiIiIikNQVbERERERERSWsKtiIiIiIiIpLWFGxFREREREQkrf3/W1CiRnbW0hAAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 1152x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(16,6))\n",
    "\n",
    "xr = min(len(dense_history.history['accuracy']), len(conv_history.history['accuracy']))\n",
    "\n",
    "# Plot training & validation accuracy values\n",
    "plt.subplot(1, 2, 1)\n",
    "plt.plot(range(1,xr+1),dense_history.history['accuracy'][:xr])\n",
    "plt.plot(range(1,xr+1),dense_history.history['val_accuracy'][:xr])\n",
    "plt.plot(range(1,xr+1),conv_history.history['accuracy'][:xr])\n",
    "plt.plot(range(1,xr+1),conv_history.history['val_accuracy'][:xr])\n",
    "plt.title(\"Models' accuracy on SVHN\")\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Dense-Train', 'Dense-Validation', 'Convolutional-Train', 'Convolutional-Validation'], loc='best')\n",
    "plt.ylim(0., 1.)\n",
    "plt.xlim(1, epochs)\n",
    "plt.xticks(range(1,epochs+1))\n",
    "\n",
    "ylim = ceil(max([max(dense_history.history['loss']), max(dense_history.history['val_loss']), max(conv_history.history['loss']), max(conv_history.history['val_loss'])]))\n",
    "\n",
    "# Plot training & validation loss values\n",
    "plt.subplot(1, 2, 2)\n",
    "plt.plot(range(1,xr+1),dense_history.history['loss'][:xr])\n",
    "plt.plot(range(1,xr+1),dense_history.history['val_loss'][:xr])\n",
    "plt.plot(range(1,xr+1),conv_history.history['loss'][:xr])\n",
    "plt.plot(range(1,xr+1),conv_history.history['val_loss'][:xr])\n",
    "plt.title(\"Models' loss on SVHN\")\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Dense-Train', 'Dense-Validation', 'Convolutional-Train', 'Convolutional-Validation'], loc='best')\n",
    "plt.ylim(0., ylim)\n",
    "plt.xlim(1, epochs)\n",
    "plt.xticks(range(1,epochs+1))\n",
    "\n",
    "plt.show()\n",
    "# fig.savefig('./acc-loss-2.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a09e0aa0-6807-4ee4-ba5b-6a4a85c3fa69",
   "metadata": {},
   "source": [
    "Now, we should compare the two models, to see which performs better in total. For the comparison, the following metrics will be used:\n",
    "* Training time\n",
    "* Accuracy\n",
    "* Log loss\n",
    "* Precision\n",
    "* Recall\n",
    "* F1 score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "72cbf3d7-5ca9-49e9-bd68-b44642f14703",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1800x1080 with 6 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(25,15))\n",
    "\n",
    "plt.subplot(2, 3, 1)\n",
    "plt.bar('dense', times['dense'][0], color='r', width=0.3,\n",
    "        edgecolor='grey', label='dense')\n",
    "plt.bar('convolutional', times['conv'][0], color='b', width=0.3,\n",
    "        edgecolor='grey', label='convolutional')\n",
    "plt.title(\"Models' training times\")\n",
    "plt.ylabel('Time (sec)')\n",
    "\n",
    "plt.subplot(2, 3, 2)\n",
    "plt.bar('dense', accuracies['dense'][0], color='r', width=0.3,\n",
    "        edgecolor='grey', label='dense')\n",
    "plt.bar('convolutional', accuracies['conv'][0], color='b', width=0.3,\n",
    "        edgecolor='grey', label='convolutional')\n",
    "plt.title(\"Models' accuracies\")\n",
    "plt.ylabel('Accuracy')\n",
    "\n",
    "plt.subplot(2, 3, 3)\n",
    "plt.bar('dense', losses['dense'][0], color='r', width=0.3,\n",
    "        edgecolor='grey', label='dense')\n",
    "plt.bar('convolutional', losses['conv'][0], color='b', width=0.3,\n",
    "        edgecolor='grey', label='convolutional')\n",
    "plt.title(\"Models' log losses\")\n",
    "plt.ylabel('Log Loss')\n",
    "\n",
    "plt.subplot(2, 3, 4)\n",
    "plt.bar('dense', precisions['dense'][0], color='r', width=0.3,\n",
    "        edgecolor='grey', label='dense')\n",
    "plt.bar('convolutional', precisions['conv'][0], color='b', width=0.3,\n",
    "        edgecolor='grey', label='convolutional')\n",
    "plt.title(\"Models' precisions\")\n",
    "plt.ylabel('Precision')\n",
    "\n",
    "plt.subplot(2, 3, 5)\n",
    "plt.bar('dense', recalls['dense'][0], color='r', width=0.3,\n",
    "        edgecolor='grey', label='dense')\n",
    "plt.bar('convolutional', recalls['conv'][0], color='b', width=0.3,\n",
    "        edgecolor='grey', label='convolutional')\n",
    "plt.title(\"Models' recalls\")\n",
    "plt.ylabel('Recall')\n",
    "\n",
    "plt.subplot(2, 3, 6)\n",
    "plt.bar('dense', f1s['dense'][0], color='r', width=0.3,\n",
    "        edgecolor='grey', label='dense')\n",
    "plt.bar('convolutional', f1s['conv'][0], color='b', width=0.3,\n",
    "        edgecolor='grey', label='convolutional')\n",
    "plt.title(\"Models' F1 scores\")\n",
    "plt.ylabel('F1 scorejupyter-lab --ip 83.212.78.90 --port 8888 --no-browser')\n",
    "\n",
    "plt.show()\n",
    "# fig.savefig('./measurements-2.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b15da6a2-b195-45cd-91dc-97ff03420718",
   "metadata": {},
   "source": [
    "### Conclusions"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d4fda538-9ff9-407f-856f-5e273a41630f",
   "metadata": {},
   "source": [
    "We can see that a very basic (not fine-tuned) convolutional neural network model always performs better than a basic (not fine-tuned) fully-connected neural network model, while needing fewer training epochs to do so.\n",
    "\n",
    "Surely, fine-tuning the hyperparameters of the model (such as #layers, #units per layer, learning rate, etc.) can result in greater accuracy and lower loss for both the fully-connected network and the convolutional neural network.\n",
    "\n",
    "However, since the convolutional neural networks were inspired by the human vision, we always expect getting better results from their use when it comes to images.\n",
    "\n",
    "Finally, the convolutional models require significantly more time for training. Because of this, tasdfhe use of a GPU is always recommended."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}