Intro_to_ML_LAB.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "<center>\n",
    "<div class=\"alert alert-block alert-info\">\n",
    "<h1><b>Fundamental & Hands-on Introduction to Machine Learning</b></h1></div>\n",
    "\n",
    "Dr. Sothea HAS <br>\n",
    "CNRS, LPSM - Université Paris Cité, LMD - École polytechnique<br>\n",
    "\n",
    "Date: 19/07/2023 <br>\n",
    "</center>\n",
    "\n",
    "---\n",
    "\n",
    "This `jupyter notebook` aims at guiding you through what you have seen during the talk. We begin the first section with `Linear Regression` models, then move to classification problems where you can start to build `Perceptron` model and `Multi-layer Perceptron` or `DNN` models. For more information, read for example, [scikit learn](https://scikit-learn.org/stable/) and [keras](https://keras.io/).\n",
    "\n",
    "**Task**\n",
    "\n",
    "1. Linear regression: Simple & Multiple\n",
    "    - Dataset: `Tips`\n",
    "    - Simple linear regression\n",
    "    - Multiple linear regression\n",
    "2. Logistic regression: Simple & Multinomial\n",
    "    - Dataset: `Spam` & `Mnist`\n",
    "    - Perceptron or Binary logistic regression with `Spam` dataset\n",
    "    - Multi-layer perceptron or Multinomial logistic regression with `Mnist` dataset\n",
    "3. Deep Neural Networks\n",
    "    - Dataset: `Spam` & `Mnist`\n",
    "    - It's your turn!"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 0. Importing datasets and useful libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Basic modules\n",
    "import random\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from sklearn.linear_model import LinearRegression, Ridge, Lasso, ElasticNet, LogisticRegression\n",
    "from sklearn.model_selection import train_test_split, GridSearchCV\n",
    "from sklearn.preprocessing import MinMaxScaler, StandardScaler\n",
    "from sklearn.metrics import mean_squared_error, accuracy_score\n",
    "\n",
    "# Tensorflow for DNN\n",
    "import tensorflow as tf\n",
    "import tensorflow.keras as keras\n",
    "from tensorflow.keras.layers import Dense , Activation, Dropout\n",
    "from tensorflow.keras.optimizers import Adam ,RMSprop\n",
    "from tensorflow.keras.utils import to_categorical, plot_model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "<div class=\"alert alert-block alert-warning\">\n",
    "<b>Note:</b> We load all the datasets in the following cell. Each dataset is mostly balanced, we therefore directly split them into training and testing parts after they are loaded. One should keep in mind that in real project, data analysis and processing are required before train-test split. You need to have the flavor of how the data feels like and what should be taken into account.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1. Tips dataset:\n",
      "Train dimension: (195, 2)\n",
      "Test dimension: (49, 2)\n",
      "\n",
      "2. Spam dataset:\n",
      "Train dimension: (3680, 57)\n",
      "Test dimension: (921, 57)\n",
      "\n",
      "3. Mnist dataset:\n",
      "Train dimension: (60000, 28, 28)\n",
      "Test dimension: (10000, 28, 28)\n"
     ]
    }
   ],
   "source": [
    "fpath = \"https://raw.githubusercontent.com/hassothea/Teaching/main/data/\"\n",
    "\n",
    "tips = sns.load_dataset(\"tips\")\n",
    "X_train_tips, X_test_tips, y_train_tips, y_test_tips = train_test_split(tips.loc[:,['total_bill', 'size']], tips.loc[:,['tip']], test_size=0.2, random_state=123)\n",
    "print(\"1. Tips dataset:\")\n",
    "print(\"Train dimension:\", X_train_tips.shape)\n",
    "print(\"Test dimension:\", X_test_tips.shape)\n",
    "\n",
    "\n",
    "spam = pd.read_csv(fpath+\"spam.txt\", sep=\" \")\n",
    "X_train_spam, X_test_spam, y_train_spam, y_test_spam = train_test_split(spam.iloc[:,1:-1], spam.iloc[:,-1], test_size=0.2, random_state=123)\n",
    "print(\"\\n2. Spam dataset:\")\n",
    "print(\"Train dimension:\", X_train_spam.shape)\n",
    "print(\"Test dimension:\", X_test_spam.shape)\n",
    "\n",
    "(X_train_mnist, y_train_mnist), (X_test_mnist, y_test_mnist) = tf.keras.datasets.mnist.load_data()\n",
    "print('\\n3. Mnist dataset:')\n",
    "print(\"Train dimension:\", X_train_mnist.shape)\n",
    "print(\"Test dimension:\", X_test_mnist.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Linear Regression\n",
    "###     1.1 Simple Linear regression: `tip` vs `total_bill`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# It's natural to plot scatter plot between predictor and target. \n",
    "plt.scatter(X_train_tips.loc[:, ['total_bill']], y_train_tips, label = \"Train\")\n",
    "plt.scatter(X_test_tips.loc[:, ['total_bill']], y_test_tips, label = \"Test\")\n",
    "plt.title(\"Tips vs Total bill\")\n",
    "plt.xlabel(\"Total bill\")\n",
    "plt.ylabel(\"Tips\")\n",
    "plt.legend()\n",
    "fig = plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**To undersatnd any `codes`, you can use `help`: `help(codes)`.**\n",
    "\n",
    "**Example:** `help(LinearRegression)`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([0.79444624]), array([[0.11358781]]))"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Loading linear regression is simple called \"reg0\"\n",
    "reg0 = LinearRegression()\n",
    "\n",
    "# Fit a linear line on the training data using `fit` method:\n",
    "reg0.fit(X_train_tips.loc[:,['total_bill']], y_train_tips)\n",
    "\n",
    "# The estimated coefficients\n",
    "(reg0.intercept_, reg0.coef_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8145329144810914"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Making prediction on testing data using variable 'total_bill':\n",
    "pred0 = reg0.predict(X_test_tips.loc[:,['total_bill']])\n",
    "\n",
    "# Measure error using RMSE compared to \"y_test_tips\"\n",
    "np.sqrt(mean_squared_error(y_test_tips, pred0))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.2 Regularized linear regression\n",
    "#### 1.2.1 Ridge regresssion\n",
    "\n",
    "$$(a^*,b^*,\\alpha^*)=\\arg\\min_{(a,b,\\alpha)}\\sum_{i=1}^n(y_i-(ax_i+b))^2+\\alpha(a^2+b^2)$$\n",
    "\n",
    "- `GridSearchCV` performs grid search algorithm over a list of parameters you want to tune, using K-fold cross-validation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best parameter: 1922.9305050351759\n"
     ]
    }
   ],
   "source": [
    "# Cross-validation with GridSearchCV is used to search for penalty parameter `alpha` (`lambda` in the talk)\n",
    "grid = {\n",
    "    \"alpha\" : 2 ** np.linspace(-10,20,num=100) # regularization strength\n",
    "}\n",
    "\n",
    "ridge_cv = GridSearchCV(estimator= Ridge(), \n",
    "                        param_grid=grid, cv=5)\n",
    "\n",
    "ridge_cv.fit(X_train_tips.loc[:,['total_bill']], y_train_tips)\n",
    "\n",
    "print('Best parameter:', ridge_cv.best_params_['alpha'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Coefficients: (array([1.04129933]), array([[0.1009982]]))\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.7682224735315277"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Train Ridge model with the optimal panelty parameter obtained above\n",
    "\n",
    "ridge0 = Ridge(alpha=ridge_cv.best_params_['alpha']).fit(X_train_tips.loc[:,['total_bill']], y_train_tips)\n",
    "print(\"Coefficients:\", (ridge0.intercept_, ridge0.coef_))\n",
    "\n",
    "# predict \n",
    "pred_ridge0 = ridge0.predict(X_test_tips.loc[:,['total_bill']])\n",
    "\n",
    "# Measure error using RMSE\n",
    "np.sqrt(mean_squared_error(y_test_tips, pred_ridge0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAjIAAAHHCAYAAACle7JuAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8o6BhiAAAACXBIWXMAAA9hAAAPYQGoP6dpAACIMklEQVR4nO3deXgTVdsG8HvSHdqmFGjTQqFlp+xbgQKWVTZZFAGVVRQVQUREARVZ9BNQQXADFV9WEUEFQbSIStmhyCbIKraA0FLZuq/JfH+EhKaZJJM0aZL2/l1Xr9fMmcycmeRlnpzlOYIoiiKIiIiI3JDC2RUgIiIishUDGSIiInJbDGSIiIjIbTGQISIiIrfFQIaIiIjcFgMZIiIiclsMZIiIiMhtMZAhIiIit8VAhoiIiNwWAxmiCmDs2LGIjIx0djXKpTlz5kAQBIv7de3aFU2bNrW4X3JyMgRBwKpVq8yeIzIyEmPHjrW2ukTlDgMZIjclCIKsv4SEBGdX1a4SEhJkX7sl169fx5w5c3DixAnHV5yIHMLT2RUgItusXbvW4PWaNWuwc+dOo+2NGzfGF198AY1GU5bVc5jGjRsbXePMmTPh7++P119/3apjXb9+HXPnzkVkZCRatmxpx1rarnbt2sjNzYWXl5ezq0LkFhjIELmpkSNHGrw+dOgQdu7cabS9vAkNDTW6xgULFqBatWrl4toFQYCvr6+zq0HkNti1RFQBlBwjoxuH8f777+ODDz5A7dq14efnh7i4OJw+fdrgvampqXjyySdRs2ZN+Pj4ICwsDIMGDUJycrLJ873//vsQBAGXL182Kps5cya8vb1x584dAMDFixcxZMgQqFQq+Pr6ombNmnjssceQnp5eqmv+559/MHToUAQHB6NSpUro0KEDtm/fri9PSEhAu3btAABPPvmkvjtKNzZl7969GDp0KGrVqgUfHx9ERETgpZdeQm5ubqnqdfToUcTGxsLPzw9RUVFYvny5QbnUGBkiMo0tMkQV2Jo1a5CZmYmJEyciLy8PS5cuRffu3XHq1CmEhoYCAIYMGYK//voLL7zwAiIjI5GWloadO3fiypUrJgcQDxs2DK+++io2btyIV155xaBs48aNePDBB1GlShUUFBSgd+/eyM/PxwsvvACVSoVr167hxx9/xN27d6FUKm26rhs3biA2NhY5OTmYPHkyqlatitWrV2PgwIH49ttv8fDDD6Nx48aYN28e3nzzTTzzzDPo0qULACA2NhYAsGnTJuTk5GDChAmoWrUqEhMT8dFHH+Hff//Fpk2bbKrXnTt30K9fPwwbNgyPP/44Nm7ciAkTJsDb2xvjxo2z6ZhEFZ5IROXCxIkTRVP/lx4zZoxYu3Zt/eukpCQRgOjn5yf++++/+u2HDx8WAYgvvfSSKIqieOfOHRGA+N5771ldn44dO4pt2rQx2JaYmCgCENesWSOKoigeP35cBCBu2rTJ6uMX16RJEzEuLk7/esqUKSIAce/evfptmZmZYlRUlBgZGSmq1WpRFEXxyJEjIgBx5cqVRsfMyckx2jZ//nxREATx8uXL+m2zZ882ed+Li4uLEwGIixYt0m/Lz88XW7ZsKYaEhIgFBQWiKN7/bIrXSeoctWvXFseMGWPxvETlHbuWiCqwwYMHo0aNGvrXMTExaN++PX766ScAgJ+fH7y9vZGQkKDvCpJr+PDhOHr0KC5duqTf9s0338DHxweDBg0CAH2Ly44dO5CTk1Pay9H76aefEBMTg86dO+u3+fv745lnnkFycjLOnDlj8Rh+fn76/87OzsbNmzcRGxsLURRx/Phxm+rl6emJZ599Vv/a29sbzz77LNLS0nD06FGbjklU0TGQIarA6tevb7StQYMG+vEvPj4+WLhwIX7++WeEhobigQcewLvvvovU1FSLxx46dCgUCgW++eYbAIAoiti0aRP69u2LwMBAAEBUVBSmTp2KFStWoFq1aujduzc++eSTUo+PuXz5Mho2bGi0vXHjxvpyS65cuYKxY8ciODgY/v7+qF69OuLi4gDA5vqFh4ejcuXKBtsaNGgAAGbHHBGRaQxkiMisKVOm4MKFC5g/fz58fX0xa9YsNG7c2GKrRHh4OLp06YKNGzcC0M6qunLlCoYPH26w36JFi/Dnn3/itddeQ25uLiZPnowmTZrg33//ddg1WaJWq9GrVy9s374d06dPx5YtW7Bz5079ANzyMpWdqDxgIENUgV28eNFo24ULF4wG8datWxcvv/wyfvnlF5w+fRoFBQVYtGiRxeMPHz4cJ0+exPnz5/HNN9+gUqVKGDBggNF+zZo1wxtvvIE9e/Zg7969uHbtmtFsHmvUrl0b58+fN9p+7tw5fTkAk0nzTp06hQsXLmDRokWYPn06Bg0ahJ49eyI8PNzmOgHavDXZ2dkG2y5cuAAAzLxMZCMGMkQV2JYtW3Dt2jX968TERBw+fBh9+/YFAOTk5CAvL8/gPXXr1kVAQADy8/MtHn/IkCHw8PDA119/jU2bNuGhhx4y6FrJyMhAUVGRwXuaNWsGhUIh6/im9OvXD4mJiTh48KB+W3Z2Nj7//HNERkYiOjoaAPR1uXv3rsH7PTw8AGi7w3REUcTSpUttrhMAFBUV4bPPPtO/LigowGeffYbq1aujTZs2pTo2UUXF6ddEFVi9evXQuXNnTJgwAfn5+ViyZAmqVq2KV199FYC2taBHjx4YNmwYoqOj4enpic2bN+PGjRt47LHHLB4/JCQE3bp1w+LFi5GZmWnUrfT7779j0qRJGDp0KBo0aICioiKsXbsWHh4eGDJkiM3XNWPGDHz99dfo27cvJk+ejODgYKxevRpJSUn47rvvoFBof8PVrVsXQUFBWL58OQICAlC5cmW0b98ejRo1Qt26dTFt2jRcu3YNgYGB+O6776we8FxSeHg4Fi5ciOTkZDRo0ADffPMNTpw4gc8//5yZfIlsxECGqAIbPXo0FAoFlixZgrS0NMTExODjjz9GWFgYACAiIgKPP/44fvvtN6xduxaenp5o1KgRNm7cKDvQGD58OH799VcEBASgX79+BmUtWrRA7969sW3bNly7dg2VKlVCixYt8PPPP6NDhw42X1doaCgOHDiA6dOn46OPPkJeXh6aN2+Obdu2oX///vr9vLy8sHr1asycORPPPfccioqKsHLlSowdOxbbtm3D5MmT9WODHn74YUyaNAktWrSwuV5VqlTB6tWr8cILL+CLL75AaGgoPv74Y4wfP97mYxJVdIJYvO2UiCqE5ORkREVF4b333sO0adOcXR0iIptxjAwRERG5LQYyRERE5LYYyBAREZHb4hgZIiIicltskSEiIiK3xUCGiIiI3Fa5zyOj0Whw/fp1BAQEmExHTkRERK5FFEVkZmYiPDxcn8RSSrkPZK5fv46IiAhnV4OIiIhscPXqVdSsWdNkebkPZAICAgBob0RgYKCTa0NERERyZGRkICIiQv8cN6XcBzK67qTAwEAGMkRERG7G0rAQDvYlIiIit8VAhoiIiNwWAxkiIiJyW+V+jIxcarUahYWFzq4GVWBeXl7w8PBwdjWIiNxKhQ9kRFFEamoq7t696+yqECEoKAgqlYo5j4iIZKrwgYwuiAkJCUGlSpX4ACGnEEUROTk5SEtLAwCEhYU5uUZERO6hQgcyarVaH8RUrVrV2dWhCs7Pzw8AkJaWhpCQEHYzERHJUKEH++rGxFSqVMnJNSHS0n0XOV6LiEieCh3I6LA7iVwFv4tERNap0F1LREREZBu1RkRi0m2kZeYhJMAXMVHB8FCU/Y8xp7bI7NmzBwMGDEB4eDgEQcCWLVsMykVRxJtvvomwsDD4+fmhZ8+euHjxonMqW85FRkZiyZIlzq4GERG5gfjTKei88Hc8/sUhvLjhBB7/4hA6L/wd8adTyrwuTg1ksrOz0aJFC3zyySeS5e+++y4+/PBDLF++HIcPH0blypXRu3dv5OXllXFNXYcgCGb/5syZY9Nxjxw5gmeeeca+lSUionIn/nQKJqw7hpR0w2dxanoeJqw7VubBjFO7lvr27Yu+fftKlomiiCVLluCNN97AoEGDAABr1qxBaGgotmzZgscee6wsq2pWWTavpaTc/4J88803ePPNN3H+/Hn9Nn9/f/1/i6IItVoNT0/LH3P16tXtW1EiIip31BoRc7edgShRJgIQAMzddga9olVl1s3ksoN9k5KSkJqaip49e+q3KZVKtG/fHgcPHjT5vvz8fGRkZBj8OVJZN6+pVCr9n1KphCAI+tfnzp1DQEAAfv75Z7Rp0wY+Pj7Yt28fLl26hEGDBiE0NBT+/v5o164dfv31V4PjluxaEgQBK1aswMMPP4xKlSqhfv362Lp1q0OuiYiI3ENi0m2jlpjiRAAp6XlITLpdZnVy2UAmNTUVABAaGmqwPTQ0VF8mZf78+VAqlfq/iIgIh9XR1ZrXdGbMmIEFCxbg7NmzaN68ObKystCvXz/89ttvOH78OPr06YMBAwbgypUrZo8zd+5cDBs2DH/++Sf69euHESNG4PbtsvtyEhGRa0nLlDe0Q+5+9uCygYytZs6cifT0dP3f1atXHXIeS81rgLZ5Ta2R2sOx5s2bh169eqFu3boIDg5GixYt8Oyzz6Jp06aoX78+3nrrLdStW9diC8vYsWPx+OOPo169enjnnXeQlZWFxMTEMroKIiJyNSEBvnbdzx5cNpBRqVQAgBs3bhhsv3Hjhr5Mio+PDwIDAw3+HMEVm9d02rZta/A6KysL06ZNQ+PGjREUFAR/f3+cPXvWYotM8+bN9f9duXJlBAYG6lPoExFRxRMTFYwwpS9MjX4RAIQptWNFy4rLBjJRUVFQqVT47bff9NsyMjJw+PBhdOzY0Yk103LF5jWdypUrG7yeNm0aNm/ejHfeeQd79+7FiRMn0KxZMxQUFJg9jpeXl8FrQRCg0WjsXl8iInIPHgoBswdEA4BRMKN7PXtAdJnmk3HqrKWsrCz8/fff+tdJSUk4ceIEgoODUatWLUyZMgVvv/026tevj6ioKMyaNQvh4eEYPHiw8yp9jys2r5myf/9+jB07Fg8//DAA7X1PTk52bqWIiMgt9WkahmUjW2PutjMGPRMqpS9mD4hGn6Zlu+itUwOZP/74A926ddO/njp1KgBgzJgxWLVqFV599VVkZ2fjmWeewd27d9G5c2fEx8fD19f5wYGueS01PU9ynIwA7Ydals1rptSvXx/ff/89BgwYAEEQMGvWLLasEBGRzfo0DUOvaJVLZPZ1aiDTtWtXiKLpwbCCIGDevHmYN29eGdZKHl3z2oR1xyAABsGMs5rXTFm8eDHGjRuH2NhYVKtWDdOnT3f4tHQiIirfPBQCOtat6uxqQBDNRRLlQEZGBpRKJdLT040G/ubl5SEpKQlRUVE2t/LEn04xal4Lc1LzGrk/e3wniYjKA3PP7+K4aGQpuVLzGhERUUXDQMYOXKV5jYiIqKJx2enXRERERJYwkCEiIiK3xUCGiIiI3BYDGSIiInJbDGSIiIjIbTGQISIiIrfFQIaIiIjcFgMZIiIiclsMZNyMIAhm/+bMmVOqY2/ZssVudSUiInI0Zva1B40auHwAyLoB+IcCtWMBhYdDTpWSkqL/72+++QZvvvkmzp8/r9/m7+/vkPMSERG5IrbIlNaZrcCSpsDqh4DvntL+75Km2u0OoFKp9H9KpRKCIBhs27BhAxo3bgxfX180atQIn376qf69BQUFmDRpEsLCwuDr64vatWtj/vz5AIDIyEgAwMMPPwxBEPSviYiIXBlbZErjzFZg42gAJRYQz0jRbh+2BogeWGbV+eqrr/Dmm2/i448/RqtWrXD8+HGMHz8elStXxpgxY/Dhhx9i69at2LhxI2rVqoWrV6/i6tWrAIAjR44gJCQEK1euRJ8+feDh4ZgWJSIiIntiIGMrjRqInw6jIAa4t00A4mcAjfo7rJuppNmzZ2PRokV45JFHAABRUVE4c+YMPvvsM4wZMwZXrlxB/fr10blzZwiCgNq1a+vfW716dQBAUFAQVCpVmdSXiIiotNi1ZKvLB4CM62Z2EIGMa9r9ykB2djYuXbqEp556Cv7+/vq/t99+G5cuXQIAjB07FidOnEDDhg0xefJk/PLLL2VSNyIiIkdhi4ytsm7Yd79SysrKAgB88cUXaN++vUGZrpuodevWSEpKws8//4xff/0Vw4YNQ8+ePfHtt9+WSR2JiIjsjYGMrfxD7btfKYWGhiI8PBz//PMPRowYYXK/wMBADB8+HMOHD8ejjz6KPn364Pbt2wgODoaXlxfUanWZ1JeIiMgeGMjYqnYsEBiuHdgrOU5G0JbXji2zKs2dOxeTJ0+GUqlEnz59kJ+fjz/++AN37tzB1KlTsXjxYoSFhaFVq1ZQKBTYtGkTVCoVgoKCAGhnLv3222/o1KkTfHx8UKVKlTKrOxERkS04RsZWCg+gz8J7L4QShfde91lQZgN9AeDpp5/GihUrsHLlSjRr1gxxcXFYtWoVoqKiAAABAQF499130bZtW7Rr1w7Jycn46aefoFBovwaLFi3Czp07ERERgVatWpVZvYmIiGwliKIo1ZxQbmRkZECpVCI9PR2BgYEGZXl5eUhKSkJUVBR8fX1tO8GZrdrZS8UH/gbW0AYxZTj1msoHu3wniYjKAXPP7+LYtVRa0QO1U6zLKLMvERER3cdAxh4UHkBUF2fXgoiIqMLhGBkiIiJyW2yRISIiKqfUGhGJSbeRlpmHkABfxEQFw0NRcoKKe2MgQ0REVA7Fn07B3G1nkJKep98WpvTF7AHR6NM0zIk1sy92LREREZUz8adTMGHdMYMgBgBS0/MwYd0xxJ9OcVLN7I+BDBERUTmi1oiYu+2MySWNAWDutjNQa8pH9hUGMkREROVIYtJto5aY4kQAKel5SEy6XXaVciAGMkREROVIWqbpIMaW/VwdA5kKIiEhAYIg4O7du2b3i4yMxJIlS8qkTu6k5P1btWqVfo0qIiJXEhIgLyu43P1cHQMZN7N8+XIEBASgqKhIvy0rKwteXl7o2rWrwb66h++lS5cQGxuLlJQUKJVKAM5/ELt7wDR8+HBcuHBB/3rOnDlo2bKl8ypERHRPTFQwwpS+RqsA6gjQzl6KiQouy2o5DAMZN9OtWzdkZWXhjz/+0G/bu3cvVCoVDh8+jLy8+02Fu3btQq1atVC3bl14e3tDpVJBEMpX/gBn8fPzQ0hIiLOrQURkxEMhYPaAaAAmlzTG7AHR5SafDAMZN9OwYUOEhYUhISFBvy0hIQGDBg1CVFQUDh06ZLC9W7du+v/WdY0kJCTgySefRHp6OgRBgCAImDNnjv59OTk5GDduHAICAlCrVi18/vnnBnU4deoUunfvDj8/P1StWhXPPPMMsrKy9OVdu3bFlClTDN4zePBgjB07Vl9++fJlvPTSS/rzmyIIApYtW4a+ffvCz88PderUwbfffmuwz9WrVzFs2DAEBQUhODgYgwYNQnJysr587NixGDx4MN5//32EhYWhatWqmDhxIgoLC/X7rF27Fm3btkVAQABUKhWeeOIJpKWlmaxX8RatVatWYe7cuTh58qT+elatWoVx48bhoYceMnhfYWEhQkJC8OWXX5o8NhFRafVpGoZlI1tDpTTsPlIpfbFsZGvmkSmvRFFETkGRU/6sWYS8W7du2LVrl/71rl270LVrV8TFxem35+bm4vDhw/pAprjY2FgsWbIEgYGBSElJQUpKCqZNm6YvX7RoEdq2bYvjx4/j+eefx4QJE3D+/HkAQHZ2Nnr37o0qVargyJEj2LRpE3799VdMmjRJdv2///571KxZE/PmzdOf35xZs2ZhyJAhOHnyJEaMGIHHHnsMZ8+eBaANDHr37o2AgADs3bsX+/fvh7+/P/r06YOCggKDe3Tp0iXs2rULq1evxqpVq7Bq1Sp9eWFhId566y2cPHkSW7ZsQXJysj7wsmT48OF4+eWX0aRJE/31DB8+HE8//TTi4+MNru/HH39ETk4Ohg8fLvt+ERHZok/TMOyb3h1fj++ApY+1xNfjO2Df9O7lKogBmNnXQG6hGtFv7nDKuc/M641K3vI+jm7dumHKlCkoKipCbm4ujh8/jri4OBQWFmL58uUAgIMHDyI/P18ykPH29oZSqYQgCFCpVEbl/fr1w/PPPw8AmD59Oj744APs2rULDRs2xPr165GXl4c1a9agcuXKAICPP/4YAwYMwMKFCxEaGmqx/sHBwfDw8NC3flgydOhQPP300wCAt956Czt37sRHH32ETz/9FN988w00Gg1WrFihb9lZuXIlgoKCkJCQgAcffBAAUKVKFXz88cfw8PBAo0aN0L9/f/z2228YP348AGDcuHH689WpUwcffvgh2rVrh6ysLPj7+5utn5+fH/z9/eHp6WlwPbGxsWjYsCHWrl2LV199VV+3oUOHWjwmEZE9eCgEdKxb1dnVcCi2yLihrl27Ijs7G0eOHMHevXvRoEEDVK9eHXFxcfpxMgkJCahTpw5q1apl9fGbN2+u/29dsKPrZjl79ixatGihD2IAoFOnTtBoNPpWG3vr2LGj0Wtdi8zJkyfx999/IyAgAP7+/vD390dwcDDy8vJw6dIl/XuaNGkCDw8P/euwsDCDrqOjR49iwIABqFWrFgICAhAXFwcAuHLlSqnq/vTTT2PlypUAgBs3buDnn382CJqIiKh02CJTjJ+XB87M6+20c8tVr1491KxZE7t27cKdO3f0D93w8HBERETgwIED2LVrF7p3725TXby8vAxeC4IAjUYj+/0KhcKoq6z4eBR7ysrKQps2bfDVV18ZlVWvXl3/3+auSddd1rt3b3z11VeoXr06rly5gt69ext0T9li9OjRmDFjBg4ePIgDBw4gKioKXbp0KdUxiYjoPgYyxQiCILt7x9m6deuGhIQE3LlzB6+88op++wMPPICff/4ZiYmJmDBhgsn3e3t7Q61WW33exo0bY9WqVcjOzta3yuzfvx8KhQINGzYEoA0gio8LUavVOH36tEE3lzXnP3ToEEaPHm3wulWrVgCA1q1b45tvvkFISAgCAwOtvh4AOHfuHG7duoUFCxYgIiICAAxmhclh6nqqVq2KwYMHY+XKlTh48CCefPJJm+pIRETS2LXkprp164Z9+/bhxIkT+hYZAIiLi8Nnn32GgoICyfExOpGRkcjKysJvv/2GmzdvIicnR9Z5R4wYAV9fX4wZMwanT5/Grl278MILL2DUqFH68THdu3fH9u3bsX37dpw7dw4TJkwwSsQXGRmJPXv24Nq1a7h586bZc27atAn/+9//cOHCBcyePRuJiYn6wcUjRoxAtWrVMGjQIOzduxdJSUlISEjA5MmT8e+//8q6plq1asHb2xsfffQR/vnnH2zduhVvvfWWrPcWv56kpCScOHECN2/eRH5+vr7s6aefxurVq3H27FmMGTPGquMSEZF5DGTcVLdu3ZCbm4t69eoZDLCNi4tDZmamfpq2KbGxsXjuuecwfPhwVK9eHe+++66s81aqVAk7duzA7du30a5dOzz66KPo0aMHPv74Y/0+48aNw5gxYzB69GjExcWhTp06RkHVvHnzkJycjLp16xp0AUmZO3cuNmzYgObNm2PNmjX4+uuvER0dra/Pnj17UKtWLTzyyCNo3LgxnnrqKeTl5cluoalevTpWrVqFTZs2ITo6GgsWLMD7778v6706Q4YMQZ8+fdCtWzdUr14dX3/9tb6sZ8+eCAsLQ+/evREeHm7VcYmIyDxBtGberxvKyMiAUqlEenq60YMtLy8PSUlJiIqKgq9v+UjVXN4IgoDNmzdj8ODBzq6KzbKyslCjRg2sXLkSjzzyiNl9+Z0kItIy9/wuzj0GhBC5IY1Gg5s3b2LRokUICgrCwIEDnV0lIqJyh4EMkYNcuXIFUVFRqFmzJlatWgVPT/7fjYjI3vgvK7k0d+75jIyMdOv6ExG5Aw72JSIiIrfFQIaIiIjcFgMZIiIiclsMZIiIiMhtMZAhIiIit8VAhoiIiNwWA5kKIiEhAYIgGK15VFJkZCSWLFlSJnVyBEEQsGXLFgBAcnIyBEHAiRMnnFonIiJyHAYybmb58uUICAhAUVGRfltWVha8vLzQtWtXg311wculS5cQGxuLlJQUKJVKAMCqVasQFBRUhjUvexEREUhJSUHTpk0ByA/miIjIfTCQcTPdunVDVlYW/vjjD/22vXv3QqVS4fDhw8jLy9Nv37VrF2rVqoW6devC29sbKpUKgiA4o9pO4eHhAZVKxYy6RETlGAMZN6Nb1TohIUG/LSEhAYMGDUJUVBQOHTpksF236nTx1oiEhAQ8+eSTSE9PhyAIEAQBc+bM0b8vJycH48aNQ0BAAGrVqoXPP//cbJ26du2KSZMmYdKkSVAqlahWrRpmzZplkNU2Pz8f06ZNQ40aNVC5cmW0b9/e4Bp0LUQ7duxA48aN4e/vjz59+iAlJUW/z5EjR9CrVy9Uq1YNSqUScXFxOHbsmMl6Fe9aSk5O1t+LKlWqQBAEjB07FmvWrEHVqlWRn59v8N7Bgwdj1KhRZq+biIicj4FMcaIIFGQ758+KVPbdunXDrl279K937dqFrl27Ii4uTr89NzcXhw8f1j+8i4uNjcWSJUsQGBiIlJQUpKSkYNq0afryRYsWoW3btjh+/Dief/55TJgwAefPnzdbp9WrV8PT0xOJiYlYunQpFi9ejBUrVujLJ02ahIMHD2LDhg34888/MXToUPTp0wcXL17U75OTk4P3338fa9euxZ49e3DlyhWDemVmZmLMmDHYt28fDh06hPr166Nfv37IzMy0eM8iIiLw3XffAQDOnz+PlJQULF26FEOHDoVarcbWrVv1+6alpWH79u0YN26cxeMSEZFzsc29uMIc4J1w55z7teuAd2VZu3br1g1TpkxBUVERcnNzcfz4ccTFxaGwsBDLly8HABw8eBD5+fmSgYy3tzeUSiUEQYBKpTIq79evH55//nkAwPTp0/HBBx9g165daNiwock6RURE4IMPPoAgCGjYsCFOnTqFDz74AOPHj8eVK1ewcuVKXLlyBeHh2vs7bdo0xMfHY+XKlXjnnXcAQF//unXrAtAGP/PmzdOfo3v37gbn/PzzzxEUFITdu3fjoYceMnvPPDw8EBwcDAAICQkxGB/0xBNPYOXKlRg6dCgAYN26dahVq5bRmCMiInI9bJFxQ127dkV2djaOHDmCvXv3okGDBqhevTri4uL042QSEhJQp04d1KpVy+rjN2/eXP/fumAnLS3N7Hs6dOhgMP6mY8eOuHjxItRqNU6dOgW1Wo0GDRrA399f/7d7925cunRJ/55KlSrpgxgACAsLMzjvjRs3MH78eNSvXx9KpRKBgYHIysrClStXrL7G4saPH49ffvkF165dA6Dt5ho7dmyFGk9EROSu2CJTnFclbcuIs84tU7169VCzZk3s2rULd+7cQVxcHAAgPDwcEREROHDgAHbt2mXUgiG7Kl5eBq8FQYBGo7HpWIB2VpWHhweOHj0KDw8PgzJ/f3+z5y0+zmbMmDG4desWli5ditq1a8PHxwcdO3ZEQUGBzXUDgFatWqFFixZYs2YNHnzwQfz111/Yvn17qY5JRERlg4FMcYIgu3vH2bp164aEhATcuXMHr7zyin77Aw88gJ9//hmJiYmYMGGCyfd7e3tDrVbbrT6HDx82eK0bw+Lh4YFWrVpBrVYjLS0NXbp0sfkc+/fvx6effop+/foBAK5evYqbN2/Kfr+3tzcASF73008/jSVLluDatWvo2bMnIiIibK4nERGVHXYtualu3bph3759OHHihL5FBgDi4uLw2WefoaCgQHJ8jE5kZCSysrLw22+/4ebNm8jJySlVfa5cuYKpU6fi/Pnz+Prrr/HRRx/hxRdfBAA0aNAAI0aMwOjRo/H9998jKSkJiYmJmD9/vlUtH/Xr18fatWtx9uxZHD58GCNGjICfn5/s99euXRuCIODHH3/Ef//9h6ysLH3ZE088gX///RdffPEFB/kSEbkRBjJuqlu3bsjNzUW9evUQGhqq3x4XF4fMzEz9NG1TYmNj8dxzz2H48OGoXr063n333VLVZ/To0cjNzUVMTAwmTpyIF198Ec8884y+fOXKlRg9ejRefvllNGzYEIMHD8aRI0esGsPz5Zdf4s6dO2jdujVGjRqFyZMnIyQkRPb7a9Sogblz52LGjBkIDQ3FpEmT9GVKpRJDhgyBv78/Bg8eLPuYRETkXIIoWjHvt4yp1WrMmTMH69atQ2pqKsLDwzF27Fi88cYbsgdiZmRkQKlUIj09HYGBgQZleXl5SEpKQlRUFHx9fR1xCRVC165d0bJlS7de2gAAevTogSZNmuDDDz90Wh34nSQi0jL3/C7OpcfILFy4EMuWLcPq1avRpEkT/PHHH3jyySehVCoxefJkZ1ePyok7d+4gISEBCQkJ+PTTT51dHSIisoJLBzIHDhzAoEGD0L9/fwDacR1ff/01EhMTnVwzKk9atWqFO3fuYOHChWZz5RARketx6UAmNjYWn3/+OS5cuIAGDRrg5MmT2LdvHxYvXuzsqlExxZcacEfJycnOrgIREdnIpQOZGTNmICMjA40aNYKHhwfUajX+7//+DyNGjDD5nvz8fIN1czIyMsqiqkREROQELj1raePGjfjqq6+wfv16HDt2DKtXr8b777+P1atXm3zP/PnzoVQq9X9y8oG48HhnqmD4XSQiso5Lz1qKiIjAjBkzMHHiRP22t99+G+vWrcO5c+ck3yPVIhMRESE56lmtVuPChQsICQlB1apVHXMRRFa4desW0tLS0KBBA6MsyEREFUm5mLWUk5MDhcKw0cjDw8NsunwfHx/4+PjIOr6HhweCgoL06/lUqlSJ6+uQU4iiiJycHKSlpSEoKIhBDBGRTC4dyAwYMAD/93//h1q1aqFJkyY4fvw4Fi9ebNfMq7rVny0tikhUFoKCgiRXJCciImku3bWUmZmJWbNmYfPmzUhLS0N4eDgef/xxvPnmm/p1cyyR2zSlVqtRWFhor6oTWc3Ly4stMURE98h9frt0IGMPcm8EERERuQ65z2+XnrVEREREZA4DGSIiInJbDGSIiIjIbTGQISIiIrfl0tOviYioYlFrRCQm3UZaZh5CAnwRExUMDwXze5FpDGSIiMglxJ9OwdxtZ5CSnqffFqb0xewB0ejTNMyJNSNXxq4lIiJyuvjTKZiw7phBEAMAqel5mLDuGOJPpzipZuTqGMgQEZFTqTUi5m47A6mkZrptc7edgVpTrtOekY0YyBARkVMlJt02aokpTgSQkp6HxKTbZVcpchsMZIiIyKnSMk0HMbbsRxULAxkiInKqkABfu+5HFQsDGSIicqqYqGCEKX1hapK1AO3spZio4LKsFrkJBjJERORUHgoBswdEA4BRMKN7PXtANPPJkCQGMkRE5HR9moZh2cjWUCkNu49USl8sG9maeWTIJCbEIyIil9CnaRh6RauY2ZeswkCGiIhchodCQMe6VZ1dDXIj7FoiIiIit8VAhoiIiNwWAxkiIiJyWwxkiIiIyG0xkCEiIiK3xUCGiIiI3BYDGSIiInJbDGSIiIjIbTGQISIiIrfFQIaIiIjcFgMZIiIiclsMZIiIiMhtMZAhIiIit8VAhoiIiNwWAxkiIiJyWwxkiIiIyG0xkCEiIiK3xUCGiIiI3BYDGSIiInJbDGSIiIjIbXk6uwJERERkf2qNiMSk20jLzENIgC9iooLhoRCcXS27YyBDRFROVJQHF1kWfzoFc7edQUp6nn5bmNIXswdEo0/TMCfWzP4YyBARlQMV6cFF5sWfTsGEdccgltiemp6HCeuOYdnI1uXqO8ExMkREbk734CoexAD3H1zxp1OcVDMqa2qNiLnbzhgFMQD02+ZuOwO1RmoP98RAhojIjVXEBxeZlph02yigLU4EkJKeh8Sk22VXKQdjIENE5MYq4oOLTEvLNP1dsGU/d8BAhojIjVXEBxeZFhLga9f93AEDGSIiN1YRH1xkWkxUMMKUvjA1V02AdhB4TFRwWVbLoRjIEBG5sYr44CLTPBQCZg+IBgCj74Tu9ewB0eVqWj4DGSKSRa0RcfDSLfxw4hoOXrrldoNH3b3+plTEBxeZ16dpGJaNbA2V0rAVTqX0LXdTrwFAEEWxfPy/2YSMjAwolUqkp6cjMDDQ2dUhckvunqPE3esvR0W4RrKOoxMkfnf0X7y86SSqB/hg07MdEVmtst2ODch/fjOQISKzTCXX0v1z6Oq/8Ny9/tZgZl9yNI1GxLwfz2DVgWSD7d9NiEWb2lXsei65z29m9iUikyzlKBGgzVHSK1rlkg9Md6+/tTwUAjrWrersalA5lJFXiCdXHsHRy3eMyt4a3NTuQYw1GMgQkUnW5ChxxQeou9efyNku/ZeFvkv2okCtMSrb8EwHdKjj/P/fMJAhIpPcPUeJu9efyFl+O3sDT63+w2h79QAfbJnYCTWC/JxQK2kMZIjIJHfPUeLu9ScqS6Io4qPf/8binReMyno0CsHHT7SGn7eHE2pmHgMZIjJJl6MkNT1PcpyJAO2UTlfNUeLu9ScqC3mFakxafxy/nr1hVDalZ3282KM+BMF1x5AxkCEik3Q5SiasOwYBMAgG3CFHibvXn8iRUtJzMfiT/biRkW9U9vmoNniwicoJtbIep18TkUXunqPE3etPZE9Hkm9j6PKDRtsVArBjygOoHxrghFoZYx6ZexjIENmHu+cocff6E5XWukOX8caW00bbW0QEYc2TMVBW8nJCrUxjHhkisit3z1Hi7vUnskWRWoM3tpzGhiNXjcpGdaiNOQObuH1Az0CGiIionLmTXYARKw7jTEqGUdm7jzbHsLYRTqiVYzCQISIiKifOpWagz5K9kmWbn49Fq1rOy8DrKAxkiIiI3Nz2P1Mwcf0xo+21givh2+c6IiSw/OZKYiBDRETkhkRRxLs7zmNZwiWjsv7Nw7B4WAv4eDo4gV3Wf0B+BlC1rmPPYwYDGSIiIjeSnV+EZ9cexb6/bxqVzezbCM88UMfxCeyOfwX88Pz912O3A5GdHXtOExjIEBERuYErt3Lw0Ed7kZFXZFS2elwM4hpUd2wFstKADU8A/x4xLguq7dhzm8FAhojIyZjjhszZd/EmRn552Gh7gI8nfpzcGbWrVnZsBY6tBbZOki574FWg6wxA4bw1mFw+kLl27RqmT5+On3/+GTk5OahXrx5WrlyJtm3bOrtqRESlxqzDZMqKvf/g7e1njbbH1q2Kz0e3hb+PAx/hmanA148D140HEKNyCDBmGxDSyHHnt4JLBzJ37txBp06d0K1bN/z888+oXr06Ll68iCpVyt/0MSKqeOJPp2DCumNGC1qmpudhwrpjWDayNYOZCqagSINpm05i68nrRmXPxtXB9N6NoHBka93R1cC2ydJlcTOAuOmAQuG489vApQOZhQsXIiIiAitXrtRvi4qKcmKNiKi8KuvuHbVGxNxtZyRX5RahXdRy7rYz6BWtYjdTBZCWmYdhyw8i+VaOUdmHj7fCwBbhjjt5Zirw9WPA9ePGZf4qYMxWoHpDx52/lFw6kNm6dSt69+6NoUOHYvfu3ahRowaef/55jB8/3uR78vPzkZ9/fyXPjAzjrIZERMU5o3snMem2wflKEgGkpOchMek2l1Yox05evYtBn+yXLNs+uTOahCsdd/I/VgI/TpEu6/oa8MArLtf6IsWlA5l//vkHy5Ytw9SpU/Haa6/hyJEjmDx5Mry9vTFmzBjJ98yfPx9z584t45oSkbtyVvdOWqbpIMaW/ci9fHf0X7y86aTR9oahAVg/vj2q+vs45sQZKcD6YUDqn8ZlAeHa1pdq9R1zbgdx6dWvvb290bZtWxw4cEC/bfLkyThy5AgOHjReghyQbpGJiIjg6tdETuSqs3LUGhGdF/5usmVEAKBS+mLf9O52r+/BS7fw+BeHLO739fgObJEpJzQaEfN+PINVB5KNyoa2qYl3HmkGLw8HtICIIvDH/4DtU6XLu70BdHnZ5VpfysXq12FhYYiOjjbY1rhxY3z33Xcm3+Pj4wMfHwdFskRkNVeelePM7p2YqGCEKX2Rmp4nOU5GF0TFRAXb9bxU9jLyCvHkyiM4evmOUdm8QU0wumOkg058/V7ryynjssCawOgfgGr1HHPuMuTSgUynTp1w/vx5g20XLlxA7drOS7xDRPK5+qwcZ3bveCgEzB4QjQnrjkEADO6Rru1n9oBol2i5Ittc+i8LfZfsRYFaY1S24ZkO6FDHAS1toggcWQH8NE26vPssoPNUl2t9KQ2XDmReeuklxMbG4p133sGwYcOQmJiIzz//HJ9//rmzq0ZEFrjDrJyQAHkL6cndz1p9moZh2cjWRi1WKhdpsSLb/Hb2Bp5a/YfR9uoBPtgysRNqBPnZ/6Tp/wJfDQPS/jIuU9YCRm9x6npIjuTSgUy7du2wefNmzJw5E/PmzUNUVBSWLFmCESNGOLtqRGSBO8zKcYXunT5Nw9ArWuWSY4hIPlEU8dHvf2PxzgtGZT0aheCTEa3h62Xn7LeiCCR+Afz8inR5j9lA55cAR6+75GSlDmQyMjLw+++/o2HDhmjcuLE96mTgoYcewkMPPWT34xKRY7nDrBxX6d7xUAgc0Oum8grVmLT+OH49e8OobErP+nixR337L+CY/i/w1VAg7YxxWVBtYNTmctv6IsXqQGbYsGF44IEHMGnSJOTm5qJt27ZITk6GKIrYsGEDhgwZ4oh6EpGbcXa3jVzs3iFbpKTnYvAn+3EjI9+o7PNRbfBgE5V9TyiKwOHPgPjp0uU95wCdppT71hcpVgcye/bsweuvvw4A2Lx5M0RRxN27d7F69Wq8/fbbDGSIyjm5U6ldodtGLnbvkFxHkm9j6HLj9B8KAdgx5QHUDw2w7wnvXgHWPQrcPG9cViUKGPU9EFzHvud0M1YHMunp6QgO1v7DEx8fjyFDhqBSpUro378/XnnFRD8dEZUL1kyldpVuG7nYvUPmrDt0GW9sOW20vWVEEFaPi4HSz8t+JxNF4NAyYMdM6fJebwGxL1TI1hcpVgcyEREROHjwIIKDgxEfH48NGzYA0C7w6Ovr3CZiInIcW6ZSl/duG1dN9Ef2UaTW4I0tp7HhyFWjsjEda+PNAU3s+3nfuQysGwLcumhcFlwXGPkdEMz1BkuyOpCZMmUKRowYAX9/f9SuXRtdu3YFoO1yatasmb3rR0QuoDRTqctrt40rJ/qj0rmTXYARKw7jTIrxWn3vPtocw9pG2O9koggc/AT45XXp8gffBjpOYuuLGTYtUfDHH3/g6tWr6NWrF/z9/QEA27dvR1BQEDp16mT3SpaG3BTHRGQa0+kbMtU6pXvUODvRH9nmXGoG+izZK1m2+flYtKpVxX4nu5MMrH0EuH3JuKxqfWDkt0CVSPudzw05dImCtm3bom3bthBFEaIoQhAE9O/f3+bKEpFrc4ep1GXFHRL9kXV+OpWC5786ZrS9VnAlfPtcR4QE2mnYhCgCBz4Cds6SLu89H+gwga0vVrIpkPnyyy/xwQcf4OJFbT9e/fr1MWXKFDz99NN2rRwRuQZ3mUpdFtwh0R9ZJooi3t1xHssSjFtEHmoehkXDWsDH004J7G4nAWsfBu4kGZdVa6htfQmqZZ9zVUBWBzJvvvkmFi9ejBdeeAEdO3YEABw8eBAvvfQSrly5gnnz5tm9kkTkXO40ldrR2Drl3rLzi/Ds2qPY9/dNo7LX+jXC+C517JPAThSB/UuBX2dLl/dZALR/jq0vdmB1ILNs2TJ88cUXePzxx/XbBg4ciObNm+OFF15gIENUDrnbVGpHYuuUe7pyKwcPfbQXGXlFRmVrxsXggQbV7XOi2/8AawYDdy8bl1VvDIzYBATZcbAwWR/IFBYWom3btkbb27Rpg6Ii4y8IEZUP5X0qtVxsnXIv+y7exMgvDxttD/DxxI+TO6N21cqlP4koAvs+AH6bK13e9z0gZjxbXxzE6llLL7zwAry8vLB48WKD7dOmTUNubi4++eQTu1awtDhrici+mDvl/qwlQLp1irOWnG/F3n/w9vazRttj61bF56Pbwt/HDmsm37qkbX1Jv2JcFtIEGLERUNYs/XkqKLnPb5sCmTVr1iAiIgIdOnQAABw+fBhXrlzB6NGj4eV1P7thyWDHGRjIEJEjMI+M6yko0mDappPYevK6UdmzcXUwvXcjKEobdIsisHcR8Ptb0uX93gfaPc3WFztwWCDTrVs3WfsJgoDff//dmkM7BAMZInIUtk65hrTMPAxbfhDJt3KMyj58vBUGtggv/Ulu/g2sGQRk/GtcFtoMeOIbQFmj9OchPYflkdm1a1epKkZEVF5wfSbnOnn1LgZ9sl+ybPvkzmgSrizdCTQabevLrrely/svBtqOY+uLk9mhk5CIiKjsfHf0X7y86aTR9oahAVg/vj2q+vuU7gQ3/wZWDwAyjbuooGqubX0JtEMrD9mFrEDmkUcewapVqxAYGIhHHnnE7L7ff/+9XSpGRESko9GImPfjGaw6kGxUNrRNTbzzSDN4eShKcwJgz3tAwjvS5Q99oG19IZcjK5BRKpX6BEFKZSmb6oiIiGTKyCvEkyuP4OjlO0Zlbw1qglEdI0t3gv8uaFtfslKNy8JaAo9vAAI5eNuVyR7sO2/ePEybNg2VKlVydJ3sioN9iahC06iByweArBuAfyhQOxZQ2Cn1vgNd+i8LfZfsRYFaY1T2zTMd0L5OKcYmaTTA7oXA7gXS5QOWAm3G2n58sgu7z1ry8PBASkoKQkJC7FbJssBAhogqrDNbIcZPh5Bxf6yHGBgOoc9CIHqgEytmmqn8L9UDfLBlYifUCPKz/eD/nQdWPQRkpxmXhbcGHv8aCFDZfnyyK7vPWrJyljY5EaeEkjtzxe+vK9bJojNbIW4cDREiitdUzLgObBwNYdgalwlmRFHEqC8TJdc/6tEoBJ+MaA1fLxtbkTQaIGE+sOdd6fIBHwJtxth2bHIJVs1asstCWuRQTNJF7swVv7+uWCeLNGrkbnsFPqKIkvGWAoBGFJG37RX4Nerv1G6mrPwiNJ29Q7KsY52qWD++ve3PnbSz2taXHOPgCDXaAo+tBwJCbTs2uRTZXUsKhcJg0K8pt2/ftkvF7KUidS3p0qaX/ECZNp3cgSt+f12xTnKo/9kDjzUDLO83ehs86jxQBjUydPFGJnp9sEeybGbfRng2rq5tB9aogV3vAHvfly4f9AnQaqRtx6Yy55CEeHPnzuWsJRel1oiYu+2M5CJ2IrT/8M7ddga9olWu3yROFY4rfn9dsU5yXfrnEhrI3a8MA5kfTlzDixtOSJaVagBv2llgZT8gV+KHdM0Y4LGvAH/3Gt9J8lkVyDz22GNuN9i3okhMum3Q9F2SCCAlPQ+JSbeZiZRcjit+f12xTnKliUGyAhm5+5XWjO/+xIYjVyXLEl/vgZAAX+sPqlEDv78N7DOxpt+gT4FWI6w/Lrkd2YEMx8e4trRM0//g2rIfUVlyxe+v3HP9eibV5QIZj8hOuL4vGCrcNhojAwAaEUhFVXhEdnJYHQqKNIhd8BtuZhUYldUKroTfX46Dpy0J7G78pW19ybtrXBbRARi+DvCvbv1xyW1x1lI5IfcXjU2/fMht2Dq7xhmzcoqf82Zmvqz3lPz+OrLecv+/svnENbzWP1ryvLbUzx7XFFO3Ol73ehrvFL4LjQiDYEZz75/y9xVPYggUUGtEu37WKem56DhfesHgsbGRmDOwifUH1aiB3+YB+5dIlz/8GdDisfv7Ju21S94ct5ytdo87191asgMZjcY4KRG5jpioYIQpfZGanifZpy8AUCm1X2Yqn2ydXeOMWTlS51QI9x+yJUl9fx1d75ioYARX9sbtbOMWheJuZxdKdi/ZUj97XZOHQkDXwePw/PoCvOm1BuG4P3YkFVUxt3AUdmha4/svD9vtnu27eBMjvzwsWbbc1kHRqae1rS/56cZltWKB4WuBytXubzuzFYifDhTLm4PAcMCGvDluOVvtHneuuy1kz1pyVxVx1hIAg2DG1WdYUOnZOrvGGbNyTJ3TFKm6lFW95237C//bn2xxv6WPtcSgljX0r22pnyOuKf50Ct7aegoRWScRgrtIQxASNY2gwf0undLes6W/XsQHv16QLNs1rSuiqlW27oDqQuCLbkDqKeny4q0vxZ3ZCmwcDZi6g1bkzXHX2WqAe9e9JLnP71KssEWupk/TMCwb2RoqpWGTuErp61ZfXrKOpdk1gHZ2jbpEc4et73NUXXVKtn6X/P6WZb17RcvL8lq8G8qW+jnqmvo0DcOeGb3wwrgnsccnDoc00QZBjK3H12hEDPp4HyJnbDcKYjwVAs7O64PkBf2tC2L+/hWYowTeqmYcxNTuDLxyCZiTLh3EaNTalhhzdzB+hnY/C5zx/wt7cee6l4ZVs5bI9fVpGoZe0aoK0zdKts+uccasHEvnBLTdS7P6N0a1AB/J729Z1lvXZWvqfFJdXrbUz5HX5KEQoBAE3M0tLPXx03MK0WLeL5JlfZqosHxUG6vqhqIC4POuQNpf0uVyZx5dPmDYnWREBDKuafeL6mL2UO48W82d614aDGTKIQ+FUK6+pGSerTN+nDFTSO6xqgX4GHTV2HIMe9TbQyFg9oBos122swcYDvS1pX6OvqbSHv/0tXQ89NE+ybK3BjfFqA61ravQxZ3AV4+aLn/pL0BZU/7xsm7YbT9XnEEnlzvXvTQYyBC5OVtnrDljpps9zlnW9dZ12ZYcPKkyMXjSlvo5+ppsPf7XiVcw83vpsSo/TOyEFhFB8itRVAB81gX475x0eZeXgR5vyj9ecf4ylxqQsZ87zwB157qXBgMZIjdn64w1Z8x0s8c5nVFva7psbamfo6/J2uNPXH8M2/9MkTzW8Vm9UKWyt/yTX9gBrB9muvylM4BSuvVNttqx2tlJGSmQHicjaMtrx1o8lDvPAHXnupcGB/sSuTld9wcAlHysmur+KM37nFFXex/DFrou20Eta6Bj3aomj29L/Rx9TXKOP7NvIzR+Mx6RM7YbBTFNawTin3f6IXlBf3lBTFE+8FEb7eBdqSAmbrp24O6c9NIHMYA2T0yfhfdemLjCPgtk5ZNx1vfLHty57qXB6ddE5YS755Gx9pyunivDmXlkrKlTdX9v/CeRfRcAXuheDy8/2FD+Cc7/DHwtMatIZ+pZbcuIo0jmkamhDWKYR8Yt6l6c3Oc3AxmicsStMvsWFeHc4R3IvXMNflVqoFH73vDwtK6329Wzlzors6+c4/929gZW7EuS3Gf1uBjENZCZ5r8oH/i0A3D7H+nyuBlAt5k21rYEjVo788hc1l45+8jk6t8vc9y57joMZO5hIEPkguyYgZWs8/aPZ0wGMPtndEeNID95Bzr7I/CNmanRU88BgXb89c/vTIXDQOYeBjJELsaOGVhJHrVGRM/Fu5F0M9uorEolLxx6rQc8FQrLv+AL84BP2gF3r0ifqNsbQNwr9r8AfmcqJLnPb85aIqKyYzEDq6DNwNqov83dAXTfzax8tH37V8myR9vUxPtDWwCQMabi7Dbgm5GmT/TyBSBA5hRoa/E7QxYwkCGismPHDKxk2h/Jt/Ho8oOSZYuHtcAjre8nmzO1Ns+d9Aw029QZ+Pam9Em6zwIemGanGpvB7wxZwECGiMqOHTOwkrEVe//B29vPSpbtmPIAGqoCDLZJrc3TV3EYy7yXmj7JtIuAf4gdaisTvzNkAQMZIio7dszASlqiKGLUl4nY97d0y8mpOQ8iwNdLsky3No8PCrDb5yWohDuS+11uPR21B75mtzpbhd8ZsoCBDBGVHTtmYK3osvKL0HT2Dsmy2LpV8dXT7SEI5qfbep7bjGTfl02Wt8lbhltQYmmtlrByNSX74XeGLGAgQ0RlR5eBdeNoaGecSCzDKDMDqzuyR26Pizcy0euDPZJlM/s2wrNxdc0foCAH+LAlkHUD7SSK3yl8HJ+rBxhsc+raPBX8O0OWMZAhorIVPVA7XVYyJ4j1GVjdRWmzrf5w4hpe3HBCsmzjsx0tr59z6lvgu6dMFrfOW47bMJzi6jJr81TQ7wzJwzwyROQcdszA6upMzQzStcUsG9naZDAz47s/seHIVcmyxNd7mG8tKcgGljQHckzMPOr9DuIDHsGEdccASLZ1mK1bmatA3xliQjw9BjJE5ExqjYjOC383aIkpTtfqsW96d303U0GRBrELfsNNiTWQagVXwu8vx8HTw8yav39uAr5/2nT5K/8AlavqX5aXtXmofGFCPCIiF6CbGWSKCCAlPQ+JSbcRWa0SOs7/XXK/sbGRmDOwiekT5WcBS5oCudIzj9B7PtDxecmiPk3D0Cta5fZr81DFxECGiMiB0jJNBzHFPf7FIcntyy117ZzcAGx+1nT5q0lAJctjXDwUAjrWrWpxPyJXw0CGiJyuPKzUa4qtM34SpnVFZLXK0oX5mcDiJkB+unR53/eA9s/YdF4id8NAhoicqryPz4iJCkaY0hep6XmSWVCK81QIODWnN/y8TQxgPbEe2DLB9AFktr4QlScc7EskU3lpNXCZ69CokZiwDet+PYI0BCFR0wgaaAewuuSMmVKIP52C5+7NDJLSKiIImyd2ki7MywAWNwYKsqTL+70PxIy3Qy2JXAsH+xLZUXlpNXCZ6zizFWL8dMRkXEeMt3bTdTEYcwtHY4cmRremMeZuO4Ne0Sq3DBh1Tl9LNxnEBPp64t1Hm0vf+2Nrga2TpA8qeACvXgL8qtixpkTuiS0yRBaUJgeIK3GZ6zizFdg4GiJEFA9PNPcqNqFwCnZoYvTbvx7fwS0HoX6deAUzvz8lWTa1VwO0iww2bg3LywDebwAU5Uof9KEPgLbjHFBbItfDFhkiO5BaHVjHnVoNXOY6NGptdtYSQQwAKARtMDPbay125rfVdzPJnfXjSNZ0x01cfwzb/0yRLDvxZi8EVfI2Lji6Gtg2WfrkHt7aFaf9gmysPVH5xkCGyAxrcoCURauBreNbXOY6Lh8wTDFfgkIAwnELMYpzOKSJBuDkdX4grzsur1CN5nN/QUGRxuj9TWsEYuvEzlCU/Jzy0oH36gFq46R3AICHlgBtn7TXZRCVWwxkiMyQ2xpQFq0GpRnf4jLXkXVD1m4huOsS6/yY6o5LTc/DhHXHMG9QE8z64S/J907uXg9TH2xoXPDH/4AfX5I+oacfMO084KssXcWJKhAGMkRmyG0NcHSrgaUHqqXxLa5yHfAPlbVbGoIAALMHRDuty85SdxwAySBm9bgYxDWobrgx9y7wbh1AVEufbODHQOtRpakuUYXFQIbIDEs5QMqi1cAe41tc4ToAaBf5CwwHMlIAiZpoRCAVVXHVvwWWDWzm1EHUlrrjSto/oztqBPkZbjyyAtj+svQbvP2BqWcBX05CICoNBjJEZngoBMweEI0J645BgPTqwLJaDUqxaq8t41ukxtLY5TpKS+EB9FkIbBx978z3ayJCgCAAmXFvYU9cL6cPnpbbzbZoaHMMaRNxf0PuHWBhpOk3DPoEaDWydJUjIj0GMkQW9GkahmUjWxuNT1HJzb9yZqt2pk7xQa6B4doHevRAi+eX+0Dd//d/SMvMQ/LNbHydeAWpGfn6Mt1YmlJdh71EDwSGrTG6J0JgONBnARrKuCdlwcdTXqAZHlRJ+x+HPwd+fsXEwZTA1L8AnwA71Y6IdJhHhkgmm2YM3cuZYtyNcu99w9ZYDGYOXrplckFBuYrninGZVY5L0UrlSH8k38ajyw9a3E8A0CCwEDsKxpjeafByoOXj9qscUQUi9/nNQIZkc5nU9lZwRJ11x0xNz8Xt7AIE+/tAFShxbI0aWNLU5HRjEQIKKqngOeUkPK4lmnygqzUiOi/8XdZaPeboxsHsm97dpnvg8M/fxsBGrRFx6NItHPznJgDtCs4d6lS1um4r9v6Dt7eflbXvkx7xmO21RrrQrwow5TTg42/V+cslFw1WyT2Uy4R4CxYswMyZM/Hiiy9iyZIlzq5OheIyqe2t4Ig6Sx3T5LEt5EwRIMInJwV33qmHKsi4X1Ci28ncOB1rlCZXjMM/fxu73+JPp2DG96dwN6dQv+3jXX8jqJIXFjxiebCwKIoY9WUi9v19U7L81JwHsf/vm5i77Qxy0//DCd9nTR/s4c+BFsPNnq9CKWWXKpFcbtMic+TIEQwbNgyBgYHo1q2b7ECGLTKl5zKp7a3giDqbOmbJ4+uPfepb4LunLB5XFAHBoPFAutvJXBBljaWPtcSgljVk7+/wz9/G7jdLCzECwHITdcvKL0LT2Tsk39OpXlWse6o9BN2HcvATYMdr0ieoVA2Y8ifgXdlsPSocO3SpEpWrFpmsrCyMGDECX3zxBd5++21nV6dCcZnU9lZwRJ3NHbMk/bFl5kwRjKpwr5bxM4BG/fVN8X2ahhmMb7mQmolPEi7JOkdxNzPz8cOJa7K6hxz++RdbssDkGUrcB1295myVTkRXXMm6XbyRiV4f7JHcd2bfRng2rq72RfYt4L06pg885Eug2aMWz18h2fiZEtnKLQKZiRMnon///ujZs6fFQCY/Px/5+fdna2RkZJjZmyxxmdT2VnBEneXmFDE4dpT5nCkWj5RxTds9FdVFv9VDIejr/OXef6w8pnYJgLeKjQOx1D3k8M/fQvebqfuQmHTbYFaWKbq6pWXm4cUNJyT32fhsx/v5c/Z/COycJX0w/1Bg8nG2vlhi42dKZCuXD2Q2bNiAY8eO4ciRI7L2nz9/PubOnevgWlUcLpPa3gqOqLO115eWmWeQM0WEAKFYMKMRtUGFRWZS+gdXllh80AJNiXjKUmZgh3/+MpcsKLmfNeczNeMr8fUe2kzGWf8Bc8wsCcDWF+vY+JkS2Urh7AqYc/XqVbz44ov46quv4OsrL3X6zJkzkZ6erv+7evWqg2tZvrlMansrOKLO1l6ffv97OVMKKhl2M92GzPFaZrqnVEo/k2UlmQqadHHN3G1noC4Z5aAMPn+Z3W8l97P1fLWCK+Hv/+uL5AX9EXJyuTaAeb+e8Y4B4cBrKcCcdAYx1rLxMyWylUu3yBw9ehRpaWlo3bq1fptarcaePXvw8ccfIz8/Hx4ehn2sPj4+8PHxKeuqllsuk9reCo6os+6YlrqXJI8dPRCeDfrh+YWfwDM7DWkIwh+aBtjjMwUq3DYRZAjabqnasaWqU5CfFyZ2q4f/+8n0tGJz3UMO//wtLFlg6j7ERAVDFegjq3sJAJ7sFInZA5oAWWnAW1VM7zh0FdDkYdnVJwk2fqZEtnLpFpkePXrg1KlTOHHihP6vbdu2GDFiBE6cOGEUxJD96ab+AvdnqeiUaWp7KziizrpjynmH1LE9PD0xcNAwbNPE4rAmGkXwxNzC0QCMu3v0teyzwOxgSEt1EgAsGNIMIYHyAnup7hqHf/667jdzZ5C4Dx4KAXMGNrF4+GcfqIPkBf0xOyj+XutLfeOdAmsCr6dqW18YxJSejZ8pka1cOpAJCAhA06ZNDf4qV66MqlWromnTps6uXoWhS9GvUho256uUvi459RpwTJ11xwxTSndrhFk4tu79oYHa9+/QxGBC4RT8J5QYJBsYLnt6qqk6Fa9LabuH7H0v1RoRBy/dwg8nruHgpVtQNxqgvd7AEsexcB/Op2aZPEegryf+92gEZiZ20AYwv80z3mnYGm3wMvUvwEt+Nx3JoFuGwsrPlMgWbpNHRqdr165o2bIl88g4ATP7Gh7TYmZfCfN/OoPP9yQZNLh7QIN5LdMxoomvzdlPzV2npczAcjP+2uNemk2sFx1iMQusRiPikWUHcOLqXaNjKwRgfJc6eCL/G9Q++YF0BYJqAxMTAS/XGdNVrjGzL5UClyi4h4EM2YM9HuLzfzqDz/YkmSx/9oEozOwXXdqqStIltQOkV74ui5a10iTWS88pRIt5v0iW9WmiwvLBNYBFDU2ffPg6oPEA6ytNRE7DQOYeBjJUWvZIz19QpEGjWT9LjIe5TyEA597qC29Px/T4OnOZCV2rkKmByaZahU5fS8dDH+2TfM/bg5tiZN7XQMJ86ZNWiQKeP1ShW1/csRWVSKdcZfYlchZTrQiW8q+UtPZgstkgBtAO+l17MBlPdTGTUbYUSmYGLssHm7WJ9b5OvIKZ35+S3Hf72HposiEGiDdxsMfWa7PGVnDuuD4akS0YyBCZYM/0/Jdv58g6p9z9bFU8M3CZ0aih/mcPBir+QBqCkKhpBI2JeQbvxp/DcYnxLwBwrvsx+B54H9ggUVi1PjBhP+DJ1AuA/QJwInfAQIbIBHum568dXEnWOeXu56qMujLy9sFjxwx0zriOzvcSEV8XgzG3cDR2aGKM3l8yiOmqKsSqu2O0Lw5InPDxb4CGfex7EW7OHddHIyoNBjJEJtgzPf+ojpH4v5/OWhwjM6pjpMzauZ6SXRm9FYno4L1E//DUUeE2lnktwZiC6dgrtpA81oZ6O9Hh35XAXYnCag2B5/ay9cUEd1wfjag0GMgQmWDP9PzengqM7xJldtbS+C5RDhvo62gluzIU0GC21xqIEmtK/appg2cKXzY6Rhhu4aDvC9oX/0qc5ImNQIPedq13eeSO66MRlQYDGSIT7J2eXze1+ou9SQYtM9r8J6anXrv6zBOprowYxTmEC7cN9nu7cARWqI0H4b7quQHPe26VPnhIE+CZBMDT+gUyKyp3XB+NqDQYyJDbcvQDXpeef8K6YxAgnX9Fl77/4KVbSMvMQ7XKPtCIIg4n3QYgomOdauhQt6q+Xq/2aYzOdavju+P/IqdAjXaRwRgTG2myJcbmmScaNdTJ+3Hpn0tIE4PgEdkJMXWrA4DZe6bWiEi89B/UyfsRItxF3Tp14RHZyWwSM6mujJB7fUJqUUDPgveRJBrWNRw3ccB3sun6j/gOqN/TdDmZFBMVjKBKXribU2hynyqVvFxqfbQKhUkC7Y6BDLmlsppaqkvPX/JcqnvnAmA2P8rHuy4hqJIXFjzSDACMjnPqWjoigv0k62zzzJMzW5G77RX45aaiAYAGAK7vC8YrinH4XWhv8IArfs/iT6cgYcv/MLlwxf3WlH1Arp8KfgPeM5lWXqqL4ppYFZF56422z/Bcj+c8f5Q8DkKbAc/sAjy8pMvJbsp18jBXdmYrED8dyLh+f1tguHZtKi7bYDMmxCO3U5oMsbaSav3ZeSZVsh7WMFVnWxPI4cxWiBtHQxRFg7Epuq6sCYVTDGYL6XZ55oEoXN73DT71WgIARu8VBAGCiTVyDl66hce/OGTyGmvgP+z3fdFkOUZ+B9Rj64u9WPo8dL4e34GDfcvSma3AxtEwDiPv/Z+Na1AZkfv8ds+RhVRhWZpaCmhbPdSWss9ZSZd/ZVDLGtp//DVqbP1hIwYoDqCD4gwU0Nh0XFN1tmbmiZ5GDTF+OkSIRgNsda9ne601qKt47+/LvZfwptcag32Lv1eECDF+hrZZvISYqGAE+Bo37r7m+RWSfZ+QDGJEAOoh/9Mu2mhFEGO04KSdP+fygIN9XZBGrW2JMfcvl4n/f5Fl7Foip7B1fItLTC09sxVFP76CTwtTARm5USyRqrNND6PLByBkXIepu6gQgHDcQoziHA5pDAcWtxWMB+cavBcAMq5p+/ajumjrLYoY9WUi9v19U79fTSEN+3ymmK3zdbEq5haOwthKXdHR7J6GmKlWHg72dUGXDxh2JxkRjf7/RfIxkKEyV5oHktN/bd5rHvYu8ctKlxulZNeNNYrX2aaHUWaKvPdIJGeR2iYp6way8ovQdPYOg82zPNfiKc+fJd9yUlMHC4seQzVkGGT27WfFZ8RMtfLZe7Yd2UHWDfvuRwYYyFCZKu0Dyam/Nos1D5ds9VAI2rEks73WYmd+W5Mp+M0pXmerH0ZntmqbpmVIQ5CsbSVd1NRAr6/8AGiDmAjhBvb6vGRy/ycKXsMBTVOT5XI/I2aqtY7c2Xa8V2XIP9S++5EBjpGhMmOP8S26B7ypf4IFaFt3HPJr00LzsEIAwgVt1401pOqsexjpykvuDxR7GOkGEebcMnsejajt1knUNDIq+0NshOtisGTm4R/UHRGZtx69Ct7TntdzNZJ9n5AMYsSaMVC/cRMdfb7HQRNBjLWfkU3jhSo43Ww7ldIwWFQpfdl65Qy1Y7Wzk8z9yxVYQ7sfWY0tMlRm7DG+xam/NmU2+8rupoH5Olua+t2naZiFQYT36QKUuYWjDFqLdGd8qktdzNs3Gp96LYHmXjbeGYVPY4O6OwCgtpCK3T5TTR7/sYI3cEgTjbD/fDH73E27fkZO7050U85c7ZxKUHhop1hvHA2Y+n9FnwXMJ2MjBjJUZuz1QJL1gHcEmc2+5rppdF1QOlUqe+HtQU1N1tniw8jiIEKt2wjEfMUzOOzbHiiWR6b4PYuv9Rymb/bBr9lRuAPtVMd5nisx2nOn5DGPaBpgeMGbBoFR8S5Ce31GHLxqO6esdk7Sogdqp1hL5pFZwKnXpcBAhsqMPR9Itv7aLFU24NqxEAPDgYwUCBItICIEFFRSYeJjo1Dlj2v4+XSq0T4lu25uZxfire1noVAIJh/u5h5GmsxUWf3DtzvPwbs9xgGQzuybkp6L59YdA9ACkUIKjvs8Z/JY6tE/ovM3BZKta8XHrOyb3t0uLQIcvErlRvRAoFF/Zva1MwYyVGbs/UCy9tdmaafvxp9JQ0LOSLwjvgsRJfOtCBAA+Dz0LmIbhOLVzX/JrpetM2/iT6dg67br+FTGvg3qNdBXuPg923fxJkZ+eRgA8H+eX2KE52+S7z+kaYwRBa9DAwWm/KNCSvoFk+cq2UVY2hYBDl6lckXhwSnWdsZAhsqMMx9I1s6WKtlycyc7HxPXH4eIlrijmILZXmsQjvuDS3P9QvVp/BMv3TI7Fqgkq2be3Fun5eTZc1i17y7+0DTAdZ9gqHDbKJGdlqBtui4xiHDprxfxwa8XUEe4jmTfaSZPNzT/TRwR7w8OFgCsPGB6Be/i7DlmxWndiY7AtXaI7IqBDJUpZzyQrJ2+K9Vyo81uq7VDE4Od+W0RoziHENxFGoJw1bsF9jTqBQ/Y9gCXlciv2DotLQBs8NYm4ttaFItnPH/UD9K9f0xtK5FuEKFGI+KRZQdw4uodLPD8Asm+CZKnOaCOxsjC1ySnkIuA2cUIi7P3mJXi3YmpGXm4nZWP4MreUPp5Q60R3aNFhmvtENkdAxkqc2U9m8Ka2VLpuQWSLTclx7ZooDDMjptRqA9CSvMANxkEmVinRYXbeMbzR3xe9BAGeh4waCUqqKSCz0PvIj2yL1rM2I66wjX85vMKYKp643bgh9sReHHDCYv1DPLzQnpuYZmPWfFQCEjPLcDCn88iNSNfv10V6IM5A5u4dsuMqbV2MlK020ustePo1d2JygsGMuQUZTmbQm4LSWp6Lt7dcd7mRSB157E0FsgcySDIzBRr3SyogZ4H8UD+ErRVXNC3EnXuMhDvr7mI9zwfQbLvHukT1ukGjPweUGhbX0IKzeei0XmyUySW/HrRKV2E2kHJhlIz8vHcumNYXhY5UmzpGrK41o6gTWjYqD+g8OByDERWYCBD5YvEQ0ZuC8ntbOmZOOYooNF3MdXLBqBRwUPhYXIskClmWzHkJOLDLbRVXNC3EtUT/sWkvTGYZOrSn9oJRBgvpSB3QPak7vXRUBVQ5l2EM74/ZXafGd+fcmyGX1u7hqxYayc+ux6XYyCyAgMZKj9MPGRiei9AmNLf4sM52N/HqtP1ViRqB/3qFlvc+TFwWPtQ69N0IJaNbI0Z35+yOKbEYiuGzER8VzTVscjrUwzx2Ce9Q90ewIhv9a0vUqwZkF3WXYSHLt2yeC/v5hTi0KVb6FS/mv0rYGXXkAGZn6EmMxVztxdwOQYiKzCQofLBxENGzEiBYtMYfNpxKR7ZVU2yhUQEMLxtTdzKyEYHxRmE4jaqChm4JQbiBoL1ixwW11uRiGVeS4zrkXEd2DgK6kdXIcC3C0TRcntMoK8nxnaKQoCvF344cc0wINCozT4E80QvPFIwBz/5vI4Dvi9K7nOi+SxUCqyKunXqwkNG+5BuQPacrX8ZjEMJlRiHUpZdhAf/uWl5p3v7lQxkLI43sdRdZGXXkBGZyRTPZlZy/urudmLunttj/A/HEJEOAxlyf2YeMgJEaAA0OvEOlo34DTM2n5H8VX9213rM9lqDp72N1+u5LgZjbuFo7NDEQBAAQdRgttcaACVzyRTz7TisLZiEdE0Hi9VPzyvC0t8uGmwLU/ri09b/otVfCyS7JK5oquOYWB+DPQ7gJ5/Xjcr3oDV+QDe8LK5Eyz/f0m7cB+T6qfTTxC0ztcqTs8g9v+F+FsebyOkusqJrSDJHiG6tnYwUSAdD2mnyf1dqBsB89xng+ssxmLvnAEo9/odjiKg4LhpJ7s/SGBIAfrmpuPVXAtIlghhd60oYpBcdVOE2lnktQW9FIkQReLdtFsIFU3lbtDygwadeH6K3ItHaqwEAtMjcgxYHJkMscV0HihoDAGop/sNgjwNG7xuUPw+ReevxVcEDeE9cBFWJa/LJSYW4cbT24W2CLudOaobhw/JGhnaMRvzpFJuuqbTktkAU3093LSVbOXTjTY7vWK1tySv5/dF1F+nuk8yuIZP76dbaAWAyQOyzACGBlWWdxpWXYzB3z59bdwzPmfk85Hy3LH2mzvp+kvMwkCH3J/Mhk3jqrNFvYQXut64IJgITXcAy22stPKDBn+fOy67abK+1UEAje39dnd7U1enetr80tQEAsZ5njfY/pq6LyLyvEJm3HifFegbXVDLY0ubDESHGz9C2ZJVQ2hXK1RoRBy/dwg8nruHgpVtmVzK3Voc6VRFUycvsPlUqeaFDnar6upi7FgU0CD84F6K5q9XdJ5ldQ2b30621E1iixSAwXD++xqmru9uBnO+PFDnfLbnHt3QMKn/YtUTuT+ZD5oYYZLQtRnHu/mBdM3Qzg9opzuFCTmXA2/L5dO+JUZwzzDljga5O6mL/FjdRXDba77PCfliofsJo/I6la1IAJrtBSrNCuaOb+z0UAhY80kxy+rXO/Eea6cdJWLqWdopzCIW56ebFuotkdg2VzKBsxMJaO+6+HIOle26OnPE/pfl+UvnFFhlyf7VjkesbapS0TkcjAtfFqkjUNDIqC8Fdq04VgrtI1DRCrq/MX+hWn0PEKx7fAAA8JJ5VpzSR+v/+S6wjmX1X9vnutWQVb0XZ/7e8AbUlx2iUVXN/n6ZhWD6yNVSBhl0rYUpfoxwylsaRWHWfZHYNyVpqQLfWTrNHtf9b4j26wdYqpeE1qpS+Lj/12h5jd8wdQ+7xXX0MEdkXW2TI/Sk8cKX9bNRPeN4oTb8uuJlbOEryoZ+GIKtOlYYgaKDAlfaz0XD387LfY0kTIRnbfV4zWf67ugW6e5xEM0WyxePKvib/UMlWFDmKj9GwdgmI0pI77dvSOBJr7hOA+11DkgODF9h1iYGyntpuL/YYu2PuGHKP78pjiMj+GMhQuVAv7gm8tj8ZkwtXGKTpT0VVzC0chR2aGH0W3OISNY1wXTS36KKWRtQe64imEcKUvqgX9wQQUhn49klAlB4Do3uPVEuQlojlXkvQx+OIZOlP6hjUFa6hoeIaunuclDyuLdekASAE1sCOrChM+Mo48Zo5Uon7nNHcL2fat6Xkfkc0jXADVRGC2xDkdhdZ6Bqyp7Kc2m4vpclqLWdpC7kJG111DBE5BruWqFzwUAjoOngcuuR/iMcK3sDkgkl4rOANdM5fil80MRAAjO8SZdQpoIECcwtHAwBMpXzRBQrz7rXq6McoNBkMDFlp9j1SLUFNhCQk+z6BZN8RkkFMv/x38HT+VPRRJKKB4prZuvh6GT9Ai1+T0RpRIiBAgKb3fMz90brlGEyN0XDV5n7deBNAujNIAwWud5x9r8yK7iILXUMVmaV7LvXfxV9bGv8j5/iuPIaIHIOBDJUbfZqG4ZORbXE5oDW2amJxSBMNDRT6sQUz+0Vj2cjWCCsx9mCHJgYTCqcgBdK/4lJRFRMKp+BkwAPGYxSaDgaGrdX+ci8mTdC+Z4dGuwxAFT9PrPRdjGTfJ7BdIu/LNnUH/cyjM2Ik/grsjJOxH0IocVxdXQ75dgIA5BQYzzwCgH2eHTFNMQ2pJa4pv5IKwrA1SPTtbHV3kqkxGq7c3G9pvEmr3mMsziQi65i758tHttaOcSrF+B93HkNEjiGIclKPurGMjAwolUqkp6cjMDDQ2dUhO7CU0VNueWp6Lm5nFyDY3wchAT6AWg3xygGE4DbqVs6DULkazmb54+9KzRASWNn8GAWNGurk/bj0zyWkiUEQasUCHh4ounoUcbuHmbyWvvnzUb9ZBwxvVwsQgJtZ+caZfS8fgCYzFWczK+HvSs1Qzb8SXt50EqkZeQZrPaUhSJ+FWBXogz2vdsfRpJtQJ+9HiHBXm9k3shOg8MAPJ67JWuV6Ure6qB8aYHaMhlojovPC3y029++b3t1pv5RLndmXrMbMvlRacp/fDGTIrUgNTg2u7I23BzVFv+bO+yWmq9eN9By0V5zF657r0FRiyjQAbFHHYkrhRGx8NtamvvyDl27h8S8OGa/1BMMsxF+P72ByjIXuGJaYO0ZxullLgPSUYf5SJiJryX1+c7AvuQ3dw7Jk5H07uwDPrz+GZ/+Nwsx+8vO12Ltez3hsxUzfDSb3652/AOfFWkh8vQeSS9HNkpaZZ3KtJ10W4gmFU5CW2dLkMew9aFLX3F+Wq2ETEQEMZMhNmJviq/PZniS0qBmEfs3Dzexl53qp1VB+9ziSfI9Lln+n7oyXCyegVgDw+6x+8PQo/bC0kMpeZjP3akRtRuHLlSeaPIYjEq+565RhInJvDGTILcjNGPrGD6fRu2mY4x+eV48AX/aEB4COEsUP5i/EBTEC4zx+wj8+I5CjCIUg9oE9xtfHeJyDh7nMvfcyCod6nAMQYnI/R7SiuOOUYSJybwxkyC4cPfBO7tTd29mFjktPrtEA6x4B/tklWbyp6AG8UvQsAAHLvRajj8cf+jL//Bt4fuEnGDhoWKm7WTyy0+y2H1tRiMjdMZChUnP0GjuAdVN37Z6v5Goi8GUvk8U989/F32JNAECC90uIVEgvYumZnYYJ646VfuCrPRYwLIatKETkzhjIUKmYGoCrW2PHXrNVYqKCEVzZG7ezCyzum3wzp9Tng0YDrB0EJO2RLN5Q1BUzisZDm15OxDmfMfAVCs0eUpcSv9Tp+u21gKGduct0WHepJxHJw0CGbObINXakHjZvD2qK59ebXvlYZ8ORK5jUvZ5tD6crh4D/9TZZ3CP/PVwSawAA+jZVYdnINog/9S9ufxsAlWhiSYBiSwrYJV2/bgHDjaMBU0N15S5gaCdl0SpnD+5STyKSj4EM2cyWNXbk/BqWetj4+3jg6c510L+ZCttPpZqtV0p6HlbtT0I1f0/UyzmFxgE5UASoTCc506iB1QOBy/skj7e+qDteK3pa/7prg2oY36UuOty7pj7NauL49TlQHZwse9HKtIxsIOmM7QnYynABQ0vKqlWutNylnuSe2NLnPAxkyGbWrrEj59ewqYdNVr4aS367CG9PebN+En9ebZQsTvuQX3j/IX/5ALCyr8ljdM9/H/+IxlO5Ey7cRMKFm6js44H3hjRHv+bhaN5rNF47chWTC780uWilTm9FIvrsnArkFAvIStZNjjJcwNCUsl752lbuUk9yT2zpcy5m9iWbWZMdNj23QDJAKZ75tVe0Cp0X/m71GkAlFU8WV/yZJN4bz4JqDYCbFyTfu66oB94oesqq8z37QBS6NgzF418cMrlkQPG6LffW1s3wcXnvVWnX9ynjVPv2zhDsKO5ST3I/pn58Mat16TGzLzmc3OywbWpXQdx7uyz+Gg7w9Sp1EKOAxmSyOEFXA4kgpmv+IiSL9/+xaV5Tie+ei0WXd3chNcN8nT7bk4SCIg0A7YrKhzTS2YWL1834N/+9OxE/Q9vKYkvwcWaria4mK1t6rOCqK1/ben5n15PcC1v6XANXvyab6bLDAsYP5uLZYY9eviNrLM3BS7dKXacYxTmEC9KDbktaXdQLkXnrEZm3Xh/ETO5RH8kL+mPrpM744/Idi0GMznfHrsmum+mqiUDGNW2LirXObNUO/i0exADamU0bR2vLHcCVV7625fzOrie5F2vGCZLjsEWGSkVOdtgfTlh+yGuVvpczBHct7hOXvxiXRZXBttXjYhDXoLrBNmt+nWfkFSG4shfuZBeavIr6flmAWsbBsqTz0JikUWtbYsz9LixNS48Z9l6zyVHcpZ7kXtjS5xoYyFCpWcoOK/dXbsc61bA+8aqsXDElKaDBy54bMdFTuuVhZVFvzC0aY7R9/4zuqBHkJ/kea3+dP9yyBv63P9loQrRO/br1AOmhOYbkJrzTuXzAuCXGQLGWnqgu1h3bAkes2eQI7lJPci9s6XMNDGTILsxlhzX1a7j4wNiiyiHoENVbdq4YnYbCFXzj/RaChGyjsqua6ni4YB5uQmmwvYqQjcNzH4a3t5fZY8dEBUMV6GvQvWRuMG/PaBXaRQUbtU7pFnKc82cQevoEQyXcluzTFSGgoJIKnhEdYVW7idwWHGtbemRyl5Wv3aWe5D7Y0ucaOGuJyoRuZD+g/TXcW5Focnr0/OR6+GxPksljKaDBK57fYILnNsnyWQVjsVbzoNH2oR4JeM/rC6tmBsWfTsFz9+otVefrYjDmFo7GnwEPYN/07vBQCPp8EjvPpOJ/+5MNjmdqRpUu38yEwin4M+AB6x6sSXuB1Q9Z3m/Mj3ZvkSnOXfJouEs9yT2U/LdNh7OWSk/u85uBDJUZXa6F5pl7JB/mxacg/1TUFq989yey8+8PKmksXMZGn7cQAOMlCE6gEZ4pmII0jfFnvNTrIwzyOAgE1rApWVz86RT8tPFzLBEWG9VZF4CcjP0QrXrf77pSa0STU8mlA6L7+Was/gdQowaWNLW8ZMGUU2WaY4aoomAeGcdgIHMPAxnXoi4qQtHiJvDOSTUxe+f+Q1cNBQ5dvAGv3W8h5vo6yb231nkTk880MtpepZIXXuxRB+1w3nJm35J1LPaLvZq/D6BWo9XmLvDLuyFZZxEChBKBgqW8JZbyzeiapHWtPBbpZi3dq1HxugGAUNr8NERkFlv67I95ZMgleVw9CI/iGW2NaAemXvplOeocm49OBZlGexwXGmNTnXew/q9c4IxhWc/Gofj4iVbw9dIFLPUMyi39YyP1y6qD4gw2eJseXyJIDKa1NEvBXL4ZQHp5B7PuLVmQu+0V+OXev78pYjA+9HoKXTXt0MfyUYjIRlxF3nkYyFDZkjngtO6h14y2vVjwPH7QdNa++CvXoGxqrwZ4oXs9CILpX0CWmn9NZeiUM6UbgMG12WuWgjXTNuM17TDxzvtoV6KlR8xX4Bsb1xLir0wicnUMZKhsWTm1+JCmMSYUvIg7kG5W/GxUG/RuopIsK87SgoGfPNEKb20/KznCJA1B8ipb7NoszWaQS25ApMswqjbR0mNLhlH2+xORO2BmXypbtWO1Y2AsmFLwPCLz1uOxglkmgxgACPQ1P4UasJxGHADe+OG0yQydiZpGuC4G6wf2GhO0A4lrx+q3WMp6LAAIquRlMsuvAG3QIHfapr0zjOoCv5LH1AV+8adTZB2HiMjRGMhQ2Ur9E8i+KVkkitpZQP3y/w9bdF1IFsjpepHzkL+dXWiyXAMF5hZqB9IaBzP3QpE+C4wGEuvylqiUhq0qKqUvlo1sjQWPNCt+hJJHtCpBmz0zjMoJ/OZuOwO16ciOiKjMsGuJHE9dCPwyCzi8zOQuRaICLxU+j22aWJP7SJHT9WKP9OA7NDGYUDhFO20aJXPfmJ7SbSnrsb0StNkzw6g1rTsc3EhEzsZAhhzn2jHgf30Adb5xWVQcMORLHDySiGd3ZCED/lYd2pqMmfYaeLtDE4Od+W0RoziHBpWyMfuJ7vCI7GRxSre52QyWAh257Jlh1B6tO64+SNjV60dE8jGQIftSFwI7XgcSP5Muf3Ql0PQRnEvNQJ+3997baF0QoyO360X3kDfXygBYzu0CaLuZDmuiMXZwa3jUsc+A19JM2yz+QH6sXQQ++PViqdcSKm3rjqsPEnb1+hGRdVw6kJk/fz6+//57nDt3Dn5+foiNjcXChQvRsGFDZ1fNLVj7q9OWX6m69+QnH8ID+0ZBIUos7xzVFWj/LFCYg59SlXh+3XaTx7MUTOjKmytzobqtwd6z7XD4yl0AAtrXViLo5h/Iv3sdPkHhyAhph5s5Rajm74PH2oTjYMKPJo9rbvmBHZoYg/M/EXoVUSk3oK5UT1aLjL1+/Zc8zp3sAry13fCBHFRJO/j5bs79MT/WdlVZCvxMte6oNSI+/v0iPvj1otF7UtLz8Ny6Y/j0iVbo19zyYG9rWHN/Lc1eszhFXaPW5gvKuqGdpRbRHrh6+P5rmUkXich+XDqzb58+ffDYY4+hXbt2KCoqwmuvvYbTp0/jzJkzqFy5sqxjVNTMvtb+6rTlV+qOPy8ja8urGKKJN1sXEQq8X/goPlEPNiob4HMcix5tjl/Fdvh9ywpMVf/PZDBhLtgAYFOZ7riW1j8ydf5cPxX8BrxncoyMvX79Sx1Hiq76U3o2QGS1SjYHTvN/OmNyvSsBxssnxJ9OwZytZwwW2JSiEICPH2+Nfs3t0/Jhzf01t2wEICOb8pmtQPx0w5XGBQUgau6/vrdeGLMoE5VeuVyi4L///kNISAh2796NBx54QNZ7KmIgY+pXp6k1fKzdH//+Ac2XD0q3vhSTI/rgmcKp2KdpZlT2uuc6jPf8Cffy4mKdYhBGqLcAkA4mPi96CM94/ihZXrwrReq9unKpsomFL2KW11qocBtSzy6NCKSiKt4qHIlPvJZKn18QJJcAsPq+mmDqOKZYvbyBled79oEozOwXLXt/KcvtsJCetffX0rIROl+P72Dc1adfAsLSVd5fL4zBDFHpyH1+u9X06/T0dABAcDCXRDfF2qmzao2IOVtl7F+YD/w4FZijBFb0MBvEXNVUR/O8LxCdv9IoiFnr9Q6SfZ+4F8RozyKKIh5X/wCg5CKS91/r9jdVLpgoE0rsV/J9b3mtRLggHcTo9gsXbuFtr/+ZPI4IEWL8DG23wz32msJs7jimWJszxprzCQC2nkwx+P5YWz+g9NO3bbm/Ng9i1qi1LTGyrvLePiW+D0TkOG4TyGg0GkyZMgWdOnVC06ZNTe6Xn5+PjIwMg7+KxNrEaB//ftFsd0BL4SL25w2Bx/+FAH98aVB2XFPX4PU+dVNE5q1Hl4KlyMD9rr8AZGO39xQk+z6BLh6njc6hEABPQTQbTHgIGpPlgqD9s7ZMIQDVBHnfj6pCpun6ARB0ay3dY68EdZaOY44t086trbet9bM10NKx5f7aPIj58gHD7iSLiq29RUQO59KDfYubOHEiTp8+jX379pndb/78+Zg7d24Z1cr1WPOrM/50iuTATG8UYo7najzh+bvkew/HfIjhe6phoOIAWnl/jBVFffF20Sij/TorTuEzr8WoLEhMvy6Piq21ZK8EdaXJgWPLtHNr612a+pXFe4vvZ/MUdZnrgxmx9X1EZBW3CGQmTZqEH3/8EXv27EHNmjXN7jtz5kxMnTpV/zojIwMRERGOrqLLkPvwqubvg2mbThpsay1cwPc+cyT3v1OjG6qMWg34KqG5dAvYcwj7NU0QmbfeaN8JHj/gVc9vTLaEuJqbYgCCId3iohGB2wiU13JTbK0leyWosyUYsSZnjK3n0+1Xmhw9ZfHe4vvplo2YsO6YdVPUrVwfrNTvIyKruHTXkiiKmDRpEjZv3ozff/8dUVFRFt/j4+ODwMBAg7+KRPer09IaPhC1Te/eKMQ7nl8g2fcJySBmfMFUdPT5HoFPbQZ8lfgvMx+vbT4FALgFpcG+H3stRbLvE5juZV0QoxGBIlEwuZaRRgTUosJkuShq/6wt04jAdbEq3ih8Uv+6ZDkAvFE41uxaSxoAYom1luR+DpaCDUvHkTouYN3yBtacr2S9ra2f1DFsYev9tbRshOQAZP36YFZ8CiW+D0TkOC4dyEycOBHr1q3D+vXrERAQgNTUVKSmpiI3N9fZVXNZlhYrBLQPOc2VA0j2fQIXfMfgCc9dBvvtVLdGs7wViMxbj52atpg9IBp/XU9H5IztaPd/vyLpZrbB/j95z0Cy7xN4yOOw1fXVTVxdUdRf+9pEMPFFUT+z5aKJMrHEfiXfN7dwFOI1HTChcApSYfjQS0VVTCicgnhNB5NrLWlnTQkQSqy1JPdzsBRsmDuOFLMPZBmsrbe19SttoGVrPYvr0zQM+6Z3x9fjO2DpYy3x9fgO2De9u+l7pvDQTqmWPFtJptfeIiLHcOnp14KJn/UrV67E2LFjZR2jIk6/BqTza9QOVGBd+LeISP5W8j1PFbyM3zRtDLb1bqLCjr9SjfZt5HkdX3nMRVUh03JlSubaKCbXT4WX0h/DDk0MHpTME1MVcwtHmckjoy0HpHLFmC+bVzgK8cWS3XlAg3aKc5jZOQj5vtXx3F4f3M69X29b88jM2foXUjPujxNSBfpgzsAmpc4jE6b0xaz+jVGlso/dU+3bIw9RlUpeEGGYnM/eGXTLNEuvrDwyNcyuvUVE8pXLPDK2qKiBDHA/42lR0j502Tdacp9f1G0wtXACslBJ1jGHta2Jt1rehc86Gf9QP/CKdk2liPZI3PMTfjxwAhdz/AAA1ZGBosohGDhwCCB46B9GljL7Kn0VGKn6F6GKdDRv1BCZ1eVl9oVaDfHKAYQId1G3Tl3szK6DuT+eN/sAVGtEHPrnFg5eugVRFBHo54XMnDxEZJ5Ayyr5qFfXcmZfqURxqkBfzBlo/YO2rNcHskdmaAAOr3OZ3hdm9iUqMwxk7qmwgUxhHrD9ZeDEOunyJzYivqA5Jqw7BsByhoy3BjXBqI6RiD+dggNbPsO8og8s12HIl0CzR2UlLusVrcKq/Ul4a/tZi4eVTFhmA0c/AO2VEI+IqCKS+/x2i1lLZIXkfcCq/tJljR4CHl4O+AQAAPpA+zCdteU0/ssqkHzLxmc76n9Z6x7M7RWVAW/LVUn8zxNtLCQuE6BNXNYrWoVqAT6WD4r7U2pLG4iUZrFGSywlbCt+3Vx1mYjIdgxkyoPCXG3W3ZPGU6EBAE9sAho8aLT593M38Ny9FpniQgN9sPn5TggP8tNvK/5gTtQ0wnUx2GJa/3EJ3vis1i1ZictW7U/CnRzpYKqkkABfu4+NsHfrjDUJ2xwVTBERVQQMZNxZ0l5g9UPSZY0HAoM/1be+6IiiiE92/Y33f7lg9JaejUPx8ROt4Otl3Mdf/MGsgQJzC0djmdcSaETpNYzmFo5ClkbEe/HnZF2KnC4lXX6UO9n5mLj+uO0rGJfgiAGj9kqIR0RE5jGQcTeFucC2KcCfG6TLR3wH1O9ptDmvUI0Xvj6OnWeMs41O7dUAL3SvZ3KWGGD8wN2hicGEwinaWTy4P4snFfdnGQHAiX/TZVyUZbqazeofjbe226/LxtQ4FluDIh17JcQjIiLzGMi4i392A2tMzBSKHgwM+gTw8TcqSknPxeBP9uNGhvEyAV+Mbote0fKyj5Z84CqgQTr8sbDwMVQVMnBLDMQNBBvNMrIX1b0WEqWft926bBw5jsXmdPhERGQVBjKurCAH2PYicGqjdPnI74F6PSSLjiTfxtDlB422eyoExE/pgnohARLvMq34g1k630sw5haOtnsQM6lbPXSqV00/ZuWHE9dkvU9Ol40jx7HYnA6fiIiswkDGFf2TAKwZJF3W5BFg0MeAd2XJ4q8OX8brm41XmG5VKwirnoyB0s/r/saSOTHM5MDQPZi3rF+OT72WGJWrcBvLvJZgQuEUfbdSSQG+nsjMK5K+LhMy8goNXlerLG9mk5z95I5PMbc6uDm6dPglx9+oLI2/seJzqSjKOocOEbkPBjKuoiAH2PoCcFo66y5GbQbqdpcsKlJr8MaW09hw5KpR2djYSMx6SOKXv0SW0lzfUFxpPxv14p6QTu0eHYI45QYg13htC4WgHeg722stdua3lWyZsTaIAYA1By9jzcHL+sG3Ab5elt8EyMqXX81fXlA0b9tf8PNS2DRWpk/TMPSKVsl/CEtljw0M16bIr6DZYss0ey8RuR0GMs526Xdg7cPSZU0fBQZ+aLL15U52AUasOIwzKcYrM78/tAUebVNT/0s2NSMPt7PyEVzZG9Hpu9Fg90QIJUZv+OTeQP2E5/Ha/mR0HTzO6CGhTt4Pv1zj5Qp0FAIQjluIUZzDIU20hQu3jm7w7ZOdImXtfzPLeEyQEZmpIO/kFJZq4K/sfDVntgIbRxtXLCNFu33YmgoXzDhqMDYRlR8MZJyhIBv4YRLw1/fS5aN/AOp0Nfn2c6kZ6LNkr2TZlomd0DIiCID0L1kFNNjn8wZEQTRqtNC1qkwu/BJd1jXHJyPb6h8S2oy+v2OejMsLwV0Ze1lHN/j2hxPXLe2qrYOM2UA3s2UEO8U4NIGdRq1tiTE39Dh+BtCof4XpZmJSQSKSg4FMWfr7V2DdEOmyZsOAAUsBb9NrHv10KgXPf2WcwC6yaiVsfK6jwcPb1C/ZGMU5g0G6JelaVdopzmHutkroFa3CzjOpVmX0TUOQ5Z1sIAK4lV2A4MpeuJNdWOrZQNZMfXZ4ArvLBwy7k6RqkHFNu19UF/uf3wUxqSARycFAxtHys4AfJgJntkiXj9kGRD1g8u2iKOL9X87jk12XjMoGtAjHoqEt4O1pOB7F3C9Zua0lIbiLQ+l5OPTPLasz+iZqGsk6h60eblkD/9ufXOrZQJamSEtxRAI7tUbEpUt/o4GcnbOM8wCVV0wqSERyMJBxlIu/Al+ZaH1p/hgwYAng5SddDiCnoAjPrDmKfX/fNCp7vV9jjH+gjsn3mvslK7e1RLffwUu3ZGX0Fe+FFXMLRzkkj0xxPaNVaBcVbP1soBKKT5GWy94J7HTdf7Uzb2KDjNYu+MvL+1MeMKkgEcnBQMae8rOALROAs1uly8f8aLFb4OrtHPT/cC8yJGb4rH0qBl3qV7dYDXO/UK1vVTFsqzCV0TfXNxRvFozEjvzWFutnq+LdRh4KwbrZQCbopkjP2XrG7DRrRySwK979dwPmPxdA0M5eqh1rt/O7OiYVJCI5GMjYw4VfgPVDpctaPAE8tNhs6wsA7P/7JkasOGy0PdDXEz++0AW1qpoeO1OSuV+octdJEqFAmNIXHetUw8clurV2aGKwM78tYhTnEIK7SEMQXhwxBt1z1fh2vfzWDWtIdRvZa/Vq3RTpj3+/iA9+vSjr3KVVsvvP3Oeir0GfBRVmoC/ApIJEJA8DGVtpNMC3T5oe+zL2JyCyk8XDrNj7D96WWDCxc71q+GxUG1T2sf4jsvRL1tI6Sb/cS2g3e0A0OtStijClr1FXlQYKgynWo3PV6Nc8DM/+G4XP9iRZrKNuhpTc19Z2G1nLQyHgxZ4N0FAVUOouKzmkuv9MfS7aPDILKtzUa6AUSQWJqMIQRFGUO87RLWVkZECpVCI9PR2BgYH2O/ChZdrpsMW1HAn0XwR4me+zLyjS4JVvT0pOJZ7QtS5e7d3Q7AKOcpiatVScBzSY1zId5y7+jYu5lfXrJJVMNvbTn9fx/PrjZs8X5OeFT0a0RrvIYHy66yK+2JeE7Hy1vlwV6IPHY2ohslplhAT4ok3tKjh6+Y6+W8jS67LM5FoWWWR/OHENL244IVmmgEbf2jWsW1t07jGwQrXESGFmX6KKR+7zmy0ytgprcf+/n/xZ1tiF/zLzMfyzg/jnZrZR2cdPtMJDzcPtVj1Tv2R1dMEKAOz8R4VUzf2cKiVj2yoy0v3fzS3EiBWHjVpSgvy88GSnKEzqXs/owVOyW8jS67Jiry4rcyx1/+laux6v06HCBzFA2XwmROSeGMjYqnYsMCdd1q5//nsXAz/eL1n20+QuiA63Y0tRMcXT4xfP7KtS+iEmKlifH6Zkq82NjHyDrKnWTG/VlDhYem4hlvx6AQ1V/uwGKIYDWYmI7IOBjAN9f+xfTN140mh7I1UA1o/vgODKcubblo6pX7LWZE0tzfRWZmCVxoGsZAm704jkYSBjZxqNiHk/nsGqA8lGZcPa1sQ7DzeDp4dj86zIYU3WVFsSx5k6FrsH7uNAVjKFC2USycdAxk4y8grx5MojOHr5jlHZW4OaYFTHyLKvlBnWZE0113rgiHNKKa+/Tq1eHZvKPS6USWQdBjKldOm/LPRdshcFao1R2cZnO7rsGAdrs6ZaGjxsz3OWVN5/nXIgK+lwoUwi6zGQsdG1u7notOB3o+2hgT7Y/HwnhAeZT4DnbLYMNtW1Hhy6dAsT1x/D3dxCWecqzcBV/jqlioQLZRJZz/mDNdxUySCmZ+NQnHurDw6/1tPlgxjg/mBT4P7gUh1zg009FAI61a+GBUOaQZB4b0mlGbhq6dcpoP11qi45VYrITXGhTCLrMZCxgVoj4s2HtEHA0DY18ff/9cWKMW3h62XffB9qjYiDl27hhxPXcPDSLdkPbLnv03UXqZSGXT4qpa/Flg5T7y0Zq8g5limHii1YKaX4r1NnsvVzIiqJC2USWY9dS1YqOV5j09F/se/vm3Yfr2HruBBr31eawaZS77VXRt740ymY8d0pWfs689dpeR+/Q2WL+YWIrMclCqxgaryG7jFtr/Eatp6nrOrnaHKWVyju6/EdnDJeoLzcb3Ituu8VIJ1fiN8rqijkPr/ZtSRTWY3XsPU85WU8ibnrKEmAtvXDGb9Oy8v9JtdTmi5fooqIXUsyldVsAlvPU15mO1i6jpKclf22vNxvck3ML0QkHwMZmcpqNoGt5ykvsx3k1i/IzwsLhjRz2q/T8nK/yXUxvxCRPAxkZCqr2QS2nqe8zHaQW79PRrRGp3rVHFwb08rL/SYicnccIyOTbjaBqYZde43XsPU8ZVU/R5N7HR3qOPeXanm530RE7o6BjEy2JpArq/OUVf0czV2uw13qSURU3jGQsUJZzSaw9TzlZbaDu1yHu9STiKg8Yx4ZG5TVSsy2nqe8rBTtLtfhLvUkInIncp/fDGSIiIjI5TAhHhEREZV7DGSIiIjIbTGQISIiIrfFQIaIiIjcFgMZIiIiclsMZIiIiMhtMZAhIiIit8VAhoiIiNwWAxkiIiJyW57OroCj6RIXZ2RkOLkmREREJJfuuW1pAYJyH8hkZmYCACIiIpxcEyIiIrJWZmYmlEqlyfJyv9aSRqPB9evXERAQAEHgQn72lpGRgYiICFy9epVrWZUh3nfn4b13Dt5353HWvRdFEZmZmQgPD4dCYXokTLlvkVEoFKhZs6azq1HuBQYG8h8XJ+B9dx7ee+fgfXceZ9x7cy0xOhzsS0RERG6LgQwRERG5LQYyVCo+Pj6YPXs2fHx8nF2VCoX33Xl4752D9915XP3el/vBvkRERFR+sUWGiIiI3BYDGSIiInJbDGSIiIjIbTGQISIiIrfFQIZk2bNnDwYMGIDw8HAIgoAtW7YYlIuiiDfffBNhYWHw8/NDz549cfHiRedUthyZP38+2rVrh4CAAISEhGDw4ME4f/68wT55eXmYOHEiqlatCn9/fwwZMgQ3btxwUo3Lh2XLlqF58+b6BGAdO3bEzz//rC/nPS8bCxYsgCAImDJlin4b771jzJkzB4IgGPw1atRIX+7K952BDMmSnZ2NFi1a4JNPPpEsf/fdd/Hhhx9i+fLlOHz4MCpXrozevXsjLy+vjGtavuzevRsTJ07EoUOHsHPnThQWFuLBBx9Edna2fp+XXnoJ27Ztw6ZNm7B7925cv34djzzyiBNr7f5q1qyJBQsW4OjRo/jjjz/QvXt3DBo0CH/99RcA3vOycOTIEXz22Wdo3ry5wXbee8dp0qQJUlJS9H/79u3Tl7n0fReJrARA3Lx5s/61RqMRVSqV+N577+m33b17V/Tx8RG//vprJ9Sw/EpLSxMBiLt37xZFUXufvby8xE2bNun3OXv2rAhAPHjwoLOqWS5VqVJFXLFiBe95GcjMzBTr168v7ty5U4yLixNffPFFURT5fXek2bNniy1atJAsc/X7zhYZKrWkpCSkpqaiZ8+e+m1KpRLt27fHwYMHnViz8ic9PR0AEBwcDAA4evQoCgsLDe59o0aNUKtWLd57O1Gr1diwYQOys7PRsWNH3vMyMHHiRPTv39/gHgP8vjvaxYsXER4ejjp16mDEiBG4cuUKANe/7+V+0UhyvNTUVABAaGiowfbQ0FB9GZWeRqPBlClT0KlTJzRt2hSA9t57e3sjKCjIYF/e+9I7deoUOnbsiLy8PPj7+2Pz5s2Ijo7GiRMneM8daMOGDTh27BiOHDliVMbvu+O0b98eq1atQsOGDZGSkoK5c+eiS5cuOH36tMvfdwYyRG5i4sSJOH36tEG/NTlOw4YNceLECaSnp+Pbb7/FmDFjsHv3bmdXq1y7evUqXnzxRezcuRO+vr7Ork6F0rdvX/1/N2/eHO3bt0ft2rWxceNG+Pn5ObFmlrFriUpNpVIBgNEI9hs3bujLqHQmTZqEH3/8Ebt27ULNmjX121UqFQoKCnD37l2D/XnvS8/b2xv16tVDmzZtMH/+fLRo0QJLly7lPXego0ePIi0tDa1bt4anpyc8PT2xe/dufPjhh/D09ERoaCjvfRkJCgpCgwYN8Pfff7v8d56BDJVaVFQUVCoVfvvtN/22jIwMHD58GB07dnRizdyfKIqYNGkSNm/ejN9//x1RUVEG5W3atIGXl5fBvT9//jyuXLnCe29nGo0G+fn5vOcO1KNHD5w6dQonTpzQ/7Vt2xYjRozQ/zfvfdnIysrCpUuXEBYW5vLfeXYtkSxZWVn4+++/9a+TkpJw4sQJBAcHo1atWpgyZQrefvtt1K9fH1FRUZg1axbCw8MxePBg51W6HJg4cSLWr1+PH374AQEBAfr+aKVSCT8/PyiVSjz11FOYOnUqgoODERgYiBdeeAEdO3ZEhw4dnFx79zVz5kz07dsXtWrVQmZmJtavX4+EhATs2LGD99yBAgIC9OO/dCpXroyqVavqt/PeO8a0adMwYMAA1K5dG9evX8fs2bPh4eGBxx9/3PW/886eNkXuYdeuXSIAo78xY8aIoqidgj1r1iwxNDRU9PHxEXv06CGeP3/euZUuB6TuOQBx5cqV+n1yc3PF559/XqxSpYpYqVIl8eGHHxZTUlKcV+lyYNy4cWLt2rVFb29vsXr16mKPHj3EX375RV/Oe152ik+/FkXee0cZPny4GBYWJnp7e4s1atQQhw8fLv7999/6cle+74IoiqKTYigiIiKiUuEYGSIiInJbDGSIiIjIbTGQISIiIrfFQIaIiIjcFgMZIiIiclsMZIiIiMhtMZAhIiIit8VAhojcniAI2LJli8nyrl27YsqUKWaPERkZiSVLlkgeMzk5GYIg4MSJE6WuKxHZFwMZIrIbQRDM/s2ZM8fke50dLBw5cgTPPPOMU85NRLbjWktEZDcpKSn6//7mm2/w5ptv4vz58/pt/v7+zqiWLNWrV3d2FYjIBmyRISK7UalU+j+lUglBEPSvQ0JCsHjxYtSsWRM+Pj5o2bIl4uPj9e/VrezdqlUrCIKArl27AtC2lPTq1QvVqlWDUqlEXFwcjh07ZnXdioqKMGnSJCiVSlSrVg2zZs1C8RVaSnYtEZF7YCBDRGVi6dKlWLRoEd5//338+eef6N27NwYOHIiLFy8CABITEwEAv/76K1JSUvD9998DADIzMzFmzBjs27cPhw4dQv369dGvXz9kZmZadf7Vq1fD09MTiYmJWLp0KRYvXowVK1bY9yKJqMyxa4mIysT777+P6dOn47HHHgMALFy4ELt27cKSJUvwySef6Lt2qlatCpVKpX9f9+7dDY7z+eefIygoCLt378ZDDz0k+/wRERH44IMPIAgCGjZsiFOnTuGDDz7A+PHj7XB1ROQsbJEhIofLyMjA9evX0alTJ4PtnTp1wtmzZ82+98aNGxg/fjzq168PpVKJwMBAZGVl4cqVK1bVoUOHDhAEQf+6Y8eOuHjxItRqtVXHISLXwhYZInJpY8aMwa1bt7B06VLUrl0bPj4+6NixIwoKCpxdNSJyAWyRISKHCwwMRHh4OPbv32+wff/+/YiOjgYAeHt7A4BRC8n+/fsxefJk9OvXD02aNIGPjw9u3rxpdR0OHz5s8Fo33sbDw8PqYxGR62CLDBGViVdeeQWzZ89G3bp10bJlS6xcuRInTpzAV199BQAICQmBn58f4uPjUbNmTfj6+kKpVKJ+/fpYu3Yt2rZti4yMDLzyyivw8/Oz+vxXrlzB1KlT8eyzz+LYsWP46KOPsGjRIntfJhGVMbbIEFGZmDx5MqZOnYqXX34ZzZo1Q3x8PLZu3Yr69esDADw9PfHhhx/is88+Q3h4OAYNGgQA+PLLL3Hnzh20bt0ao0aNwuTJkxESEmL1+UePHo3c3FzExMRg4sSJePHFF5kAj6gcEMTiiRSIiIiI3AhbZIiIiMhtMZAhIiIit8VAhoiIiNwWAxkiIiJyWwxkiIiIyG0xkCEiIiK3xUCGiIiI3BYDGSIiInJbDGSIiIjIbTGQISIiIrfFQIaIiIjcFgMZIiIiclv/DyT8qeAHHHsqAAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Let's see how the two lines differ\n",
    " \n",
    "# It's natural to plot scatter plot between predictor and target. \n",
    "plt.scatter(X_train_tips.loc[:, ['total_bill']], y_train_tips, label = \"Train\")\n",
    "plt.scatter(X_test_tips.loc[:, ['total_bill']], y_test_tips, label = \"Test\")\n",
    "plt.plot(X_train_tips.loc[:, ['total_bill']], reg0.predict(X_train_tips.loc[:, ['total_bill']]), label = \"Without penality\")\n",
    "plt.plot(X_train_tips.loc[:, ['total_bill']], ridge0.predict(X_train_tips.loc[:, ['total_bill']]), label = \"With penality\")\n",
    "plt.title(\"Tips vs Total bill\")\n",
    "plt.xlabel(\"Total bill\")\n",
    "plt.ylabel(\"Tips\")\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.2.2 Lasso\n",
    "\n",
    "If you are not too lazy, do the same thing for `Lasso` regression ($\\ell_1$ penalty)\n",
    "\n",
    "$$(a^*,b^*,\\alpha^*)=\\arg\\min_{(a,b,\\alpha)}\\sum_{i=1}^n(y_i-(ax_i+b))^2+\\alpha(|a|+|b|).$$\n",
    "\n",
    "**Don't hesitate to add more cells as you are working on it!**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# To do\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1.3 Multiple Linear Regression\n",
    "\n",
    "You can do pretty much the same as what we have done by including `size` variable in the training and testing data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8210834735838907"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Direct MLR\n",
    "reg1 = LinearRegression().fit(X_train_tips.loc[:,['total_bill', 'size']], y_train_tips)\n",
    "\n",
    "pred_mlr = reg1.predict(X_test_tips.loc[:,['total_bill', 'size']])\n",
    "\n",
    "np.sqrt(mean_squared_error(y_test_tips, pred_mlr))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Check what regularization parameters are called in Elastic net?\n",
    "# help(ElasticNet)\n",
    "\n",
    "# Let do elastic net in this case\n",
    "param = {\n",
    "    'alpha' : 10 ** np.linspace(-5,3, num=50),\n",
    "    'l1_ratio' : np.linspace(0,1, num=20)\n",
    "}\n",
    "\n",
    "elas_cv = GridSearchCV(\n",
    "    estimator=ElasticNet(),\n",
    "    param_grid=param,\n",
    "    cv=10\n",
    ")\n",
    "\n",
    "elas_cv.fit(X_train_tips.loc[:,['total_bill', 'size']], y_train_tips)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-block alert-danger\">\n",
    "<b>Q0. A.</b> What are the warnings about? When <code>l1_ratio</code> is null, what does it mean? What type of regularized model should you consider? Let's build the final model.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Your answer:*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best parameters: {'alpha': 10.985411419875572, 'l1_ratio': 0.0}\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Sothea Has\\AppData\\Roaming\\Python\\Python39\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:628: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations, check the scale of the features or consider increasing regularisation. Duality gap: 1.229e+02, tolerance: 4.212e-02 Linear regression models with null weight for the l1 regularization term are more efficiently fitted using one of the solvers implemented in sklearn.linear_model.Ridge/RidgeCV instead.\n",
      "  model = cd_fast.enet_coordinate_descent(\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.7640642937177642"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(\"Best parameters:\", elas_cv.best_params_)\n",
    "\n",
    "mlr_elas = ElasticNet(alpha=elas_cv.best_params_['alpha'],\n",
    "                      l1_ratio=elas_cv.best_params_['l1_ratio']).fit(X_train_tips.loc[:,['total_bill', 'size']], y_train_tips)\n",
    "\n",
    "pred_mlr_elas = mlr_elas.predict(X_test_tips.loc[:,['total_bill', 'size']])\n",
    "\n",
    "np.sqrt(mean_squared_error(y_test_tips, pred_mlr_elas))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-block alert-danger\">\n",
    "<b>Q0. B.</b> Is this better than simple linear regression?\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Your answer:*"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Logistic Regression\n",
    "### 2.1 Perceptron (Binary Logistic Regression) with `Spam` dataset\n",
    "$$\\mathbb{P}(Y=1|X=x)=\\sigma(x^t\\beta),\\text{ where }\\sigma(t)=\\frac{1}{1+e^{-t}}$$\n",
    "\n",
    "You have seen how it all works in linear regression case. Everything is similar, just replace `LinearRegression` or `Ridge` in the previous case by `LogisticRegression`. Remember to follow this steps: `load model`, `fit`, `predict`, and then `evaluation` using suitable metrics."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Feel free to look at the distributions of predictors and target of the dataset\n",
    "\n",
    "# TO DO"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Just build a logistic regression model on the training data using."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Sothea Has\\AppData\\Roaming\\Python\\Python39\\site-packages\\sklearn\\linear_model\\_logistic.py:460: ConvergenceWarning: lbfgs failed to converge (status=1):\n",
      "STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.\n",
      "\n",
      "Increase the number of iterations (max_iter) or scale the data as shown in:\n",
      "    https://scikit-learn.org/stable/modules/preprocessing.html\n",
      "Please also refer to the documentation for alternative solver options:\n",
      "    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression\n",
      "  n_iter_i = _check_optimize_result(\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[-2.58653847e-01, -5.25702603e-02,  1.66977608e-01,\n",
       "         2.82685135e-01,  6.23944607e-01,  7.41330076e-01,\n",
       "         1.65942603e+00,  1.07871059e+00,  4.92254730e-01,\n",
       "        -1.78877800e-01,  3.54135491e-02, -1.18973091e-01,\n",
       "         1.90574262e-01,  1.95521845e-01,  3.86512970e-01,\n",
       "         1.24815974e+00,  8.83756481e-01,  4.91825445e-01,\n",
       "         8.00678704e-02,  6.58421425e-01,  1.94212922e-01,\n",
       "         2.28210012e-01,  1.45357947e+00,  6.28833484e-01,\n",
       "        -2.28892148e+00, -1.17780804e+00, -3.33314749e+00,\n",
       "         3.20247292e-01, -6.48856511e-01, -3.84272665e-01,\n",
       "        -3.39579478e-01, -1.97513943e-01, -8.41296667e-01,\n",
       "        -2.05173285e-01, -5.55330389e-01,  8.86075536e-02,\n",
       "        -3.32435205e-01, -3.42832885e-01, -5.62647775e-01,\n",
       "        -2.43218930e-01, -5.09819740e-01, -1.20249379e+00,\n",
       "        -2.97950447e-01, -8.18513390e-01, -7.82352585e-01,\n",
       "        -1.44079523e+00, -1.61224355e-01, -5.03618399e-01,\n",
       "        -7.24652498e-01, -1.78907413e-01, -1.21502270e-01,\n",
       "         3.17081705e-01,  1.14023533e+00,  4.16007635e-01,\n",
       "        -8.71996644e-03,  7.09520114e-03,  8.34731411e-04]])"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "logit0 = LogisticRegression().fit(X_train_spam, y_train_spam)\n",
    "logit0.coef_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-block alert-danger\">\n",
    "<b>Q1.</b> What is the warning about? What convergence does it refer to?\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Your answer:*"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-block alert-danger\">\n",
    "<b>Q2.</b> Let's fix the warning by increasing the <code>max_iter</code> option in <code>LogisticRegression</code> object, then fit it on the training data. Finally, evaluate how well does the model perform on the testing data?\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# TO DO"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# TO DO"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.2 Regularized Logistic Regression\n",
    "\n",
    "<div class=\"alert alert-block alert-danger\">\n",
    "<b>Q3. A.</b> You should read <code>help(LogisticRegression)</code> or look it up. The goal is to realize that the penalty parameter are already built in <code>LogisticRegression</code> object.<br> \n",
    "    <li> Now, what are the arguments to control the regularization parameter in <code>LogisticRegression</code>? </li>\n",
    "    <li> Find the best ones using <code>GridSearchCV</code> as we did in the previous section. </li>\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# TO DO"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# TO DO"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-block alert-danger\"> \n",
    "<b>Q3. B.</b> Make prediction on the test data and compute its performance.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# TO DO"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.3 Multinomial Logistic Regression or Multi-layer Perceptron (MLP)\n",
    "\n",
    "If you reach this question, you have done well. I'm sure you know what to do with `Mnist` dataset. I'll process the data for you a bit, then you will build your own model using what you done up until now. Let's first look at some training examples."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x1000 with 18 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "f, axs = plt.subplots(3,3,figsize=(10,10))\n",
    "ind = np.random.choice(range(len(y_train_mnist)), 9)\n",
    "for i in range(9):\n",
    "    k = ind[i]\n",
    "    j = i%3\n",
    "    jj = int(np.floor(i/3))\n",
    "    im = axs[jj,j].imshow(X_train_mnist[k, :, :])\n",
    "    plt.colorbar(im, ax=axs[jj, j])\n",
    "    axs[jj,j].set_title(\"Number : \"+str(y_train_mnist[k]))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Convert labels 0-9 to one-hot vector. Let create new target called 'y_train_mnist1' and 'y_test_mnist1' resp.\n",
    "y_train_mnist1 = to_categorical(y_train_mnist) \n",
    "y_test_mnist1 = to_categorical(y_test_mnist)\n",
    "num_labels = len(np.unique(y_train_mnist))\n",
    "\n",
    "# Dimension of images\n",
    "image_size_mnist = X_train_mnist.shape[1]\n",
    "input_size_mnist = image_size_mnist** 2\n",
    "\n",
    "# resize and normalize\n",
    "X_train_mnist1 = np.reshape(X_train_mnist, [-1, input_size_mnist])\n",
    "X_train_mnist1 = X_train_mnist1.astype('float32') / 255\n",
    "X_test_mnist1 = np.reshape(X_test_mnist, [-1, input_size_mnist])\n",
    "X_test_mnist1 = X_test_mnist1.astype('float32') / 255"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-block alert-danger\"> \n",
    "<b>Q4.</b> Now, you can build MLP using <code>X_train_mnist1</code> and <code>y_train_mnist1</code> and <code>LogisticRegression</code> object. Don't forget to test it using the testing data <code>X_test_mnist1</code>.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# TO DO"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# TO DO"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. Deep Neural Networks\n",
    "\n",
    "<div class=\"alert alert-block alert-success\">\n",
    "It is your entire choice to design your networks now. Try to compare DNN result with the previous methods on <code>Spam</code> and <code>Mnist</code> dataset.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "To train neural networks, we need the following methods:\n",
    "    <li> `Sequential`: `keras` model allowing us to build a sequent of layers in the networks.</li>\n",
    "    <li> `Dense`, `Dropout`: `keras` layer module allowing us to design the size and property of different types of layers in the NN.</li>\n",
    "    <li> `to_categorical` : to perform `one-hot` encoding on categorical target (must be in `int` type).</li>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense , Dropout\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "from tensorflow.keras.optimizers import Adam\n",
    "from tensorflow.keras.regularizers import l1, l2\n",
    "\n",
    "# This function allows you to keep track of the learning curves\n",
    "def plot_history(fitted_model, key, save = False, showfig = False, img_path = None):\n",
    "    plt.plot(fitted_model.history[key])\n",
    "    plt.plot(fitted_model.history['val_'+key])\n",
    "    plt.xlabel(\"Epochs\")\n",
    "    plt.ylabel(key)\n",
    "    plt.legend([key, 'val_'+key])\n",
    "    if save:\n",
    "        image_name = 'training_curve.png'\n",
    "        if img_path is None:\n",
    "            plt.savefig(image_name, format='png', dpi=600, bbox_inches='tight')\n",
    "        else:\n",
    "            plt.savefig(img_path + image_name, dpi=600, bbox_inches='tight')\n",
    "        plt.clf()\n",
    "    if showfig:\n",
    "        plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.1 `Spam` data with DNN\n",
    "\n",
    "We first encode the target from categorical to binary vector: `spam` = (0,1) and `nonspam` = (1,0). This can be done using `to_categorical` function."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "# One hot encoding\n",
    "temp = np.repeat(1, len(y_train_spam))\n",
    "temp[y_train_spam == \"nonspam\"] = 0\n",
    "y_train_spam1 = to_categorical(temp)\n",
    "\n",
    "temp = np.repeat(1, len(y_test_spam))\n",
    "temp[y_test_spam == \"nonspam\"] = 0\n",
    "y_test_spam1 = to_categorical(temp)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-block alert-success\">\n",
    "The most complicated part of building DNN is to choose its architecture and its hyperparameters. The followings are chosen based on a few tries. In practice, you can start with small networks, then progressively increase the size of the networks according to the training process.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# network parameters\n",
    "input_size_spam = X_train_spam.shape[1]\n",
    "n_epochs_spam = 200                       # The number of network-complete-update iterations in the training process.\n",
    "batch_size_spam = 200\n",
    "hidden_units_spam = 16\n",
    "learn_rate = 0.001\n",
    "val_size_spam = 0.05                      # The size of validation data to keep track of model performance at each epoch\n",
    "\n",
    "# Building the sequent of the NN\n",
    "model = Sequential()\n",
    "model.add(Dense(hidden_units_spam, input_dim=input_size_spam, activation='relu', bias_regularizer = l2(100)))\n",
    "model.add(Dense(2, activation='sigmoid'))\n",
    "\n",
    "# Set up compiling method of the model \n",
    "model.compile(\n",
    "  loss = 'categorical_crossentropy',\n",
    "  optimizer = Adam(learning_rate = learn_rate),\n",
    "  metrics =['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/200\n",
      "18/18 [==============================] - 1s 15ms/step - loss: 3.4207 - accuracy: 0.5169 - val_loss: 2.3338 - val_accuracy: 0.4402\n",
      "Epoch 2/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 1.2624 - accuracy: 0.5807 - val_loss: 1.1797 - val_accuracy: 0.6522\n",
      "Epoch 3/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 0.8069 - accuracy: 0.6453 - val_loss: 0.9004 - val_accuracy: 0.6848\n",
      "Epoch 4/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 0.6861 - accuracy: 0.7140 - val_loss: 0.6193 - val_accuracy: 0.7500\n",
      "Epoch 5/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 0.5299 - accuracy: 0.7772 - val_loss: 0.6022 - val_accuracy: 0.7880\n",
      "Epoch 6/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 0.4928 - accuracy: 0.8055 - val_loss: 0.5436 - val_accuracy: 0.7772\n",
      "Epoch 7/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 0.5415 - accuracy: 0.7812 - val_loss: 0.5963 - val_accuracy: 0.7826\n",
      "Epoch 8/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.4293 - accuracy: 0.8370 - val_loss: 0.4687 - val_accuracy: 0.8587\n",
      "Epoch 9/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.3931 - accuracy: 0.8636 - val_loss: 0.4557 - val_accuracy: 0.8207\n",
      "Epoch 10/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.4156 - accuracy: 0.8670 - val_loss: 0.4559 - val_accuracy: 0.8261\n",
      "Epoch 11/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.3599 - accuracy: 0.8847 - val_loss: 0.3754 - val_accuracy: 0.8804\n",
      "Epoch 12/200\n",
      "18/18 [==============================] - 0s 7ms/step - loss: 0.3375 - accuracy: 0.8939 - val_loss: 0.3654 - val_accuracy: 0.8696\n",
      "Epoch 13/200\n",
      "18/18 [==============================] - 0s 7ms/step - loss: 0.3337 - accuracy: 0.8856 - val_loss: 0.3374 - val_accuracy: 0.9022\n",
      "Epoch 14/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.3136 - accuracy: 0.8959 - val_loss: 0.3329 - val_accuracy: 0.8750\n",
      "Epoch 15/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.3167 - accuracy: 0.8864 - val_loss: 0.3042 - val_accuracy: 0.9239\n",
      "Epoch 16/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.3080 - accuracy: 0.8916 - val_loss: 0.3325 - val_accuracy: 0.9076\n",
      "Epoch 17/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.3156 - accuracy: 0.8887 - val_loss: 0.3170 - val_accuracy: 0.9076\n",
      "Epoch 18/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.3208 - accuracy: 0.8859 - val_loss: 0.2731 - val_accuracy: 0.9402\n",
      "Epoch 19/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.3032 - accuracy: 0.8982 - val_loss: 0.2812 - val_accuracy: 0.9130\n",
      "Epoch 20/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2812 - accuracy: 0.9082 - val_loss: 0.2714 - val_accuracy: 0.9130\n",
      "Epoch 21/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2735 - accuracy: 0.9119 - val_loss: 0.2625 - val_accuracy: 0.9239\n",
      "Epoch 22/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.3091 - accuracy: 0.8976 - val_loss: 0.3541 - val_accuracy: 0.8696\n",
      "Epoch 23/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.3160 - accuracy: 0.8916 - val_loss: 0.3577 - val_accuracy: 0.8533\n",
      "Epoch 24/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.3377 - accuracy: 0.8847 - val_loss: 0.3404 - val_accuracy: 0.8913\n",
      "Epoch 25/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.3303 - accuracy: 0.8876 - val_loss: 0.3163 - val_accuracy: 0.8804\n",
      "Epoch 26/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2901 - accuracy: 0.9088 - val_loss: 0.2469 - val_accuracy: 0.9239\n",
      "Epoch 27/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2712 - accuracy: 0.9090 - val_loss: 0.2603 - val_accuracy: 0.9076\n",
      "Epoch 28/200\n",
      "18/18 [==============================] - 0s 10ms/step - loss: 0.2597 - accuracy: 0.9173 - val_loss: 0.2297 - val_accuracy: 0.9402\n",
      "Epoch 29/200\n",
      "18/18 [==============================] - 0s 9ms/step - loss: 0.2537 - accuracy: 0.9182 - val_loss: 0.2373 - val_accuracy: 0.9348\n",
      "Epoch 30/200\n",
      "18/18 [==============================] - 0s 7ms/step - loss: 0.2603 - accuracy: 0.9108 - val_loss: 0.2326 - val_accuracy: 0.9402\n",
      "Epoch 31/200\n",
      "18/18 [==============================] - 0s 7ms/step - loss: 0.2585 - accuracy: 0.9159 - val_loss: 0.2237 - val_accuracy: 0.9402\n",
      "Epoch 32/200\n",
      "18/18 [==============================] - 0s 9ms/step - loss: 0.2557 - accuracy: 0.9193 - val_loss: 0.2255 - val_accuracy: 0.9348\n",
      "Epoch 33/200\n",
      "18/18 [==============================] - 0s 9ms/step - loss: 0.2631 - accuracy: 0.9182 - val_loss: 0.2203 - val_accuracy: 0.9457\n",
      "Epoch 34/200\n",
      "18/18 [==============================] - 0s 8ms/step - loss: 0.2508 - accuracy: 0.9199 - val_loss: 0.2826 - val_accuracy: 0.9022\n",
      "Epoch 35/200\n",
      "18/18 [==============================] - 0s 9ms/step - loss: 0.2426 - accuracy: 0.9222 - val_loss: 0.2165 - val_accuracy: 0.9402\n",
      "Epoch 36/200\n",
      "18/18 [==============================] - 0s 9ms/step - loss: 0.2380 - accuracy: 0.9285 - val_loss: 0.2166 - val_accuracy: 0.9348\n",
      "Epoch 37/200\n",
      "18/18 [==============================] - 0s 8ms/step - loss: 0.2443 - accuracy: 0.9228 - val_loss: 0.2137 - val_accuracy: 0.9457\n",
      "Epoch 38/200\n",
      "18/18 [==============================] - 0s 7ms/step - loss: 0.2392 - accuracy: 0.9248 - val_loss: 0.2142 - val_accuracy: 0.9348\n",
      "Epoch 39/200\n",
      "18/18 [==============================] - 0s 7ms/step - loss: 0.2349 - accuracy: 0.9256 - val_loss: 0.2370 - val_accuracy: 0.9293\n",
      "Epoch 40/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2353 - accuracy: 0.9288 - val_loss: 0.2168 - val_accuracy: 0.9348\n",
      "Epoch 41/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2320 - accuracy: 0.9276 - val_loss: 0.2085 - val_accuracy: 0.9457\n",
      "Epoch 42/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2337 - accuracy: 0.9268 - val_loss: 0.2078 - val_accuracy: 0.9402\n",
      "Epoch 43/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2525 - accuracy: 0.9153 - val_loss: 0.2204 - val_accuracy: 0.9293\n",
      "Epoch 44/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2375 - accuracy: 0.9231 - val_loss: 0.2213 - val_accuracy: 0.9239\n",
      "Epoch 45/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2413 - accuracy: 0.9182 - val_loss: 0.2607 - val_accuracy: 0.9130\n",
      "Epoch 46/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2290 - accuracy: 0.9251 - val_loss: 0.2044 - val_accuracy: 0.9457\n",
      "Epoch 47/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2259 - accuracy: 0.9268 - val_loss: 0.2202 - val_accuracy: 0.9239\n",
      "Epoch 48/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2344 - accuracy: 0.9213 - val_loss: 0.2113 - val_accuracy: 0.9402\n",
      "Epoch 49/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2496 - accuracy: 0.9179 - val_loss: 0.2054 - val_accuracy: 0.9457\n",
      "Epoch 50/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2768 - accuracy: 0.9073 - val_loss: 0.2450 - val_accuracy: 0.9076\n",
      "Epoch 51/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2271 - accuracy: 0.9236 - val_loss: 0.2203 - val_accuracy: 0.9293\n",
      "Epoch 52/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2344 - accuracy: 0.9216 - val_loss: 0.2163 - val_accuracy: 0.9402\n",
      "Epoch 53/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2227 - accuracy: 0.9288 - val_loss: 0.1970 - val_accuracy: 0.9511\n",
      "Epoch 54/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2189 - accuracy: 0.9311 - val_loss: 0.2452 - val_accuracy: 0.9239\n",
      "Epoch 55/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2271 - accuracy: 0.9245 - val_loss: 0.2002 - val_accuracy: 0.9402\n",
      "Epoch 56/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2455 - accuracy: 0.9205 - val_loss: 0.2540 - val_accuracy: 0.9185\n",
      "Epoch 57/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2451 - accuracy: 0.9159 - val_loss: 0.2451 - val_accuracy: 0.9348\n",
      "Epoch 58/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2439 - accuracy: 0.9191 - val_loss: 0.4044 - val_accuracy: 0.8641\n",
      "Epoch 59/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2758 - accuracy: 0.9056 - val_loss: 0.2419 - val_accuracy: 0.9293\n",
      "Epoch 60/200\n",
      "18/18 [==============================] - 0s 7ms/step - loss: 0.2325 - accuracy: 0.9179 - val_loss: 0.1993 - val_accuracy: 0.9402\n",
      "Epoch 61/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2197 - accuracy: 0.9279 - val_loss: 0.2027 - val_accuracy: 0.9402\n",
      "Epoch 62/200\n",
      "18/18 [==============================] - 0s 7ms/step - loss: 0.2198 - accuracy: 0.9268 - val_loss: 0.1907 - val_accuracy: 0.9457\n",
      "Epoch 63/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2128 - accuracy: 0.9293 - val_loss: 0.1912 - val_accuracy: 0.9565\n",
      "Epoch 64/200\n",
      "18/18 [==============================] - 0s 7ms/step - loss: 0.2169 - accuracy: 0.9276 - val_loss: 0.1915 - val_accuracy: 0.9402\n",
      "Epoch 65/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2357 - accuracy: 0.9191 - val_loss: 0.2184 - val_accuracy: 0.9511\n",
      "Epoch 66/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2598 - accuracy: 0.9211 - val_loss: 0.5619 - val_accuracy: 0.8804\n",
      "Epoch 67/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2697 - accuracy: 0.9116 - val_loss: 0.2242 - val_accuracy: 0.9402\n",
      "Epoch 68/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.3114 - accuracy: 0.9025 - val_loss: 0.1932 - val_accuracy: 0.9402\n",
      "Epoch 69/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2444 - accuracy: 0.9219 - val_loss: 0.1973 - val_accuracy: 0.9402\n",
      "Epoch 70/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2132 - accuracy: 0.9262 - val_loss: 0.1899 - val_accuracy: 0.9457\n",
      "Epoch 71/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2103 - accuracy: 0.9265 - val_loss: 0.1930 - val_accuracy: 0.9348\n",
      "Epoch 72/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2116 - accuracy: 0.9291 - val_loss: 0.2052 - val_accuracy: 0.9293\n",
      "Epoch 73/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2192 - accuracy: 0.9242 - val_loss: 0.1944 - val_accuracy: 0.9402\n",
      "Epoch 74/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2127 - accuracy: 0.9262 - val_loss: 0.1884 - val_accuracy: 0.9457\n",
      "Epoch 75/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2057 - accuracy: 0.9296 - val_loss: 0.1862 - val_accuracy: 0.9457\n",
      "Epoch 76/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2065 - accuracy: 0.9308 - val_loss: 0.2496 - val_accuracy: 0.9293\n",
      "Epoch 77/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2148 - accuracy: 0.9273 - val_loss: 0.1862 - val_accuracy: 0.9457\n",
      "Epoch 78/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2016 - accuracy: 0.9316 - val_loss: 0.2213 - val_accuracy: 0.9348\n",
      "Epoch 79/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2280 - accuracy: 0.9222 - val_loss: 0.1947 - val_accuracy: 0.9348\n",
      "Epoch 80/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2063 - accuracy: 0.9319 - val_loss: 0.2091 - val_accuracy: 0.9348\n",
      "Epoch 81/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2113 - accuracy: 0.9248 - val_loss: 0.2682 - val_accuracy: 0.9239\n",
      "Epoch 82/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2645 - accuracy: 0.9139 - val_loss: 0.1932 - val_accuracy: 0.9457\n",
      "Epoch 83/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2051 - accuracy: 0.9302 - val_loss: 0.2236 - val_accuracy: 0.9293\n",
      "Epoch 84/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2048 - accuracy: 0.9305 - val_loss: 0.2117 - val_accuracy: 0.9348\n",
      "Epoch 85/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2055 - accuracy: 0.9336 - val_loss: 0.1860 - val_accuracy: 0.9348\n",
      "Epoch 86/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 0.1991 - accuracy: 0.9331 - val_loss: 0.2037 - val_accuracy: 0.9348\n",
      "Epoch 87/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2124 - accuracy: 0.9302 - val_loss: 0.2064 - val_accuracy: 0.9293\n",
      "Epoch 88/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2217 - accuracy: 0.9259 - val_loss: 0.2193 - val_accuracy: 0.9457\n",
      "Epoch 89/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.2229 - accuracy: 0.9288 - val_loss: 0.2136 - val_accuracy: 0.9348\n",
      "Epoch 90/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2111 - accuracy: 0.9291 - val_loss: 0.2243 - val_accuracy: 0.9348\n",
      "Epoch 91/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2041 - accuracy: 0.9328 - val_loss: 0.2877 - val_accuracy: 0.9076\n",
      "Epoch 92/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2214 - accuracy: 0.9233 - val_loss: 0.1820 - val_accuracy: 0.9348\n",
      "Epoch 93/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1971 - accuracy: 0.9331 - val_loss: 0.1827 - val_accuracy: 0.9457\n",
      "Epoch 94/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 0.1928 - accuracy: 0.9351 - val_loss: 0.1844 - val_accuracy: 0.9457\n",
      "Epoch 95/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1932 - accuracy: 0.9348 - val_loss: 0.1788 - val_accuracy: 0.9457\n",
      "Epoch 96/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 0.1916 - accuracy: 0.9356 - val_loss: 0.1785 - val_accuracy: 0.9511\n",
      "Epoch 97/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1944 - accuracy: 0.9339 - val_loss: 0.1809 - val_accuracy: 0.9348\n",
      "Epoch 98/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1921 - accuracy: 0.9374 - val_loss: 0.2006 - val_accuracy: 0.9348\n",
      "Epoch 99/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2359 - accuracy: 0.9231 - val_loss: 0.3275 - val_accuracy: 0.8750\n",
      "Epoch 100/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 0.2639 - accuracy: 0.9145 - val_loss: 0.1905 - val_accuracy: 0.9457\n",
      "Epoch 101/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.3122 - accuracy: 0.8987 - val_loss: 0.2255 - val_accuracy: 0.9293\n",
      "Epoch 102/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2568 - accuracy: 0.9142 - val_loss: 0.2375 - val_accuracy: 0.9185\n",
      "Epoch 103/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2090 - accuracy: 0.9299 - val_loss: 0.2144 - val_accuracy: 0.9293\n",
      "Epoch 104/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1961 - accuracy: 0.9354 - val_loss: 0.1822 - val_accuracy: 0.9402\n",
      "Epoch 105/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1894 - accuracy: 0.9354 - val_loss: 0.1843 - val_accuracy: 0.9457\n",
      "Epoch 106/200\n",
      "18/18 [==============================] - 0s 7ms/step - loss: 0.1967 - accuracy: 0.9345 - val_loss: 0.1739 - val_accuracy: 0.9457\n",
      "Epoch 107/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.1974 - accuracy: 0.9336 - val_loss: 0.1726 - val_accuracy: 0.9457\n",
      "Epoch 108/200\n",
      "18/18 [==============================] - 0s 7ms/step - loss: 0.1981 - accuracy: 0.9314 - val_loss: 0.2425 - val_accuracy: 0.9293\n",
      "Epoch 109/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2173 - accuracy: 0.9251 - val_loss: 0.1945 - val_accuracy: 0.9402\n",
      "Epoch 110/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1956 - accuracy: 0.9305 - val_loss: 0.1823 - val_accuracy: 0.9402\n",
      "Epoch 111/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1860 - accuracy: 0.9362 - val_loss: 0.1747 - val_accuracy: 0.9457\n",
      "Epoch 112/200\n",
      "18/18 [==============================] - 0s 11ms/step - loss: 0.1857 - accuracy: 0.9368 - val_loss: 0.1718 - val_accuracy: 0.9457\n",
      "Epoch 113/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1855 - accuracy: 0.9359 - val_loss: 0.1763 - val_accuracy: 0.9402\n",
      "Epoch 114/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1955 - accuracy: 0.9339 - val_loss: 0.1776 - val_accuracy: 0.9402\n",
      "Epoch 115/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1846 - accuracy: 0.9362 - val_loss: 0.1726 - val_accuracy: 0.9457\n",
      "Epoch 116/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1851 - accuracy: 0.9368 - val_loss: 0.2057 - val_accuracy: 0.9348\n",
      "Epoch 117/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1901 - accuracy: 0.9339 - val_loss: 0.1846 - val_accuracy: 0.9402\n",
      "Epoch 118/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1879 - accuracy: 0.9351 - val_loss: 0.1713 - val_accuracy: 0.9511\n",
      "Epoch 119/200\n",
      "18/18 [==============================] - 0s 8ms/step - loss: 0.1843 - accuracy: 0.9359 - val_loss: 0.1796 - val_accuracy: 0.9402\n",
      "Epoch 120/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.1944 - accuracy: 0.9299 - val_loss: 0.1781 - val_accuracy: 0.9293\n",
      "Epoch 121/200\n",
      "18/18 [==============================] - 0s 7ms/step - loss: 0.2291 - accuracy: 0.9208 - val_loss: 0.2213 - val_accuracy: 0.9293\n",
      "Epoch 122/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.1939 - accuracy: 0.9331 - val_loss: 0.1731 - val_accuracy: 0.9402\n",
      "Epoch 123/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.1812 - accuracy: 0.9368 - val_loss: 0.1709 - val_accuracy: 0.9348\n",
      "Epoch 124/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1834 - accuracy: 0.9356 - val_loss: 0.1699 - val_accuracy: 0.9457\n",
      "Epoch 125/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 0.1887 - accuracy: 0.9345 - val_loss: 0.1766 - val_accuracy: 0.9293\n",
      "Epoch 126/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1821 - accuracy: 0.9365 - val_loss: 0.1731 - val_accuracy: 0.9402\n",
      "Epoch 127/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 0.1835 - accuracy: 0.9354 - val_loss: 0.2763 - val_accuracy: 0.9185\n",
      "Epoch 128/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2092 - accuracy: 0.9248 - val_loss: 0.1693 - val_accuracy: 0.9402\n",
      "Epoch 129/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 0.2185 - accuracy: 0.9296 - val_loss: 0.4066 - val_accuracy: 0.8967\n",
      "Epoch 130/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2183 - accuracy: 0.9199 - val_loss: 0.2297 - val_accuracy: 0.9239\n",
      "Epoch 131/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2005 - accuracy: 0.9334 - val_loss: 0.1877 - val_accuracy: 0.9293\n",
      "Epoch 132/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1819 - accuracy: 0.9371 - val_loss: 0.1702 - val_accuracy: 0.9402\n",
      "Epoch 133/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 0.1843 - accuracy: 0.9325 - val_loss: 0.2430 - val_accuracy: 0.9239\n",
      "Epoch 134/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 0.2172 - accuracy: 0.9276 - val_loss: 0.2046 - val_accuracy: 0.9293\n",
      "Epoch 135/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2050 - accuracy: 0.9276 - val_loss: 0.2485 - val_accuracy: 0.9185\n",
      "Epoch 136/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1869 - accuracy: 0.9345 - val_loss: 0.1834 - val_accuracy: 0.9348\n",
      "Epoch 137/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1799 - accuracy: 0.9359 - val_loss: 0.1683 - val_accuracy: 0.9402\n",
      "Epoch 138/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1786 - accuracy: 0.9365 - val_loss: 0.1755 - val_accuracy: 0.9348\n",
      "Epoch 139/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1762 - accuracy: 0.9379 - val_loss: 0.1664 - val_accuracy: 0.9402\n",
      "Epoch 140/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.1767 - accuracy: 0.9365 - val_loss: 0.2088 - val_accuracy: 0.9293\n",
      "Epoch 141/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1987 - accuracy: 0.9282 - val_loss: 0.2603 - val_accuracy: 0.9293\n",
      "Epoch 142/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2610 - accuracy: 0.9168 - val_loss: 0.2229 - val_accuracy: 0.9293\n",
      "Epoch 143/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2158 - accuracy: 0.9233 - val_loss: 0.2291 - val_accuracy: 0.9293\n",
      "Epoch 144/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1875 - accuracy: 0.9342 - val_loss: 0.1893 - val_accuracy: 0.9293\n",
      "Epoch 145/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1847 - accuracy: 0.9336 - val_loss: 0.1683 - val_accuracy: 0.9402\n",
      "Epoch 146/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1744 - accuracy: 0.9399 - val_loss: 0.1903 - val_accuracy: 0.9293\n",
      "Epoch 147/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1744 - accuracy: 0.9391 - val_loss: 0.1670 - val_accuracy: 0.9402\n",
      "Epoch 148/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1862 - accuracy: 0.9362 - val_loss: 0.2542 - val_accuracy: 0.9239\n",
      "Epoch 149/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 0.1959 - accuracy: 0.9334 - val_loss: 0.1696 - val_accuracy: 0.9402\n",
      "Epoch 150/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 0.1774 - accuracy: 0.9365 - val_loss: 0.1764 - val_accuracy: 0.9348\n",
      "Epoch 151/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1750 - accuracy: 0.9388 - val_loss: 0.1781 - val_accuracy: 0.9348\n",
      "Epoch 152/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1749 - accuracy: 0.9396 - val_loss: 0.1659 - val_accuracy: 0.9402\n",
      "Epoch 153/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1719 - accuracy: 0.9394 - val_loss: 0.1672 - val_accuracy: 0.9348\n",
      "Epoch 154/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1717 - accuracy: 0.9405 - val_loss: 0.1653 - val_accuracy: 0.9402\n",
      "Epoch 155/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1737 - accuracy: 0.9411 - val_loss: 0.1665 - val_accuracy: 0.9402\n",
      "Epoch 156/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1710 - accuracy: 0.9399 - val_loss: 0.1675 - val_accuracy: 0.9402\n",
      "Epoch 157/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1789 - accuracy: 0.9388 - val_loss: 0.1885 - val_accuracy: 0.9293\n",
      "Epoch 158/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1893 - accuracy: 0.9345 - val_loss: 0.2057 - val_accuracy: 0.9293\n",
      "Epoch 159/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1862 - accuracy: 0.9316 - val_loss: 0.2142 - val_accuracy: 0.9293\n",
      "Epoch 160/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 0.1877 - accuracy: 0.9356 - val_loss: 0.1675 - val_accuracy: 0.9348\n",
      "Epoch 161/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2373 - accuracy: 0.9253 - val_loss: 0.3157 - val_accuracy: 0.9130\n",
      "Epoch 162/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2582 - accuracy: 0.9153 - val_loss: 0.2542 - val_accuracy: 0.9185\n",
      "Epoch 163/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.3329 - accuracy: 0.8985 - val_loss: 0.5431 - val_accuracy: 0.8587\n",
      "Epoch 164/200\n",
      "18/18 [==============================] - 0s 7ms/step - loss: 0.2716 - accuracy: 0.9170 - val_loss: 0.1987 - val_accuracy: 0.9348\n",
      "Epoch 165/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.1925 - accuracy: 0.9345 - val_loss: 0.1856 - val_accuracy: 0.9293\n",
      "Epoch 166/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.1823 - accuracy: 0.9371 - val_loss: 0.1688 - val_accuracy: 0.9402\n",
      "Epoch 167/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1724 - accuracy: 0.9391 - val_loss: 0.1732 - val_accuracy: 0.9402\n",
      "Epoch 168/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1743 - accuracy: 0.9371 - val_loss: 0.1975 - val_accuracy: 0.9348\n",
      "Epoch 169/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.1757 - accuracy: 0.9388 - val_loss: 0.1757 - val_accuracy: 0.9348\n",
      "Epoch 170/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1697 - accuracy: 0.9419 - val_loss: 0.1692 - val_accuracy: 0.9402\n",
      "Epoch 171/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1793 - accuracy: 0.9382 - val_loss: 0.2055 - val_accuracy: 0.9293\n",
      "Epoch 172/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1942 - accuracy: 0.9314 - val_loss: 0.1755 - val_accuracy: 0.9348\n",
      "Epoch 173/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1711 - accuracy: 0.9416 - val_loss: 0.1664 - val_accuracy: 0.9402\n",
      "Epoch 174/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1889 - accuracy: 0.9365 - val_loss: 0.2427 - val_accuracy: 0.9239\n",
      "Epoch 175/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1957 - accuracy: 0.9342 - val_loss: 0.2265 - val_accuracy: 0.9293\n",
      "Epoch 176/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1768 - accuracy: 0.9405 - val_loss: 0.1659 - val_accuracy: 0.9402\n",
      "Epoch 177/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1771 - accuracy: 0.9368 - val_loss: 0.1788 - val_accuracy: 0.9348\n",
      "Epoch 178/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1670 - accuracy: 0.9419 - val_loss: 0.1746 - val_accuracy: 0.9402\n",
      "Epoch 179/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1661 - accuracy: 0.9422 - val_loss: 0.1644 - val_accuracy: 0.9402\n",
      "Epoch 180/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1660 - accuracy: 0.9428 - val_loss: 0.1812 - val_accuracy: 0.9293\n",
      "Epoch 181/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1665 - accuracy: 0.9419 - val_loss: 0.1769 - val_accuracy: 0.9348\n",
      "Epoch 182/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1725 - accuracy: 0.9405 - val_loss: 0.1706 - val_accuracy: 0.9348\n",
      "Epoch 183/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 0.2005 - accuracy: 0.9339 - val_loss: 0.2071 - val_accuracy: 0.9348\n",
      "Epoch 184/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1795 - accuracy: 0.9405 - val_loss: 0.2612 - val_accuracy: 0.9239\n",
      "Epoch 185/200\n",
      "18/18 [==============================] - 0s 4ms/step - loss: 0.1811 - accuracy: 0.9379 - val_loss: 0.1838 - val_accuracy: 0.9293\n",
      "Epoch 186/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1646 - accuracy: 0.9428 - val_loss: 0.1669 - val_accuracy: 0.9402\n",
      "Epoch 187/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1639 - accuracy: 0.9416 - val_loss: 0.1729 - val_accuracy: 0.9348\n",
      "Epoch 188/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1646 - accuracy: 0.9436 - val_loss: 0.1779 - val_accuracy: 0.9348\n",
      "Epoch 189/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1707 - accuracy: 0.9411 - val_loss: 0.1715 - val_accuracy: 0.9348\n",
      "Epoch 190/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.1703 - accuracy: 0.9419 - val_loss: 0.2750 - val_accuracy: 0.9293\n",
      "Epoch 191/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.1767 - accuracy: 0.9399 - val_loss: 0.1803 - val_accuracy: 0.9402\n",
      "Epoch 192/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.1683 - accuracy: 0.9414 - val_loss: 0.1820 - val_accuracy: 0.9348\n",
      "Epoch 193/200\n",
      "18/18 [==============================] - 0s 7ms/step - loss: 0.1660 - accuracy: 0.9419 - val_loss: 0.2064 - val_accuracy: 0.9348\n",
      "Epoch 194/200\n",
      "18/18 [==============================] - 0s 9ms/step - loss: 0.2376 - accuracy: 0.9239 - val_loss: 0.2743 - val_accuracy: 0.9239\n",
      "Epoch 195/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.1791 - accuracy: 0.9385 - val_loss: 0.1860 - val_accuracy: 0.9293\n",
      "Epoch 196/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1624 - accuracy: 0.9439 - val_loss: 0.1955 - val_accuracy: 0.9293\n",
      "Epoch 197/200\n",
      "18/18 [==============================] - 0s 6ms/step - loss: 0.1654 - accuracy: 0.9448 - val_loss: 0.1849 - val_accuracy: 0.9293\n",
      "Epoch 198/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.1635 - accuracy: 0.9445 - val_loss: 0.1766 - val_accuracy: 0.9293\n",
      "Epoch 199/200\n",
      "18/18 [==============================] - 0s 8ms/step - loss: 0.2000 - accuracy: 0.9339 - val_loss: 0.2407 - val_accuracy: 0.9239\n",
      "Epoch 200/200\n",
      "18/18 [==============================] - 0s 5ms/step - loss: 0.2031 - accuracy: 0.9311 - val_loss: 0.1709 - val_accuracy: 0.9402\n"
     ]
    }
   ],
   "source": [
    "# Fitting the model\n",
    "history = model.fit(\n",
    "  X_train_spam,\n",
    "  y_train_spam1,\n",
    "  epochs = n_epochs_spam,\n",
    "  batch_size = batch_size_spam,\n",
    "  validation_split = val_size_spam\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-block alert-success\">\n",
    "This is the most important part where you can access the state of your model. In general, as the network has just been trained, the accuracy on both training and validation data should be low and progressively increasing. After awhile, the accuracy on the training data should keep increasing while the accuracy on the validation set might drop when the network is trained long enough and starts to overfit the training data. One should stop the training process before the accuracy on the validation set drops too much (too overfitting). <code>Early stopping</code> is a method allowing to end the training process when the validation accuracy starts to decrease. However, we don't discuss it here because the training process is not so complicated. This graph can also provide many insignts related to the state of the model, yielding the adjustment of the model's architecture and hyperparameter tuning. For more information,see for example, <a href=\"https://rstudio-conf-2020.github.io/dl-keras-tf/notebooks/learning-curve-diagnostics.nb.html\" class=\"blank\">https://rstudio-conf-2020.github.io/dl-keras-tf/notebooks/learning-curve-diagnostics.nb.html</a>.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the history\n",
    "plot_history(history, 'accuracy', showfig=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "loss = 0.16145944595336914 , accuracy = 0.9478827118873596\n"
     ]
    }
   ],
   "source": [
    "# Measure the performance of the model\n",
    "loss, acc = model.evaluate(X_test_spam, y_test_spam1, verbose=0)\n",
    "\n",
    "print(\"loss = {} , accuracy = {}\".format(loss, acc))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-block alert-danger\"> \n",
    "<b>Q5.</b> Play with different values of hyperparameters including <code>learning_rate, hidden_units, batch_size</code> and <code>n_epoch</code>. Notice the change in learning curves according to each option of these hyperparameters.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.2 `Mnist` data with DNN\n",
    "\n",
    "Use what you have learned to build your own DNN classifier. Good luck!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# TO DO"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# TO DO"
   ]
  }
 ],
 "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": "3.9.7"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}