From logistic to deep learning.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<img style = \"float: right\" src = \"https://brighterion.com/wp-content/uploads/2016/07/Neural.jpg\" width = \"300\"/>\n",
    "\n",
    "## From Logistic Regression to Deep Learning\n",
    "\n",
    "Learning objectives:\n",
    "* [Objective 1] OLS\n",
    "* [Objective 2] Logistic Regression\n",
    "* [Objective 3] Neural Network\n",
    "* [Objective 4] OLS Cannot Handle Multicolinearity\n",
    "* [Objective 5] Deep learning is imune of multicolinearity\n",
    "* [Objective 6] Decision Tree is imune of multicolinearity\n",
    "* [Objective 7] Which one is more complex (more weighting coefficients)?\n",
    "* [Objective 8] Build more hidden layers\n",
    "* [Objective 9] Standardize data for better modeling performance\n",
    "* [Objective 10] L1 and L2 Regularization\n",
    "* [Objective 11] Dropout Regularization"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load a dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(506, 6)\n"
     ]
    }
   ],
   "source": [
    "from sklearn.datasets import load_boston\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "X, y = load_boston(return_X_y=True)\n",
    "X = pd.DataFrame(X)\n",
    "y = pd.DataFrame(y)\n",
    "X.columns = load_boston().feature_names\n",
    "          # 'CRIM', 'ZN', 'INDUS', 'CHAS', 'NOX', 'RM', 'AGE', 'DIS', 'RAD',\n",
    "          # 'TAX', 'PTRATIO', 'B', 'LSTAT'\n",
    "features = ['CRIM','ZN','RM','AGE','DIS','LSTAT']\n",
    "X = X[features]\n",
    "y.columns = ['MEDV']   # MEDV - Median value of owner-occupied homes in $1000's\n",
    "y['Y'] = np.where(y>y.mean(),1,0)\n",
    "print(X.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>MEDV</th>\n",
       "      <th>Y</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>24.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>21.6</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>34.7</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>33.4</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>36.2</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   MEDV  Y\n",
       "0  24.0  1\n",
       "1  21.6  0\n",
       "2  34.7  1\n",
       "3  33.4  1\n",
       "4  36.2  1"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.model_selection import train_test_split\n",
    "#split the final dataset into train and test with 80:20 \n",
    "x_train, x_test, y_train, y_test = train_test_split(X[features], y['Y'], test_size=0.2,random_state=1234)\n",
    "\n",
    "scaler = StandardScaler() \n",
    "scaler.fit(x_train)    # Only the training data are used to fit the scaler transformation,\n",
    "x_train_scaled = scaler.transform(x_train) \n",
    "x_test_scaled = scaler.transform(x_test)  # then the scaler is used to transform the test input data."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### [Learning Objective 1] OLS"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>OLS Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>            <td>Y</td>        <th>  R-squared (uncentered):</th>      <td>   0.682</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                   <td>OLS</td>       <th>  Adj. R-squared (uncentered):</th> <td>   0.677</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>             <td>Least Squares</td>  <th>  F-statistic:       </th>          <td>   142.1</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>             <td>Thu, 19 Nov 2020</td> <th>  Prob (F-statistic):</th>          <td>1.08e-95</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                 <td>12:01:31</td>     <th>  Log-Likelihood:    </th>          <td> -159.89</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>No. Observations:</th>      <td>   404</td>      <th>  AIC:               </th>          <td>   331.8</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Df Residuals:</th>          <td>   398</td>      <th>  BIC:               </th>          <td>   355.8</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Df Model:</th>              <td>     6</td>      <th>                     </th>              <td> </td>   \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>      <td>nonrobust</td>    <th>                     </th>              <td> </td>   \n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "    <td></td>       <th>coef</th>     <th>std err</th>      <th>t</th>      <th>P>|t|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>CRIM</th>  <td>   -0.0025</td> <td>    0.002</td> <td>   -1.008</td> <td> 0.314</td> <td>   -0.007</td> <td>    0.002</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>ZN</th>    <td>    0.0028</td> <td>    0.001</td> <td>    2.694</td> <td> 0.007</td> <td>    0.001</td> <td>    0.005</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>RM</th>    <td>    0.1901</td> <td>    0.015</td> <td>   12.486</td> <td> 0.000</td> <td>    0.160</td> <td>    0.220</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>AGE</th>   <td>   -0.0042</td> <td>    0.001</td> <td>   -3.898</td> <td> 0.000</td> <td>   -0.006</td> <td>   -0.002</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>DIS</th>   <td>   -0.0548</td> <td>    0.014</td> <td>   -3.930</td> <td> 0.000</td> <td>   -0.082</td> <td>   -0.027</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>LSTAT</th> <td>   -0.0247</td> <td>    0.003</td> <td>   -7.417</td> <td> 0.000</td> <td>   -0.031</td> <td>   -0.018</td>\n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "  <th>Omnibus:</th>       <td>17.764</td> <th>  Durbin-Watson:     </th> <td>   2.037</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Prob(Omnibus):</th> <td> 0.000</td> <th>  Jarque-Bera (JB):  </th> <td>  13.090</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Skew:</th>          <td> 0.334</td> <th>  Prob(JB):          </th> <td> 0.00144</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Kurtosis:</th>      <td> 2.424</td> <th>  Cond. No.          </th> <td>    82.4</td>\n",
       "</tr>\n",
       "</table><br/><br/>Notes:<br/>[1] R² is computed without centering (uncentered) since the model does not contain a constant.<br/>[2] Standard Errors assume that the covariance matrix of the errors is correctly specified."
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                                 OLS Regression Results                                \n",
       "=======================================================================================\n",
       "Dep. Variable:                      Y   R-squared (uncentered):                   0.682\n",
       "Model:                            OLS   Adj. R-squared (uncentered):              0.677\n",
       "Method:                 Least Squares   F-statistic:                              142.1\n",
       "Date:                Thu, 19 Nov 2020   Prob (F-statistic):                    1.08e-95\n",
       "Time:                        12:01:31   Log-Likelihood:                         -159.89\n",
       "No. Observations:                 404   AIC:                                      331.8\n",
       "Df Residuals:                     398   BIC:                                      355.8\n",
       "Df Model:                           6                                                  \n",
       "Covariance Type:            nonrobust                                                  \n",
       "==============================================================================\n",
       "                 coef    std err          t      P>|t|      [0.025      0.975]\n",
       "------------------------------------------------------------------------------\n",
       "CRIM          -0.0025      0.002     -1.008      0.314      -0.007       0.002\n",
       "ZN             0.0028      0.001      2.694      0.007       0.001       0.005\n",
       "RM             0.1901      0.015     12.486      0.000       0.160       0.220\n",
       "AGE           -0.0042      0.001     -3.898      0.000      -0.006      -0.002\n",
       "DIS           -0.0548      0.014     -3.930      0.000      -0.082      -0.027\n",
       "LSTAT         -0.0247      0.003     -7.417      0.000      -0.031      -0.018\n",
       "==============================================================================\n",
       "Omnibus:                       17.764   Durbin-Watson:                   2.037\n",
       "Prob(Omnibus):                  0.000   Jarque-Bera (JB):               13.090\n",
       "Skew:                           0.334   Prob(JB):                      0.00144\n",
       "Kurtosis:                       2.424   Cond. No.                         82.4\n",
       "==============================================================================\n",
       "\n",
       "Notes:\n",
       "[1] R² is computed without centering (uncentered) since the model does not contain a constant.\n",
       "[2] Standard Errors assume that the covariance matrix of the errors is correctly specified.\n",
       "\"\"\""
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import statsmodels.api as sm\n",
    "from sklearn.metrics import mean_squared_error\\\n",
    "# Fit and summarize OLS model\n",
    "ols_mod = sm.OLS(y_train, x_train)\n",
    "ols_res = ols_mod.fit()\n",
    "ols_res.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.12065916075669458"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "OLS_ypred = ols_res.predict(x_test)\n",
    "mean_squared_error(y_test, OLS_ypred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### [Learning objective 2] Logistic regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.10680833477413351"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "clf = LogisticRegression(random_state=0).fit(x_train, y_train)\n",
    "clf_pred = clf.predict_proba(x_test)[:,1]\n",
    "clf.score(x_test, y_test)\n",
    "mean_squared_error(y_test, clf_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'CRIM': array([-0.04244221,  0.02904445,  0.81027456, -0.03071539, -0.67809648,\n",
       "        -0.38792817])}"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "coef_dict = {}\n",
    "for coef, feat in zip(clf.coef_,X[features].columns):\n",
    "    coef_dict[feat] = coef\n",
    "coef_dict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Var</th>\n",
       "      <th>Coef</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>CRIM</td>\n",
       "      <td>-0.042442</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ZN</td>\n",
       "      <td>0.029044</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>RM</td>\n",
       "      <td>0.810275</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>AGE</td>\n",
       "      <td>-0.030715</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>DIS</td>\n",
       "      <td>-0.678096</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>LSTAT</td>\n",
       "      <td>-0.387928</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     Var      Coef\n",
       "0   CRIM -0.042442\n",
       "1     ZN  0.029044\n",
       "2     RM  0.810275\n",
       "3    AGE -0.030715\n",
       "4    DIS -0.678096\n",
       "5  LSTAT -0.387928"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "out = pd.DataFrame({'Var':features,'Coef':clf.coef_[0]})\n",
    "out "
   ]
  },
  {
   "attachments": {
    "c38dac4d-00d0-42db-91a1-bacd4340399b.png": {
     "image/png": "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"
    }
   },
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### [Learning objective 3] Design a deep learning model doing logistic regression\n",
    "\n",
    "* **Dense** implements the operation: output = activation(dot(input, kernel) + bias). It calculates the dot product of the input X matrix and the weights matrix, then adds a bias vector. The activation is the element-wise activation function passed as the activation argument. \n",
    "* **input_dim** is the number of dimensions. For example, if there are six variables, input_dim = 6. Suppose there are 500 observations and six variables. X becomes a matrix of (500,6). W is a matrix of (6,1). The dot product for X and W becomes (500,1).\n",
    "\n",
    "\n",
    "![image.png](attachment:c38dac4d-00d0-42db-91a1-bacd4340399b.png)\n",
    "\n",
    "* Notice that the deep learning model take much more time than the logistic model.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['loss', 'mean_squared_error'] [0.11228352040052414, 0.11228352040052414]\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Activation \n",
    "#from keras.utils.visualize_util import plot\n",
    "dl_mod = Sequential()\n",
    "dl_mod.add(Dense(1, input_dim=6)) \n",
    "dl_mod.add(Activation(activation='sigmoid'))\n",
    "#dl_mod.add(Dense(1))\n",
    "#dl_mod.add(Activation(activation='sigmoid'))\n",
    "dl_mod.compile(optimizer='RMSprop', loss='mse', metrics=['mean_squared_error'])\n",
    "dl_mod.fit(x_train, y_train, epochs=600, verbose=0)\n",
    "score = dl_mod.evaluate(x_test,y_test, verbose=0)\n",
    "print( dl_mod.metrics_names, score)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.002381</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.221999</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.025847</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.990582</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.005449</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>399</th>\n",
       "      <td>0.992567</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>400</th>\n",
       "      <td>0.458391</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>401</th>\n",
       "      <td>0.250994</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>402</th>\n",
       "      <td>0.000392</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>403</th>\n",
       "      <td>0.920506</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>404 rows × 1 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "            0\n",
       "0    0.002381\n",
       "1    0.221999\n",
       "2    0.025847\n",
       "3    0.990582\n",
       "4    0.005449\n",
       "..        ...\n",
       "399  0.992567\n",
       "400  0.458391\n",
       "401  0.250994\n",
       "402  0.000392\n",
       "403  0.920506\n",
       "\n",
       "[404 rows x 1 columns]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dl_ypred = dl_mod.predict(x_train)\n",
    "pd.DataFrame(dl_ypred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.11228351080999227"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.metrics import mean_squared_error, r2_score\n",
    "dl_ypred = dl_mod.predict(x_test)\n",
    "mean_squared_error(y_test, dl_ypred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Notice that Dense() and Activation() can be combined into one line of code as below:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['loss', 'mean_squared_error'] [0.5588235259056091, 0.5588235259056091]\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Activation \n",
    "dl_mod = Sequential()\n",
    "dl_mod.add(Dense(1, input_dim=6, activation='sigmoid'))\n",
    "dl_mod.compile(optimizer='rmsprop', loss='mse', metrics=['mean_squared_error'])\n",
    "dl_mod.fit(x_train, y_train, epochs=1000, verbose=0)\n",
    "score = dl_mod.evaluate(x_test,y_test, verbose=0)\n",
    "print( dl_mod.metrics_names, score)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### [Learning objective 4] Regression cannot handle multicolinearity\n",
    "* Notie the warning notes in the end of the summary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>OLS Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>            <td>Y</td>        <th>  R-squared (uncentered):</th>      <td>   0.692</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                   <td>OLS</td>       <th>  Adj. R-squared (uncentered):</th> <td>   0.688</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>             <td>Least Squares</td>  <th>  F-statistic:       </th>          <td>   187.2</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>             <td>Thu, 19 Nov 2020</td> <th>  Prob (F-statistic):</th>          <td>2.22e-124</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                 <td>12:01:57</td>     <th>  Log-Likelihood:    </th>          <td> -196.40</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>No. Observations:</th>      <td>   506</td>      <th>  AIC:               </th>          <td>   404.8</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Df Residuals:</th>          <td>   500</td>      <th>  BIC:               </th>          <td>   430.2</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Df Model:</th>              <td>     6</td>      <th>                     </th>              <td> </td>    \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>      <td>nonrobust</td>    <th>                     </th>              <td> </td>    \n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "     <td></td>       <th>coef</th>     <th>std err</th>      <th>t</th>      <th>P>|t|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>CRIM</th>   <td>   -0.0009</td> <td>    0.002</td> <td>   -0.424</td> <td> 0.672</td> <td>   -0.005</td> <td>    0.003</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>ZN</th>     <td>    0.0025</td> <td>    0.001</td> <td>    2.715</td> <td> 0.007</td> <td>    0.001</td> <td>    0.004</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>RM</th>     <td>    0.1932</td> <td>    0.013</td> <td>   15.055</td> <td> 0.000</td> <td>    0.168</td> <td>    0.218</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>AGE</th>    <td>   -0.0042</td> <td>    0.001</td> <td>   -4.514</td> <td> 0.000</td> <td>   -0.006</td> <td>   -0.002</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>DIS</th>    <td>   -0.0548</td> <td>    0.012</td> <td>   -4.664</td> <td> 0.000</td> <td>   -0.078</td> <td>   -0.032</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>LSTAT</th>  <td>   -0.0009</td> <td>    0.000</td> <td>   -8.629</td> <td> 0.000</td> <td>   -0.001</td> <td>   -0.001</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>LSTAT2</th> <td>   -0.0017</td> <td>    0.000</td> <td>   -8.629</td> <td> 0.000</td> <td>   -0.002</td> <td>   -0.001</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>LSTAT3</th> <td>   -0.0026</td> <td>    0.000</td> <td>   -8.629</td> <td> 0.000</td> <td>   -0.003</td> <td>   -0.002</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>LSTAT4</th> <td>   -0.0035</td> <td>    0.000</td> <td>   -8.629</td> <td> 0.000</td> <td>   -0.004</td> <td>   -0.003</td>\n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "  <th>Omnibus:</th>       <td>19.421</td> <th>  Durbin-Watson:     </th> <td>   1.325</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Prob(Omnibus):</th> <td> 0.000</td> <th>  Jarque-Bera (JB):  </th> <td>  13.852</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Skew:</th>          <td> 0.292</td> <th>  Prob(JB):          </th> <td>0.000982</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Kurtosis:</th>      <td> 2.439</td> <th>  Cond. No.          </th> <td>1.14e+17</td>\n",
       "</tr>\n",
       "</table><br/><br/>Notes:<br/>[1] R² is computed without centering (uncentered) since the model does not contain a constant.<br/>[2] Standard Errors assume that the covariance matrix of the errors is correctly specified.<br/>[3] The smallest eigenvalue is 4.48e-28. This might indicate that there are<br/>strong multicollinearity problems or that the design matrix is singular."
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                                 OLS Regression Results                                \n",
       "=======================================================================================\n",
       "Dep. Variable:                      Y   R-squared (uncentered):                   0.692\n",
       "Model:                            OLS   Adj. R-squared (uncentered):              0.688\n",
       "Method:                 Least Squares   F-statistic:                              187.2\n",
       "Date:                Thu, 19 Nov 2020   Prob (F-statistic):                   2.22e-124\n",
       "Time:                        12:01:57   Log-Likelihood:                         -196.40\n",
       "No. Observations:                 506   AIC:                                      404.8\n",
       "Df Residuals:                     500   BIC:                                      430.2\n",
       "Df Model:                           6                                                  \n",
       "Covariance Type:            nonrobust                                                  \n",
       "==============================================================================\n",
       "                 coef    std err          t      P>|t|      [0.025      0.975]\n",
       "------------------------------------------------------------------------------\n",
       "CRIM          -0.0009      0.002     -0.424      0.672      -0.005       0.003\n",
       "ZN             0.0025      0.001      2.715      0.007       0.001       0.004\n",
       "RM             0.1932      0.013     15.055      0.000       0.168       0.218\n",
       "AGE           -0.0042      0.001     -4.514      0.000      -0.006      -0.002\n",
       "DIS           -0.0548      0.012     -4.664      0.000      -0.078      -0.032\n",
       "LSTAT         -0.0009      0.000     -8.629      0.000      -0.001      -0.001\n",
       "LSTAT2        -0.0017      0.000     -8.629      0.000      -0.002      -0.001\n",
       "LSTAT3        -0.0026      0.000     -8.629      0.000      -0.003      -0.002\n",
       "LSTAT4        -0.0035      0.000     -8.629      0.000      -0.004      -0.003\n",
       "==============================================================================\n",
       "Omnibus:                       19.421   Durbin-Watson:                   1.325\n",
       "Prob(Omnibus):                  0.000   Jarque-Bera (JB):               13.852\n",
       "Skew:                           0.292   Prob(JB):                     0.000982\n",
       "Kurtosis:                       2.439   Cond. No.                     1.14e+17\n",
       "==============================================================================\n",
       "\n",
       "Notes:\n",
       "[1] R² is computed without centering (uncentered) since the model does not contain a constant.\n",
       "[2] Standard Errors assume that the covariance matrix of the errors is correctly specified.\n",
       "[3] The smallest eigenvalue is 4.48e-28. This might indicate that there are\n",
       "strong multicollinearity problems or that the design matrix is singular.\n",
       "\"\"\""
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# OLS\n",
    "X_dup = X[features].copy()\n",
    "X_dup['LSTAT2'] = X_dup['LSTAT'] *2\n",
    "X_dup['LSTAT3'] = X_dup['LSTAT'] *3\n",
    "X_dup['LSTAT4'] = X_dup['LSTAT'] *4\n",
    "mod = sm.OLS(y['Y'], X_dup)\n",
    "res = mod.fit()\n",
    "res.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.09566923935424808"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Logistic regression cannot handle multicolinearity\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "clf = LogisticRegression(random_state=0).fit(X_dup, y['Y'])\n",
    "clf_pred = clf.predict_proba(X_dup)[:,1]\n",
    "clf.score(X_dup, y['Y'])\n",
    "mean_squared_error(y['Y'], clf_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[-0.03126084,  0.0202013 ,  1.15042326, -0.02406476, -0.55078628,\n",
       "        -0.01245061, -0.02490122, -0.03735183, -0.04980244]])"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf.coef_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### [Learning objective 5] Deep learning is imune of multicolinearity\n",
    "* Try to change epochs from 1000 to 3000\n",
    "* Keras does not have a parameter to set the random seed. There are two ways to set seed. \n",
    "    * use numpy seed - the best\n",
    "    * use tensorflow set_random_seed"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "from numpy.random import seed\n",
    "seed(1)\n",
    "\n",
    "#from tensorflow import set_random_seed\n",
    "#set_random_seed(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['loss', 'mean_squared_error'] [0.09585237503051758, 0.09585237503051758]\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Activation \n",
    "#from keras.utils.visualize_util import plot\n",
    "dl_dup = Sequential()\n",
    "dl_dup.add(Dense(1, input_dim=9, activation='sigmoid')) \n",
    "\n",
    "dl_dup.compile(optimizer='rmsprop', loss='mse', metrics=['mean_squared_error'])\n",
    "dl_dup.fit(X_dup, y['Y'], epochs=600, verbose=0)\n",
    "dl_dup_score = dl_dup.evaluate(X_dup,y['Y'], verbose=0)\n",
    "print( dl_dup.metrics_names, dl_dup_score)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### [Learning objective 6] Deep learning is imune of multicolinearity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DecisionTreeClassifier(max_depth=10, min_samples_leaf=10)"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.tree import DecisionTreeClassifier # for classification\n",
    "from sklearn.tree import DecisionTreeRegressor # for regression\n",
    "\n",
    "# First, specify the model. \n",
    "dtree = DecisionTreeClassifier(min_samples_leaf = 10, max_depth = 10)\n",
    "# Then, train the model.\n",
    "dtree.fit(X_dup,y['Y'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.06957760715607728"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dtree_pred = dtree.predict_proba(X_dup)[:,1]\n",
    "mean_squared_error(y['Y'], dtree_pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### [Learning objective 7] Understanding Model Complexity\n",
    "\n",
    "Assume you are given \n",
    "* 10 billion rows, 1111 columns (variables)\n",
    "* Of these 1111 variables, 1100 are numerical and 10 are categorical with 400 levels each\n",
    "* 1 target variable (0 or 1)\t\n",
    "\n",
    "Consider two models, which DL model is more complex (more weighting parameters)?\n",
    "* Model 1: 4 hidden layers with 400 neurons each.\n",
    "* Model 2: 1 hidden layer with 500 neurons\n",
    "\n",
    "Let’s compute the number of weightings for each model and compare. In general we multiply the number of neurons in adjacent layers, then add up all the pairs.\n",
    "(Input layer nodes x hidden layer 1 nodes) + (hidden layer 1 nodes + hidden layer 2 nodes) + … + (final hidden layer nodes x output layer nodes)\n",
    "\n",
    "In this question we are told we have some categorical variables as inputs. These must be one-hot encoded, with a separate neuron each, for each input. So the 10 categorical will be transformed to 10 x 400 = 4000 dummy variables. Now we have a total of 1100 + 4000 = 5100 input neurons.\n",
    "\n",
    "Model 1: \n",
    "5100 x 400 + 400 x 400 + 400 x 400 + 400 x 400 + 400 x 1 = 2,520,400 weights\n",
    "\n",
    "Model 2: 5100 x 500 + 500 x 1 = 2,550,500 weights\n",
    "\n",
    "So Model 2 is more complex."
   ]
  },
  {
   "attachments": {
    "192210c0-87eb-4236-8ecf-9ea58b870c2a.png": {
     "image/png": "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"
    },
    "bb80ccac-f394-4386-938a-69e6f479494a.png": {
     "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZkAAAERCAYAAACpRtp7AAAgAElEQVR4Ae2dCXxURbb/1XGc+ftmHJd5Om/mvXGZmec8R3TGBQmIKCqDo7iyI0FZlEVAQBwQVGQTFQRURBYBRUQW2REU2VGWAFlZQtjCEpJAIEBIWD3/z7fa23aa7qTX27eTcz6fpPveW7fq3F9V1+9W1alzLhAVRUARUAQUAUUgSghcEKV8NVtFQBFQBBQBRUCUZLQRKAKKgCKgCEQNASWZqEGrGSsCioAioAgoyWgbUAQUAUVAEYgaAkoyUYNWM1YEFAFFQBFQktE2oAgoAoqAIhA1BJRkogatZqwIKAKKgCKgJKNtQBFQBBQBRSBqCCjJRA1azVgRUAQUAUVASUbbgCKgCCgCikDUEFCSiRq0mrEioAgoAoqAkoy2AUVAEVAEFIGoIVChSOaLL76QHTt2RA0szbhyILB582b55ptv/D7swoULZcuWLX6vh3JhzZo1snjx4lBu1XsqKQIHDhyQTz/9VE6fPu1oBMImmRkzZkhiYqIjHvKXv/yljBgxwhG6qBKxQ8Bfm2zZsqVMmzZN5s6dK3/6059kz549PpV8+eWX5S9/+YvPa5y8/vrr5ZVXXvF7PZQLzZs3l3vvvTeUWyN6z8SJE+Vf//qXZGZmRjTfeMjs7Nmz8uGHH0qzZs2kZs2aQnuZOXOmY1WnHV9wwQWSn5/vWB1RLGySeeONN+Q3v/mNLQ/57LPPytNPP+23LCUZv9BUqgv+2uRvf/tbee2112TVqlXy2GOP+f1xVkaS2bt3r/zzn/80nRYdFxhVJikpKZG7775bfv7zn0vDhg2le/fuUr16dYMH3wOR2bNny1VXXSWnTp0KJHm5ac6dOye02enTp/tMW2lJZuvWrVJcXCyHDx+WFStWyKFDh9wAcX7btm3mOD09XVavXi0nT550X8/JyZF9+/a5j8+cOSObNm0SPvfv3y916tSR++67T7Zv3y4FBQXudNYXb5LJy8uTpUuXmnKsit+1a5dw3lN4o7XOUbGU+d1338nx48fdydCdtzuGpsuWLRPyUXEmAuWRDB0KU2KeQju12qsvkqHdMqVFO/IeyfhrM0xn8Eeb4d6yRgfeIxl05J4lS5bIwYMHjar8pqzfj6X70aNHS53jN0T79Pwdkdb6XaalpZl8rfutz3fffVcaNWoks2bNMh1rZSMZnh9yXbBggQWJ/PDDD2Y087Of/czgd+LEifPaDdPz9EXUw9ChQ00e9G3Z2dkmH+qfvoV2Q/uiHjyFaVfPfoZ2uHPnTqFN0V7Q6Z133jF9ntWHWff7Ihna9aJFi9zTuYzOMjIySvWzPBd9HG0MKSoqku+//17Qm/SWWO03NzfXTOV66mmlCeQz4iMZOvoGDRrIJZdcYgD6f//v/5mHRhlA4TpvkRdddJG5/j//8z8GQK7Xr1/fvE1Ziq9fv96kSU5OlltuucV8B3T+WrdubSVzf3qSTI8ePUwZV155pVx88cXy5z//2RAeUxJ33HGH+x4IjLcPOhZ+mLfffrv84he/MHr++te/lnnz5pm06I7O99xzjyn/oYcecuehX5yFQHkk4/3jHDdunGkjtCvq/uabb3ZPl9E+nnjiCXfbYxrt0ksvdU+XldVmII4777zTTM1Z7bZz584+wfIkGUji8ssvN+X8x3/8h/AbmjNnjpnKIR86IUv4HUB6SMeOHeXCCy+Uyy67zOj7wgsvWMlMe6btW3r46zCs31xlI5m///3v5rftBuzHL2ANZsOGDZMpU6aY78eOHXMns144+vfv78aW9NbsDvVatWpVufbaa919Hu0JEkFIO3bsWHd+1guORTBWffH59ddfu9PxxbMdQxj0kRAifR7pn3vuOYEg6Lfefvtt973kw3XqmnaFrvxx7tZbbzUvRiRGd3ChLXLNU093ZgF8iQrJXHHFFYb5ePP6v//7P6ldu7ZRxQLl/vvvN1MVMCwPZ63plEUysPijjz5qSIhRBT9+b/EkmfHjx0tqaqpJwtsDRMN865gxYwxg1lslby4AuGHDBmncuLH89a9/NWSE7vXq1ZPf//73pkFYutNgGMWogYE3+s45hmTobHnZ8fzjxYfpMqsumcs+cuSI/OpXvzJ1zxvdl19+aTp1a03m888/N+1j0qRJUlhYaDpy2ou1JlNWm+FHSlr0IW8Ihh+8rzl0T5JhBMMUCW+VjPQTEhLMCJ77mM6hQ0MYIdGh8EKF3uSN8QD3vf/++6Zs3p4RfhuQ47fffut+yzUXvP5VVpKBmLt27eqFhuuQ9vHiiy+WSTKQxoQJEwzmjGysGRqrDQwcONCMZoYPH27S0AYR787bIhmu8SLA9VGjRpnZIZc2P/33bMeQzODBg90zPG+++aa5l1Hwgw8+aMjDupNlB9Ykac+0nw4dOphRG/0xfTcvK4ile8+ePYXpVM9ZKSuvQD6jQjIMGy156aWX5IYbbjCHFihMfVkCwTDKQMoiGa7zBvDII49Yt5736UkyXKSiGT4y9XX11VebDoZhLW+GdDZIixYt5KabbjLfGbmw6Iee/HXr1s1UFENfS3fvaRZzo/5zFAJ06hBKmzZtSv3RPrxJxnqrYxrJEjobi2QYKdx4443WJfPJ6NsimbLaDD9S5vUtWbdunWlPvNB4iyfJWNd2795tprYgMl7WkMcff9z9ff78+SY/OgcWqa+77jp328XAgQ6Kjgfh2cGlPKmMJANBMAKw6tQbI97keUEoayTDPZMnTzaY8xJsiXcb4OWYNkPHjZRFMrwscP2TTz6xsiv1afVJni8tENzGjRtl9OjR5l7aBi9I5MN0GC/rPM/rr78uWEly/qOPPnK3m3/84x9y1113mXLQnVE902vhSFRIxtPCq0+fPsKPEvEFCh35NddcY65HkmQYyVCZkAvDSOstloKaNm1qiA0SYiTFW4b11sCwlgVA6481IN4GfOlulNZ/jkOAzpR69RZr4d+zLjEB5Yfmucbn+TbJaNYaiVv5WVMk5bUZb+JgHpyyGKl4i2da1l0gNkYef/vb30wbZoSNYO1EHnQk3EOngNStW1eYWrParfXJCAfxfgEzJ338q4wkAwz0UeDpLYx0wZs1m1BJxttqkD6mffv2pijy9pyG8mx7wZAMo1r6NciSl3ZeksgbYoH0GKkx4rXaflZWljBNTBpmZ6z2wqc1zerZJr1xCeY4JiTDwr0lrM8w74dgOWaxKMcM/QGBNRmEkQwWMP7E+iHxtsAPrlOnTm4WpmOwRi/W2ytvE0yrWIt0VMSrr77qM3urcjzfGnwm1JMxRyAYkvnqq69MG6NztaRt27bukcwzzzzjXvPgOj/m3/3ud+633rLajPePNFCS4XdAJ8GoG6HjsUiG8iFLpjgoe8iQISZNq1at3GnMCa9/1m/D6/R5h5WVZJiK5w3fe60KfOmDwAXrMb5bpu/ebcEayVj1Bri0gdtuu82NM+s5rPvxYovw3apDjp9//nl327NIxpOE3Bl5vbRbulnTo1Y9QjIII3LIDSKqVq2aOWeNZDxH8Z75e7dfz2vBfI8JyTA9BRjMdzPHzNAN6du3r2FiFttZK/nP//zPUiTDlAAjE6YbfG2Gs35IjFDIF5AYhbBoB8NbJMPw+A9/+IPJu1atWm68mCqjTOatma9MSUkx86EkUJJxw+T4L8GQDPXMiAFDDtbpLAsha7rMmmpg2om1Paar6GisqZWy2oz3jzRQknnqqadM+2S6bOXKlfLf//3f7ikywOflCR1o01iTIZZVGHPxvAgxh876o3Xd+m2YxD7+We3dWoNifYH272vt08ftcX8KS1deOHmr57e+du1aGTBggFnLpT4Q+hxwZ+qMkeTDDz9cqi1YnTZTVUlJSeYe2gD30HeBJ1O4lGN1/lWqVDEzLXT01jqK1fbIgNmYJ5980qwve6+JePZJbESnHEZbGKNAmhxb5dCOOObvgw8+MLpBhrywsGbDyIZRGy9dWKch3u3XnAzhX9gk069fP7NYZJXND3bkyJHWoSEOGBSxQKFyAJo/RjLW2wOEgHUXQDASgRT4bi3gAxSLcJyzhnTugkRMZ2GVTYWx2E9apsuYfmDqzhJGMVzjh2gJpoboxgIq17i/SZMm5rI1/+1d0da9+ukcBLzbpKUZLyi0Ae+6pGNlVECd/+///q8wkuEToZNlNEOHzh8/eN4Ee/fuba6X1WZY7/OcarM6KdZmvMUzLdNpjJbQh2k/Nkda64bcx8iea5j0ewrkyu+Pa/zRWUFUiPfv0vM+vmMUY93n+envLdf7/opwzAwHI0br988LJ+tznqbDXbp0MddJw/Q+lqpWW7AsvMCPexE6atakmfbkPOvBvPRaQpnUMdeoL0jIcw2QlxmrPpYvX27dZj492zFTYpb1ILph8MJ9vNhYwjQaL9/0s5ZgIo9+VhnoYq0BebZJK30on2GTDKMCz0pgFOG5UOR53SIZ3rR4UM9hpafymN0xFEXIz1MoC6b2LMO67l02du3khdBZoIsljJp4u+MNzltoLEyh8ekp3see1/S7cxDwbHOeWtF2rHbjXZfcY7UV0ni2afLgRchqr7RNz7bEdV9thukOqx1beniXa533Tssx7ZxPb30ssmJ3vregF6MYrCM9xfu34XlNv5dGwLPfKH3FdYSlIG/9iHdboK4YPVqL/9ZogPNMs/kaGVLH1ojTu64pg7L8vdx6tyf6Vuul3bvvZNTEGqMvIX8MsjzbtXeb9HVfIOfCJplACrHSeJKMdS4WnwDJ6Aq2V1EE4g0B3q4Z6dPZqTgbAYtkYq0l04GMVqZOnWq7KraSDNNemGPG+sfBFAfTYL6sfGyvAS1QEQgSAaZ62W+h4nwE2OPC2k6sBcMA9sd4j3zs0MtWkrHjgbQMRUARUAQUAecgoCTjnLpQTRQBRUARqHAIKMlUuCrVB1IEFAFFwDkIKMk4py5UE0VAEVAEKhwCSjIVrkr1gRQBRUARcA4CSjLOqQvVRBFQBBSBCoeAkkyFq1J9IEVAEVAEnIOAkoxz6kI1UQQUAUWgwiGgJFPhqlQfSBFQBBQB5yCgJOOculBNFAFFQBGocAgoyVS4KtUHUgQUAUXAOQgoyTinLlQTRUARUAQqHAJKMhWuSvWBFAFFQBFwDgJKMs6pC9VEEVAEFIEKh4CSTIWrUn0gRUARUAScg4CSjHPqQjVRBBQBRaDCIaAkU+GqVB9IEVAEFAHnIKAk45y6UE0UAUVAEahwCCjJVLgq1QdSBBQBRcA5CMQNyZw794Ns2p4v46dvlM4DF0i9dpMloclYuf2pUVKl3gjzyTHnuU460nOfiiKgCCgCikBsEHA8yWRk5Umf95fJXQ3HSELTCZKQOF0SEr+Su1stkZrPr5J72q2WWu3Wmk+OOc91k67pBHMf95OPiiKgCCgCioC9CDiWZNak7JXGL30pVRt9LNUSZ0rN51bKvR3WBf3HfdxPPuRHviqKgCKgCCgC9iDgOJI5kH9c2vaZJ3c1HifVWnxlRimhkIv3PYx2yI98yZ9yVBQBRUARUASii4CjSGbB8m1SteEYqd58ltRqvyboUYs3sfg6Jl/ypxzKU1EEFAFFQBGIHgKOIBkW5/uMYN1lnNRsvSwq5OJNOJRDeZSrxgHRa2CasyKgCFRuBGJOMqdPn5X2b8yXhKafmcV7bzKI5jFGA5RL+eihoggoAoqAIhBZBGJKMowg6OCrN5sctemx8kjKTJ81m2z00BFNZBuX5qYIKAKKQExJhqkqRhLRWn8pj2Cs65SPHuijoggoAoqAIhA5BGJGMl8t32bWRJiysjr7WH6iB2s0agwQucalOSkCioAiEBOSwXwY6y67FvkDJS/0QS81b9YfhiKgCCgCkUEgJiTDPhXMiAPt/O1Mh17op6IIKAKKgCIQPgK2kww77tkQGet1GH/EhV7op54Bwm9cmoMioAgoAraTTONu083Oe3+dvBPO4xkAPVUUAUVAEVAEwkPAVpLBSSU+xHDx4gQy8acD+qGnOtUMr3Hp3YqAIqAI2EoyeEPGWaW/zt1J59ETfVUUAUVAEVAEQkfANpJhoyPu+kP1pmw3AaEn+uoGzdAbl96pCCgCioBtJEMAMeLB2E0W4ZSHvuitoggoAoqAIhAaAraRDJEqCSQWTqdv973oi94qioAioAgoAqEhYBvJEBKZiJV2E0U45aEveqsoAoqAIqAIhIaAbSRTr91kExo5nE7f7nsJ5YzeKoqAIqAIKAKhIWAbySQ0HiM1n18V1kjmpRGZMmN5vrwyOsvkk9g/XaYuyZWBE3eGla8/8kLfhCZjQ0NW71IEFAFFQBEQ20jmtic/klphOMPsOSpLzp79QU6fOSdFJWelWd80KTx+xlThlMW5USEZ9EVvFUVAEVAEFIHQELCNZKrUGxHWJsxOw7dIn4+3m5EMj1py6qzs2F8skE/DV1OjRDJr5Z7nV0q97htl6Be7Ja/gZGgo612KgCKgCFRSBGwjGUYEkXDr/+qY7aaqCk+c8UkuLQdmRIxw0Lda0y+kzovr5ZGXNsi97dfJ4/9OlsGf7ZJ9B0sqaZPRx1YEFAFFIHAEbCMZ1jbCXZNh7WTw5N3m6ZYkHy5FJm3e2iQbtx011+r3Sil1zd+aS3nnrTWZrOwieapHssxclifDpuyWp3omuwjn5WR589OdsvtAceCIa0pFQBFQBCoRAraRTCSsy54dkCEnSlzrMLkFpwyRPNgpSRr0TpGCo6dlafJhU3VPQgId1oX952ldlp1bLI16pciXy/JMGYXHT8uIL7OlUa9Uua9DkplS6zd+h2zfd6ISNR99VEVAEVAEykbANpIJZ58MI5PNu4vMQn9y1jFZlFRgnmr2qnxh2gxCqf3COsH6DIkUyXjvk8k9dFKavZYqkxfllEK16MQZ+WjmHmnc20U4j7y0UV4fmyVbs4+XSqcHioAioAhUNgRsI5lwdvz3Hp1l6mXXgWJ5vEey/PujbfLDDz9I8cmzwujBGrVEnmTO3/FfUHhaWvZPl5Ez9vhsK+j08dx90vS1VLnvhXXyr64bBP0371DC8QmYnlQEFIEKjYBtJBOO77L7OyaZ0QmjFYtQ6r28UR7snOQ+5nzEScaP77JjRWek3dubZPDnu8p0oHny9Fn5ZP5+afaGi3DqdlkvPT7MlORtxyp0o9KHUwQUAUXAQsA2krHDCzMmzkjT19NKkY9FTMF8lueFmRFLl2Fbpe+47Wb/jgWov8/TZ3+QSQtzJLFvmtTumCR1u2yQ7h9kyrpNhf5u0fOKgCKgCMQ9AraRDEhFM57Mms2Fcur0OVMhx4vPypSleWERTSDxZE6eOSevfLRNeozMdJcdSItgU+m0JbnybP90s5ZUp/N66TY8U9ZkKOEEgp+mUQQUgfhBwFaSqYiRMSEM1oU6D93itnwLpvpZW5q9PF9av5khTAtiLffi0K2yIsVlKRdMXppWEVAEFAGnIWAryfDwjbtNl2otnO2NGf3QM1CBKN79fLc8/9YmKTxxOtDbfKab+12+PD8oQ+7vlCQPdEqSFwZvkcXrDvlMqycVAUVAEXA6AraTzJqUvXJX43FSq/2asKazgllfCSYteqEfegYr7JthCuxgYXhEY5X79ZqD0v7tzYZsIB2MDb5efdC6rJ+KgCKgCDgeAdtJBkTa9pkn1ZvPciTJoBf6hSos7mO+nHMosn7OlqwvkI5DtrgIp2OSPDcoQ+auzDem3KHqqvcpAoqAIhBtBGJCMgfyj0vVhmOkZutljiIa9EEv9AtHZi/Pkwa9UoR9PdGQ71KOmHWbBzuvN4YDrQdmyIxluQFZuUVDH81TEVAEFAF/CMSEZFDmq+Xb5K6G4yLiNDOY6TB/aXGGiT4Llm/zh1VQ579ee1Ce6JEsmXuKgrov2MSYQGOZ9s8XNxjT6GcGpMvUxQcEk2kVRUARUARijUDMSIYH7zNimSQ0/Szm6zOsw6AH+kRSVqUeNl6b07Ls2XzJJs+XR2wTNn3ibaDFG+ny2cIcYVOoiiKgCCgCsUAgpiTDBs32b8yX6s0mx4xoIBjKRw/0ibSs33JUHnt5o6y1edMlbmx6jsyUh7puMITzdJ9UmTB/n3HFE+ln1PwUAUVAEfCHQExJBqVOnz5rOnhGEpGIN+NvOszXecqjXAgGPaIlGTuPmxHNso0ux57RKsdfvjjqxH/aw902GI/RTV5NkTGz9wiOPcuSjIwMSUtLk5MnfzJi4LioyDUFuGfPHnM9Pz+/rGz0miKgCFRiBGJOMmDPCIKpKtZE7DIGoBzKo9xojGC82xQhAOr3TJGvvotth4weeIjGUzQhChq9lmI8SBO6wFPY+9OwYUO54IILZN26debS1q1bzfHcuXPN8ZgxY8xxjx49PG/V74qAIqAIuBFwBMlY2mAMgHUXZsTR2kdjpseazzLlRGqR39K/vM89eSXSsHeqTF2SW15SW64TbG3A+B0mFg5RPxv0SpUPpu+Rw0ddhJOUlGRIZMGCBUaf9957zxxDLsjChQvloosuEshHRRFQBBQBXwg4imRQEPNh9qmwIZKd97XarY2ImTP5kB/5kn+4Zsq+wAzkXF7BKWn+RppMXLA/kOS2pSGc9Fuf7jTrRxAO0T/7j02Tn11yqUycONHsx7npppvk6quvlv79+xu92rZtK40bNy6l45kzvqfg/J0vdbMeKAKKQIVDwHEkYyHMjntcu1Rt9LHgrBKvyL7WVco7x33cTz7kF8pOfkunSH0yUmBvC6MGpqWcJnkFJ00YA0ywa7ZZJve1niOtek6Tpxq1kLp168oLL7wgWVlZcvnll8vOnTuN+jk5OdKmTRu5//77Sz2Ov/OlEumBIqAIVFgEHEsyFuI41cR7810Nx0hC0wmSkDhdiFhJaOSaz6+SWu1Wm9EOnxxznusmXdMJ5j7uJx8nyfHiM9Jh8GYZNHGnLWtCoT77/1z7J3nkmXfkH499IPdiiff0F1Lzidek9oOPy4ABA0y26enp8tBDD8kTTzwhtWvXdhfl77w7QQX+giHJuvR98v5na6R177ly/7Ofyp0NRsutj39o/vjOOa6RhrTRND6pwFDrozkcAceTjIUfi/MEPiPCJqGc67WbLAlNxsptT34kVeqNMJ8cc57rpCO9HYv6lo7BfpacOivd3t9qFuLPnHWFKQg2j2inv/POO+Xiiy+Wnj17yuGjZ+SJ1kPlb3UHSc3nlssjL20wHqh35rg8G7z//vulSMbSzd9563pF+tyyPV96DV1sCKVGs4mS0HyGJDyzUO5us0zuafu9WWtkXZDvnOMaaUgL8XAveagoAhUFgbghmYoCuPdznD5zTnqP2mYCmDlx0+Rjjz0m1113nRQXu4hk+PDhcumll8qGlE3GKq1x71RjpYa1WoOOE+Xu2o96P6JUBpLJ3HlQnn1ltlRt8LEkJM4yo+rypnK9rzMS517yIC/yVFEE4h0BJRkH1ODZcz/IwE92GAeYRSW+F85jpWZBQYEcPvxTbBv2yHjvizl+8qx8PHefPNhmplnDYT8O+3I27XT5gKvIJHPy1BnpN3K53NlgjFRrMT8iVpGMdMiLPMmbMlQUgXhFQEnGITWHAcDwqdnS5s0MKTwWn53K0KFD5Z5a98on8/dLszdSjacBXNw81v5TqXbvYw5BOnJqZO8vlIefmyTVn55ipr+8RybhHjOlRt6UQVkqikA8IqAk47BaGzVzj7Tomyb5R045TDP/6pw+fVpatGghv//97+WSSy4xZs1s4DxRdErqNu4hVer0kxotv5EaLWZJ69eXStKWo/4zi5MrqZm5ktB4jCQkzgvJ6jEYAqIMyqJMFUUg3hBQknFgjU3+5oA0eS1F9ueXOFC70FTCKzTeofESXbtDktTpvN54j/4+/UhoGcbwLjp7Ng0nPPtN1AnGIiPKokwlmhhWvBYdEgJKMiHBFv2b5qzKM25oLMut6JdoXwlnz/4gs5fnS+s3M+T+jklCXJwXh26VFSk/rf3Yp01wJTFtZUYwNhKMJ9FQtk6dBVdnmjq2CCjJxBb/MktfnHTIxKTZsiu8IGplFhLji6xFzf0uX54flCGEmH6gU5IxgCASqNOEBXjWR+yYIrOIxfuTstFBjQGc1jpUH38IKMn4Q8Yh51enHzEenJMz7YlJE+vH/nrNQWn/9mZ5oNN6M8pp9/Ym+Xq1M0x5sfRiId6747f7GB3QRUURiAcElGTioJYIRvb4v5MFwqlMwmim45AtZjqtdsckeW5QhsxdmR8TVzzsWcGkGIsvu0nFuzx0QBfdR1OZfg3x+6xKMnFSd0yZ4UuMKbTKKKzXsG7D+k3tF9YZ328zluUK6zt2SIses8zeFe8OP1bH7KNhw6aKIuB0BJRknF5DHvphBFD/lRTzNu9xutJ9XZNxxFim/fPFDcZSDYs1LNewYIuG4OaFXfjRCj8RClGhCzqpC5po1LjmGUkElGQiiaYNeeUcPGnMm50Sk8aGRy6ziI1bj8rLI7YJmz7ve2GdJPZNk0kLc+T0qcgRDv7EcPcSChlE8x50QjcVRcDJCCjJOLl2fOi2d+9eWbE6VZ5+PVnGztlrUmRnZ8v+/a74NEePHjUhkXHFX9kkbfsx6TkyUx7qusEQztN9UmXC/H1SfLLs0NpYt70/NdsnXHhGxnElfsWiSRih5I1O6Kbem31WnZ50CAJKMg6piEDVmDJlilx44YXSpVtvsxA+fMpuqVKliiQmJposiFJ5zTXXyA033BBolhUy3ZbdRcZ/Gn7UCDPd5NUUGTN7j+BnzVtw5UOgNtz6eAsu+PGQHAoJ2HEPuqGjiiLgVASUZJxaM2Xodf3110vLli0FZ5ptB6XJX+/vI3X+WdfcQQTKK664wh3rpYxsKs2lrH0nTDgFPEVDOI1eSzEepI+dOG0s1dgQagih/ToZ+sXuUrgQ6wVX/HYQRihloBs6qigCTkVAScapNVOGXjVr1pRHH3W51H+l92tyx1MjpXqjj4SwAUuWLJErr7xSCgt/cqjoKw6+zjEAACAASURBVPQx55haq2yyM+eEDBi/Q+p132hGL+azw7qfSKT9OhMV1MKFoGLEfAmFAOy4B93QUUURcCoCSjJOrZky9GrUqJHUqFFDDh06JIxq3h06TG577F0TAK363fcKMV+QkpISGTRokAmTbIV5zs3NlXvuuUdq1apl8rjjjjtKEVIZxVa4S/vySuTxl5PPJ5D26+Sdz3aZ5yV6JcHFQiGMJ3smy3vTs2Xmijz5cOZemb4sT54dkBFSXv7KRzd0VFEEnIqAkoxTa6YMvbp06SI33nijtG/f3pDIvHnz5KKLfiatei+QmolfyNGiU8JI5YEHHpDOnTvLRRdd5M4NA4GuXbuaaaJTp07JL3/5S5kzZ477emX68u26Q8ZAwGcH3n6dvPXpTrOwHsoGzG4fZApxgg4fOy2LN/zkk63dO5sjSjKujZmjK1O16bPGGQJKMnFWYaj7zjvvmMX/m2++WXCzv2HDBrngggvksssukzdGJZuNioePutYb0tPTS5GM5+NiJIBr/m3btnmerjTfP1uY47vDb7/O+FBjKu22+hOC3h/zRM9kQy5Hi87II903mDLyjpwSIp+61386rJN6L2+UJq+n+dbBcwqvjO/sl7mr+QwZOGGHYF2nogg4DQElGafVSAD6TJo0yZDKsmXLTOqcnBxzPGTIEHP88dy9JmhYXsEp8UcyR44ckVtuuUVGjRoVQIkVM8mHX2YbDwL+RjJ1u2yQWx8bGTTJDJ/mslJjqoy8IRo8E6RkHTPHhDkYNXuvFBWflfnfHwybZP7+5CizGZU9Qs/0S5MZS3PleHF8Br6rmC2tcj+Vkkwc1n9xcbF7Xwzqs96yZ8+eUj69pi3JlYa9U2XxqrTzRjKEUE5ISHCv3cQhBBFRGRLuPHSL305+8Ge7QpouW7jW5frnmR/XX4b9SDoTvzlgymJfzorUI5JfeErmrMr3W75P8vMa1XhPl+Hnru+47YLp9puf7nSHwI4IYJqJIhACAkoyIYAWL7fM/z5fHu++Xn7127+UIqAGDRpIv379zLmlS5fKwoUL4+WRwtYz99BJ+eLbA4J353rdNxhPAf46czxfh7Lwv3l3kdGTTaGP/nujIRNO/PujbfJg5yTje40yN+0qCptk/C38E8J78qIcYUNqywHpMnt5njF5DxtAzUARCBIBJZkgAYun5AMGDJCqD7QyJrgPPNZKJk6cKCtXrjRTa6zhWH/jx4+Pp8cKWldc8RBtFGJ57OWNxnJs7aZCM4W1N79E8PDsTTT1e6bIuXM/GPPgYE2Y5/8YmmDK0jzZmn3CrS8jHLwP4AInUiRTngkzo9z1W47+uE9og3n2LdkVNz6RG2z94hgElGQcUxXRU4QOlc6VzqayCOTx+dc58vygTSZMwuDPd5nnx+LLWwZP2mn2zHgSzYgvXesqoWzG7Dhsixm9sA6zNPmwDPh0pyEs1mDeGLfDTWiRGMkEsxkTYxD8uuH9AC8HhE0oz+WON1Z6rAgEi4CSTLCIxWn6tCxXTJpVqT+Z08bpo/hVOzu3WLAYowN9qkey2b2/YetR08H7u4k3/dfHbpfmb6S5Nmj+uOZhve2H6laGCJ//6uqyLIO8sCSr3dFj02eHdbIrp0S+WXfITTqeJBfo91DcyvDM6zYXSu9R2+SRlzbIu5/vlm17XVN8/nDS84pAqAgoyYSKXBzel7mnyHS+X691RqTJSEAIsXz61X5pNSBdmOIaNmW3sPjNVFd5Qpq3J+6UjkM2y4mSM5KVXWQWzFnHsCQaDjJ7jspyr9PQ4TPqat4vPWiyiYSDzIOFpw1+GIm0fWuTfPVdvpScOt+/m4VHKJ/sx8ITheXElWdevny5bN68OZTs9J44Q0BJJs4qLFx1dx8olga9UmTm8rxws4rZ/bsOFBvvypjrEl/nvanZkpp1tJRxQ3nKMW2Ge5kuw7aWmjLavOu4MQzwvL8yuPqHcFdnHJFXPtomj7600ZD1jv0/rSd54hHMd4jktttuk1//+tdSp04ds0kYl0h//vOf5Te/+U1QdRZMuZrWOQgoyTinLmzTJOfQSWn2WqpZs7Ct0DALosMbP2+fiRfTqFeKfDAtW9K3Hwupk2KtpM/YLOn+QaacOn2uXM0qW9Cy/MOnDNYQeIfBm2XhmoNmI2m5QPlIsHjxYsnMzJQHH3xQHn74YcF8ftq0abJo0SLjLdzHLXqqgiGgJFPBKjTQxykoPC3P9k+XUTP3BHqL7em27zthYuY0eyNVGvdOFTZPZuwMzzIKJ6K9Rm0zcWdOnimfYKyHfqZn5Qu/zGiPNbweIzLNetX707Jld06xBUlQn/jY69atm/ue1157TV544QX38blz50J6YXBnoF8ci4CSjGOrJvqKFZ44bebhWfgNZA0j+hqJWRcZM3uvGWk1fS1VRs7YI1t2hUcslt6MWugwXx2dJWfOBk4w3J+586Dc2WCMhOLHLNBF/EDTuTZgjjE6Wc8W7U82rhIkj3WvjkO2yKK1ByVQkj558qTZEDx6tMvH2sGDB+WPf/yj2UCM3oxs/vKXvwh7tlQqHgJKMhWvToN6IkxYXxy6VfqP2272jQR1c4QSb91dZEZUmNay6D561h7hXCSFxeyu7201u+F9mTEHUla/kcul+tNTgl6gD5Q8Ak2HDugSC2GqcUXKYTPViFk8o8u9uSVlqoKPPPZksfiPEcA///lPef/99809HTp0kF69eskvfvELWbFiRZn5xOPFvEPHZd7STOnz/jJp1n2G1Eocb7xI3Pr4h8IfkU05xzXSkJZ7KpIoyVSk2gzxWcwb/shMM4UUyBpFiMW4b8O6iAV2RilYNWE+zOglWma0EGnnd7cYNyvhjNhOnjojDz83SRIS58WMaCgbHdAl1sImV14InuiRbF5UliQdMjGNvPVas2aNIZlPPvlEateuLe3atXMnoS0gV111VYUhmfyCIhk3faM89NwkuaP+aKmZOEXuenqOJDz7jdz93AozGsaxKX+MSjnHNdKQlnu4lzzIK95FSSbeazBC+vOGis8rRjXR2KBHZ8J6CpscsW7DmSOOPFl3iaYQPbT9O5tNIDKrQwunvOz9hZLQeIzpFAIdeUQqHR0RZaODk4SpxyUbC8xI8fF/J5uoo/vzfxrd4FfvZz/7mYnYOmzYMJ+qVwSSyc4plB5DvjUkUSNxmlRv9a3Uarc26BcS7uFe8oBwyJO841WUZOKg5jIyMiQtLU2Y27aEYyx1EH7EHOfn51uXQ/rkLX/IpF3G/cqxovDflOnUcT+PJRhz+ZgcYyEW6uJxsA/FM7Djf/hU1+79YO/3lz41M1eqNrSXaCAYyqRsJwt7fhihQjZMTy5PLjDrX4cPHxYW9/1JPJNMcclpeWvsKjP1ldBiVkTX7RjpkCfTapRBWfEmSjIOrzE66oYNG5rphnXr1hltrTnuuXNdYXfHjBljrvfo0SMiT8M8e8v+6YIFWrACUbFnhY6dXffk88n8/cKmSTsFB5GtB2aYDi8a5dLZmxGNDVNnTJFRltMJxhNnrPgICsc0JdNpTIdiOu9P2DMTjwv/1Mm9ieOlRvOpck/b74IetQQ6yiVvyqCseGoH1LeSjL9W76DzSUlJhkQWLFhgtHrvvffMMeSC4EWZ6JeQT6Rk4oL9xsILr8XlCQvpeCxmtz0dCp07u/D35P00ZVJeHpG8jo8uRk3RNs9m2or1ERbio2F1Rp7kTRlOmyILpr54wfhg+h559OWNxmBgZcpht5HJl19+KXXr1jXtmfhGGAHEi0yam2osDqs/83XUyMWbhCgLK0fKjhdRkomDmsrNzTU/QrwoM7K56aab5Oqrr5b+/fsb7du2bSuNGzd2PwnxZk6c8L3WQSTNQGXmsjxh4yNeArwFYsEvGObPTI3gLwzni55z8d732HGcf+SUMSSA5OwQFuCx9OKHX63F/KADnHl3IhyzIExe5EneTljkjwSWGJXg0gg3PkyfsiaHaXQ8yqAxq+SuRuOl5vMrbSMYq63UfG6lKRsd4kHihmToXNmr8MnsZOk66Gup136yJDQZK7c/NUqq1BthPjnmPNdJR3rui3dhLvviiy8WFk0nT55sCIW3PzazZWVlyeWXXy47d+400w1VqlSRJ554wrjtaN26tfvRS0pKZNCgQSZtMJh8veagmfbCrxfGAUmbj5pFdMxXn39rk/EagJWRE4QOi701uPW3W2hrLXrMkqoNPpaExFmCXzGrUwj0k3u4lzzY/EmeFVV25hSbKVVCXPcYmSnfpx2RUE3L7cZo4OiVUq3xJ1EZvQbaVhjlogO6OF0cTzJZ2QXy5uhVhlCqNRkvNRKny12J86VGq8Xmh3xPu9XGgoNPfqScN9cTpwvpqzcZKzA++cSz/Pd//7d07drVkMeWLVukRYsWUr9+feMPirgxCC48rHDK7Dlgb8KxY8eMv6gHHnhAOnXqdF6UzPIwwXJo3Lx9UufF9fLwSxuMUQBBv5z2BgrR4RWA0MOxFFzQ4OuMhVo8JOOKn5gvBBejYyhlutpmmblGGtJyD/eSR2WRk6fPyoLv86X925uM1eGE+fuE0ahTZezUDWYUEY3p0UAJxkqHDoym0MnJ4liSycjKMwGjmDLgR3j3c8uDfjOkMriP+8mnde+5Qr7xKHfeeacZzfTs2dOozyL/JZdcIrfffrshEe9nYnf1dddd57boYfSSnp4eEMmwaIuzxLc+3WmcJeK/augXu41rEWLTOE3YDIhZNPFRnCJ4byZMAPFoaHdE2IREPDfhcY5rpCEt91RmwZyddoaDThx1rskodIwnCuplTcpe04+EMkq1iCHSn+hC34ZuThXHkUzh0RJjF1614Viz6Y03v0hUDPlgpUO+2J1TTjzJY489ZkiD9RZk+PDhcumll/pc7E9OTpY//OEPsn79+lKPWBbJ4CKEKYuBE3aYGCPMm09fmlvqrTJjhysmDWapThHMoZnf521YpWIgwD6ted/lG/NzRqcYoYRi6RhJNOgvmBWp3vLbiPRHkejTrDzQCd2c2qc5imRgY8BKSJwuTH9ZIEbyk3zJn3KczP7eP5CCggJhr4El7JHxtS+GqbQbbrhBvvvuOyup+9ObZFiIxQEiLu8JXoW56ZfL8sr8QWftO+GYTh1dMJPGVFalYiJADCSimtI+8TnHmmAwa4qRQoUXU/qNSPZFkcwL3dDRieIYkmFekWGfXW8KlEN5Tp/PDKbRsP7y17/+1fiI4oc4YsQItxNC8mGEc+GFF5oNcuzuf7jbBrPDf/byPMHsN1Bhegp3MIx0YiX4NsOqzUmjqlhhURnKJajc7JX50vrNDGPcMembnKDabDgYMcXODEi0XnwjQTboho5OXA5wBMlgIeEyBwzeIiecCmI+k3LjwUIjkB9J3759zWI/C/7W365du0ykw1Ydesnv/niTOX/tX2tI11c/lCPHQt/Vz8I/LvjtMhX2fH7c00Aw36Ud8Tyt3ysJAnjlJqIpo5s+H2+XjZlHo/rkrJsx1R5OX2PHveiIrk6TmJMMll+xNAe0TAHjxeY80AbEvPaS9QXy2pgsM2J56b1MmbsqT44eD3zEUl5ZjH4Ie4w/MrsEbwIQjBMNEOzCQMtxIYBfOvZy4VWCFx6sHiPZvikFq1RmPCK1NhxNskFHdHWaJW1MSYZdq4wkYm0OaJkCxtMuWl8dDVMKrE+8OnqbIZaXR2w1MduJGxMtOV7sckD5zme7om4JtH7LUWF/DptAVRQBTwQwShn4yQ7T7glbkbLtmOflkL+b7RPNZzh+FGORF5a0TnthjhnJpG49YFg3FjtmrQrx/GQXLW8B6BVPwtscAaQw+fxX1w3S48NMWbj6oND52yWMmnCGSEjjYIOBBaojIxdGMGlZkek8Ai1X08UXArR71goT+6VJi75pMm1JrgTi7BVfd97CuiYbvEPdPuHZv9j1HV0xaoqFcYQ3ftZxTEgGT6I4erPT508glWz0SRzveE+n/JCIu95zZKYhFgiGnfkQTqwEE2jCGhN5kg12kRQs4CCYTWGGXo6kTpqX8xFgarXfeNfoBtP89O2+X1DwZgEheZMRHhfY0B1I3xFKmkVJBXK06IzsP3hSandcF7Fy0NlJ3iJiQjK4rDYeRTtEDthQKtnXPeiFfk4TfgDsBaETxyqMDp2pMabInCK4BeHH3OndLREjvGUbC4zTTUxZVRSBUBBgnWbKtweMT7tn+6cbM/2iEz/9bghvQV9AWGnPkNK4psLDiK9+IhLn6nReb0b+kEwk8rPyQGd0d4rYTjIE32HnczTdYltgh/KJXujnhCBB/DjYlNb9AxexsE9gcdKhqAQVi1SDZJiON2biuIS7CMs0IPtgoh3YLFLPrvk4HwG8hff92GW+j0cLLNMe6rrB3cm/PjbLPdWED0RcVIXSjwRyT+PXXJ6Uv1l3KKJloDO6O0VsJxmzqanFrIiCGkiFBpOGIEGx2tiEWfGcVXnS7f2tZsRCoyfqIDHq40kIy8u8eKh+qOZ/n282fe7y4QE6nnBQXZ2JAL+zz7/OMW3Mu28gyB6Cs118IXpfD+f4yZ7J8t70bJm5Ik+WpbhM8N+dkh3RMtAZ3Z0itpIM8aoJJxpra7LyGgn6oadd8bUxBZ61Is8snjMVxpsWmwwjvbZhd6ObvChHmryaIsF6aQYLQgzEKh6N3ThpebFDgBAVvvqDqYsPmEX/SPop6/ZBpvE0ffjYaVma/JP3jlZvbvKpgy+9AjmHzhgsOEVsJZlx0zeauNWBABXrNMTXRt9oCb6YZi7Pk85Dt5hNZZhdEswJVy8VSdilXf+VlIBDLmMN1OS14ImpImGmz2IPAluyj/vv3Nuvk2rNpkRsl/8TPZMFcmGh/5Hurum548VnzdR37RciuzbN7n9CoDhFbCWZh4gi2Mp5DuZ8ERp6om8khakjXNGzMM5uZRbJsZzyXGyMZHlOyQsDBSJm8qMuS5i+aPZaquPCCJSls16LTwTYc/V0n1S5t/06YQGeGQQcrT79eprZYIzn8buazTBhRHz1D8GeG/7jFBxTZdwL0RCqfEPmUf9EF6JhVK12a02MLafUjG0kk3fouJmCAoBgKygW6dGTKTP0DkfyD58ytvovDN5sXJizYWx1+hHBnX5lEjw8Y4ac7GeT3GcLcySxb+hrOJUJS33W8BHAWtNzwd+7j2HKmtFApPyVLVzrcuL6zADX9NzAiTvNQ3yyMCfi/WGlHcnMXZopNROnRBxQ78YRyWP0Re9gBb9emEy2f2ezicEyaOJOExsjWhsVg9UvVumx5IFoiFXjKWNm7xVMS4Nx0ul5v35XBEJBYOSMPT77IwxW2GDMukak1mQ273aZ4LNhmr/cH8NO9xi5TR7snORTj1D7skq7JtPn/WVy19NzIgpmqJUQ6H3oi96BSM6hkybsb9u3NhnXJ7hZWbe50IQsDuT+ypJm8y6Xc8slSa43uw+/zDaedcM1d64s+OlzRg4BtgN49wWMbrJzXTGbImldNn+1K5T2xG8OCA5eLWGEA6HV7bL+PF28dQv0uNJalzXpPkMSnv0mYkAGCng46dC3WfcZVns473N/fongcpw9Ibyhs6mLud54iVV+3gPZdGLHfldMmi7DtwikbKcLHJseUYtxMAKMpAn33PyNNOkybGupPskzdEQk98l0HLZF8gtPmZdOLMsGfLrTrMkUFZ+V18ftKKVDOH0W91bafTK1EsdLzedWhAXmSyMyZcbyfHlldJbJJ7F/ukxdkivMb4ZbMb7uv/u5FYLenoJZLesHxLVgMZtwsUwDKbF4olT2dzZsMuf9YOf1Juph2an1qiIQPgIssrPfDK/hrQdmCCNpzuUe+smlC6NqT4n0jv8HOiWZqTKrr6n38saIupOx8q20O/5du/y/D5kMeo7KMm8BLJgXlZyVZn3TpPC4yzXElMW5IedrVYyvT/bLmN3/ucUmbgoNlB3ow6fsNgvYNFKV4BAAM6zqMN2GsFnsHzd3X3CZaGpFIEAEWAfFHROhALAYwwDFW978dKex+Dx7tvTvOdq+y3z1OZE4V63xuMrpu+zWxz8MKyZDp+FbTIAiRjIIO+B37C8WyKfhq6lRIRniM9Ro9a1548bs+Js1B6Puzt77B8Dx3r17JS0tTfLy8tyXs7OzZf/+/eb46NGj5npWVpb7uhO/8CNmBIM3A2ujKd5v2RD33tRstzsPJ+quOsUXAuw3I5Q4m3ppb2UFNtubX+Iz5Lh6YY5MndtmwhwuyVgM/+qY7ebJC0+cKUUubGhq89YmwVojsV96REgHkvn7E6NlRcphs97CJkGmyPB5xHDbLmuoKVOmmLDJ3bt3d9d6lSpVJDEx0Rxv3bpVrrnmGrnhhhvc1532hTdK4twQisB7XxDOCjsO2Sy8Ueq0o9NqLr70wWEsnib4neKdnCia4QixWYjRYvU/Tv+s1PFkwp0usyp38OTdps0sST7srnjcZB85fsYsrK3/MaBVJPwBWdNlno0UFylE46MBs4Hr+bc2ydg5e02QpGjufbn++uulZcuWRpWcnBxDOnXq1DHHZ86ckSuuuEIGDBjgqapjvkMqPUZmSu9R2/zGm2FkiiNQiCiaODoGFFUkoghgnTh+3j559OWN0nfcdtmZ47IQC7cQjYwZLoIito1kWEBnId0ii1A+nx2Q4XZtj505eTzYKUnue2GdfLX6oDzd1zWC2ZdfIstTj4RVFnn7Wvj3hJy3biLwQTJYmGH/DvlAQgzBIyk1a9aURx991GTZu3dv+d3vfie33nqrOV6yZIlceeWVUlhYaI4Z5h8+fNjn9BNTa5CSXcK0GKGficXuPeftrQPkQrho0sebQ1DvZ9FjexA4WHhaRs7INhud2TaAxWekpXXvuZKQOC/s/iSUPi+Ye9ARXZ0mtpEMpsChmjDX75UibGZioT8565gQ7AeZvSpfmDbzrAiIhqmZsfP3lTrvmSbQ7+WZMHtXJruIsb1nOo25YP7e/Xy3mW7zDojkfW95x40aNZIaNWrIoUOHhFHN0KFDzRQZhFK9enUZPny4yYIpNMjniSeekD/84Q+SlJRkzr/33nty0003SePGjc19Y8dG34Ee9v8vDt1qFvoDnQYjHfjhIcEz5kd5+Oj1yoUAMwpYdj760kZ5f1p2yN6+A0EtIytPqjYcG7Hd/4H2P8GkY5c/OqKr08Q2kglnM2bv0a4Fbdy+P94jWf790Tbzlk4nRuQ7qzK4hsUSU2b3dwx/F20wmzF9VSxDdry5Mg3EKIeOc8L8fWYzVqCdrpVvly5d5MYbb5T27dvLoEGDZN68eXLRRRfJqFGj5Pbbb5ezZ12hADp27Og2EKhdu7Z07drVZDFixAhZvny5+d63b1+57bbbrKyj8kmUTp6Xt8tgrfAgTtytYyaOS3YVRcBCgD4A60SmxfAU4StsspU2kp8mREkUA5hZfVionwmJ02MWnqQ8nG0jmXDcykAYxGHw9FaKjbmnOwbWR7A2wzfWP1+MzO5Z3MrMC8GtjC/QmQpioyZD+5YD0s16DlNDc1flBeQQ8p133jHrMDfffLOcPn1aNmzYIBdccIFcdtllxrLMu8yTJ0+aEc+4ceO8L0nTpk3dRgPnXYzACTZXssmS4GUQRqjy8dy9ZsMc/t9UKjcCW3cXCUH72PQ8ccH+iEVeDRTVwqMlUr3JWKne0nkOftEJ3dDRiWIbyUTbQSZrMKfOnJOuH2yVQZ/tkmUpPxkGhPJ2ECkHmf4qHcu0hWsOmpEYljB4H8aMF+eZvkIqT5o0yZDKsmUuNzcs/kMyQ4YMOa8IOnamzZo1a3ZeJ//xxx/L3/72N7Nmc96NETjBAiwjkBFeG9tCzZqRYOPeqRFf4wpVH73PXgTSso6ZmYAGvVKMo1nL9N1eLVylrUnZK3c2GBMxf2ah9Eve9+CnDJ3QzaliG8kAQLRc/Td9Pe08fJky866QYI6j4er/PCU9TmRlFxkXNaxh4D+Jz0kLc4TzTDcVFxe798VwG0SyZ8+e80iEa0yZNWjQwD2FZhWDKTQjodzcXOtURD8hzpb90+WjmXsimi8hqHHDjjsalcqBwNpNhdJxyBbz8sVo3ykWh2OnbpC7Go13ROBFrF/RBZ2cLLaSjAYtC6wp8LaGfyVGNuyIZ/6ZoGaMfMoLZ8zaS926dc2UGmTy7rvvmkKZXvvLX/5iiOrUqVOChVokBSsfdGXNKRqCSxCmSjZ5OBeMRjmaZ+wQ4MVp+cbDZnMuLyvEIQp27dIO7QeOXinVGn8SU6KBYNABXZwutpKMhl8OrTlALLzNvT42yxAOP0DWdpI2H3XvnCfnEydOyK9+9SszjcZUGn8tWrQwhWIEYJ3j89prrw1NGR93EdqAAFA4C42mrMkoNESz4ce9UNEsS/O2DwFM279ec1Ba9E2Tdm9vMoH8wlnLs0NzNmkyiqj5/MqwZkyCmV2x0lImZaNDPIitJAMgxkqjxSzbK8aqoEA+E1rMcqylBlNnvM1/Mn+/sd5iag3rNeLXRGoDWjANF1NSPCHgqNQOSc1yxaQhoqhK/CFw4MAB+eabb8xIm026kxdul3uafiBtB641hjHx9EST5qaa9ZDqz3xtW39GWazBUHa8iO0kk51TaJxO3tP2O9sqJhBisdKgF5WInvEgGAmsTDls9uM0eTXFOPDEPQv7daLt9gZzcfYCzV5ur20+lkYYSyxa64rREQ/1pDqKYHTypz/9yYyon+86WOo8N1Ou+q8/y5VX/VZatWoVlxClbj0g9yaOlxrNp0o0+zTypgzKosx4EttJBnDeGrvKBViIMawtQojGJxWJfvEqxu3Ncg+3N4M2mf0EmHZHcvGU/Qosxn/1ncthqd14UT4WR3YTnN3PWZHKGzFyjLz/eaohmSdavSOz5q+U9PR0adOmjTFWiddnLS45bfoMXGcxC8J6SaT6JvIiT1586ZcoK94kJiQDUDCyncPMQCrd6JM4Pi4r0lfDY66b6SXj9uYtjtrThgAAHsNJREFUl9ubniMzjXdaRiGhClZehDyI9UgCQm36WvTXgkLFSe9zIcCIetTMPSYUece+sw3JrFmzxg0PniisjcKctNv1kVuRML8w+8FywB31R0uNxGmChSpbIQLpezzTcA/3kgd5kWe8zKz4gjAmJIMiDPlg51gsnHlWqPUdPdAnNdOetQVflRHtc7i2wXv0oIk/ub0hmicWPYG6vcnc45qqwtrLCYJV2zP90mT0rMiaTYfzbAWFJ+Sr5duk97AlUr/zNKnR9GO5/alRUqXeCPPJMee5TjrSV0TBIGT4FJfrFz45tvZ7HTniiusybdo0wSgFiYXro2jgjoETlrR123xmSIJN3XgPwU0VgRsZneDhnT++c45rpCEtxMJ2D/Igr3iXmJEMwLF45QSbcyoaPeJpMS0SDW83bm+W5MrLI7YatzcEdcKTbcaOYz5NRzE4wIzYaYvubADFGza+rGJllYRBxjertkuLHrMMkdzdfLJUTZxjdojzAoNvKd5Q+eSYXdpcJx0ExH3cH6wLnki0g0jnsTe3xPifq9d9oxnBeK4N9uvXT6666ipTZGZmplx33XVixUH64IMP3CMaO1wfRfq5feXHJnS8neBWixD0OApmWo3QJ/zxnXNcIw1puaciSUxJBiAxw4ulzTkEQ/nxYg4Yrcb3k9ubPSZELW56cOMxZ1WeMC3FzmsIBjNiJwoGEJ3f3SIDxu8o19tzpPVf9P12ebDlRKnedKJUa7FAarVdE9QUCem5j/vJh/ziUbbtLZI+Y12uX7B+xL2Qt3Tr1s04bp01a5Yxo1+wYIF3EnMcbddHPgvVk1FBIOYkw1Oxochlc74qqB+nNdUV6icuGSg3HjY0RaX2y8iUt0/2LtBp80aK/7heo7aZ8LW+3N6UkZVtl4iGSFA0wi14B0aLhhIHC4qkZe85Uq3xBKneclFE2i75kB/5kn88SPr2YwZ3DEEwpcdxrT/BWzj7tP7+97/L6tWrfSaLtusjn4Xqyagh4AiS4elwjcCaiF0O6CiH8pzukiFqNR9gxus2uzZA4trl869zpMswl9ubzkO3yGcLc4Q1GidN8RDmgRDPXd/bWmZnF+Dj+02WlL7POCWsljgzrLDivl6QmKsnX5weUo5TBYevuD/CdH72yvyAiJ3pzIIC/+t50XZ95FQsK7JejiEZQMbJGz8s3FYzd+3rBxjuOfIlf8pxslM5JzS679OOmCky1mg8Bbc3TJsRx4Nd2lY0woWry3d745lPtL5DeoM/2yXt394UsEFDMLos+i7rxxeiyIxe/LVpRjVVG44RynOKQBKsybEGltgvTb5ee9Dn+l0o+nq7Pnr11VdDyUbvcRgCjiIZsMFdNSZ7BOAh0htvdf5+hMGcJx/yI1/yd6pbbKe0DyzOWIPZkl3+IiRubxjpMB8P4TzbP10+/DJbGAXF0muuFVbBc+E5XHxZL2EEfHebZRFpl+W1YcqhvG9jvE6DDzF8iVG3RIGlfUTayCKaro/CrXe9P3QEHEcy1qMQ4Y1QovzAEprPkLufWx7Sj5r7uJ98yM+JkeOsZ3bKJ94C2AeTtS9401pGEZt3HZdPv9ovHYdsNh6lCaf8xbcHYuJFGT3wq4b5bLhizO4bjpEarZeG1BbLIxR/1ynvzoZjYrLTG4MQjD/Yj9Tp3S2Cd2QVRSAYBBxLMtZDZGUXyJujV0lCk7FSrcl4qZE4Xe5KnC81Wi02cR1Km4auMufN9cTpJj3TYtxPPirlI8CUFwu4kfKDhpEA0yuYF9NRQV4DP9lh3ortino5Y2mucX+TnVtcPgB+UhQWlUjNp8eFHELcH4EEet7ssXh6nKCHHVJy6qyJ31L/lRRj4p62vfSUqR06aBkVAwHHk4wFM0PzzJ0H5ZPZydJ10NdSr/1kQzyem9wgIs5znXSkj/SQ3tLHaZ979+41ETLz8n7yI5adne2OQcMu6rS0NPeeBF/688aKqxb2OURLMIfGFQyWaphJPzcoQ0bP2ivJmZF1e+OtP+QJwWFmG4p0eXOhJDSfbusIxpuAKB89oilFJ86YyJNMlRK5lXhGKopAOAjEDcmE85CV4V6sci688ELp3r27+3GrVKniDrO8detWueaaa+SGG25wX/f88uWyPBOBEhKwS3B7w/4bwiwTrvlfXTdIj5GZwsgjGkS3IsW1zoTJbTCyPmO/VG30sdSKkjGKN5n4O6Z89NiwaX8w6geUllHlmNl7jbn6wAk7ZPeB0Ed9ARWoiSoNAkoyFaiqr7/+emnZsqV5IsIzQzp16tQxx2fOnJErrrhCBgwYcN4Ts17CVFbuIfsI5jwlRIwl2JL1BfL2j25vGvZOlcGf75LlyQURsxIjBs9jL280sXh86eDr3FOdpprNkv46fzvPs2kTfSIl+YdPGSvBR1/aaKY0Y90GIvVcmo9zEFCScU5dhK1JzZo15dFHHzX5EPnyd7/7ndx6663meMmSJXLllVdKYaFr4ZZpRL4T4rn5G2mlIm6eO3fOEdOMeFqetiRXeozINKOc9u+43N4wEgknYiL3Mx2EhZQl4AGheVuiMWq4q/G4kBwdRoN8cE2DPhs2hRcgbm9+ibzz2S6BXLDCKyiMP+++Vt3pp7MRUJJxdv0EpV2jRo2kRo0acujQIWFUM3ToUDNFRgdavXp1Ybc1wmjnH//4h9x8+73yH7+5RpZ9l+QuB4eFhGleunSp+5wTvmDlRETMj2bukdYDM+SRl3B7s03mrsw3bm+C1ZG1GdZoWKtBPpiWbdZbpi4uHauj21tfG9P3aBBGqHliio9eoQgetPuO225Gc/ipKzyh5BIKjnpP4AgoyQSOleNTdunSRW688UZp3769DBo0SObNmycXXXSRjBo1Sm6//XY5e9bl7qNTp07yzvgU46Ps73+/TV5//XXzbB06dJBevXrJL37xC1mxYoWjn5cRB6EGWD8ggBlmynj6/S7tSMA7/bE2e/TlZBNZ1OrwITBLSk6ekTueGh3VYFRWucF8EsAKvdAvUMGsHHc7EOvkRTkBYxRo/ppOEfCHgJKMP2Ti8Pw777xj1mFuvvlmE96WHdT4ibrsssuMZZn1SO9NzTYb6r5etFT+67/+S5KSXCMZyxIPL7lOJxnrWaxP9vTQeeJOhpDUOMucuGC/EEXTn9sbLKnue2HdeRZjvO0jq1P2SI1mE8+7HgwhRCsteqFfebIx86jBhPWtmcvzBP9uKoqAnQgoydiJdpTLsmJ1LFu2zJTE4j8kM2TIEHMMiRA/hrWNJ5+qLxdffLE0btzYBInyVC0eScZTf7wMsGmQKTBcnxi3Nx9vlwXf55daexo22TVF5k0ErFEgQz/53mzk9b7uhGM2GKOfP8ElEPWc2DfNTAliyaeiCMQCASWZWKAepTKLi4vd+2IoAlLZs2eP+eRt/q1Pd5pd25aX3GPHjhmTZqbPPCXeScbzWfiOBRVhovt8vN2Y6BLk7NUxWXJv+/NHMRAIU0oYFrTsNcdx0VstgiOKK/p5CnVMULpWA9Kl9ZsZsmxjgd9RnOd9+l0RiCYCSjLRRNchedNh9h/n8kzMTm4sz6z1mX/961/SunXrUpr+5je/cdzCfykFwzigI97yo9ub2j6myqxOnJHQ/c9+Ine3Cc2dUZu3Nsmkbw7IlMW58uHMvSbktZV3JD7RC/0QPE9Dos3eSJUXBm+W1RmuqJNhwKS3KgIRQ0BJJmJQOjMjOiB2bhP90pqPJ6b6fffdJ4888ohUrVpVdu7caZT/8ssvpW7dumaK7ZZbbjFGAM58qvC1Iu6Jv87+vg5JxhiAqIUEtfOXzt/52avyjYKZe05I+g6Xg9HcglNB5+Mvf86j150NxxryatQrRfAPl7wtuE2m4aOoOSgC5SOgJFM+RnGbArNfLIr447unnDhxQoqKKq/LkBcGbzm/02+/zhgNNHnd5a+L8LjBegF/fdwOA/Pi9QUm/3aDt5jjb388hiCa9U2T7h9mSvshm8/XoYPvKTxvwjExZ56eLb1HbQvIU7Zn3et3RcBOBJRk7ETbxrIYtXT/IFNeH5tlezhiGx8z5KIIvubdcVvHbNTck1diYrAHSzJp248bvBv2TjH5D5q0y+g4bOpuc7wi1WVivSq90BD/kuTDfvWw9PH1iV6QoIoi4HQElGScXkMh6Me6C6a8/Yh3f06tinxBiFEEHoa9O3CcdlohDoKdLmONB6MKPBVY+UI6SKs3N5lzo2bvlZ6jssz3ed8flGMnzrrTWvcE8mmmyxqM9vVoek4RcBQCSjKOqo7wlcG1fschW2TQxJ1qWeQDThyATpi/z/hqq9d9Y6kO/p8vrhdP55nBLvwzekHWbz1q8u3x0TZzDPHc3zFJHuiU5C6vTuf1wmbQDZmutIEQi2caz4V/H4+ppxQBxyCgJOOYqghfETYXEnKY2C3Wxsrwc43/HOjkv15z0IzucI4JPuyAxxmkZcYMAXgH5ArWhJmRTFHxWSk8fkbGfbXffAe9I8fPyPcZhbIsxTU1VrvjOlmeekTyjpwSa1rNk0AC+e7LhDn+a0qfoCIioCRTQWoVH1TEZmEDooprjxDWVuwNYgqMEALsG/E2gHhx6Faz63/JxvOD2oWyGRNyoYyjRWdk+LRsWZTkynf7vmJp/FqqGcksXHtI8gtPSdPX09wjm0CIxTNNeZsxtQ0oAk5BQEnGKTURhh7EAmk5IF1GzSzfzUgYxcTFrSzYj5u7z0TCBBNMlcuKwLlwzUGZu+qnQG+eD/n9xtDcykBqTI9ZpPB4j2T3d4gHwSCgy/tbZcf+4lLTaNY95X0G6lbG83n0uyIQCwSUZGKBegTLxEU7rlPGztkbwVzjKyumCed9ly8dBm82zjLfn5Ytlv+x8p6krGnFaDjIZATjKbh7KY9QvK+H4iDTs0z9rgjYiYCSjJ1oR7gs3KWwy/uzheHFFomwWrZkh9UcAciwoGPkgNv/VamHI26u7URX/9US54bs6t+WytFCFAEPBJRkPMCIp68sWjd5NcVMB8WT3uHqujunWEbP2iP1e6YYT9KEjT5WFLjL+2DLJzhYRQxaFiwOml4RCBUBJZlQkYvhffvzS8S4bl/mey0hhqpFpeijx0/L7OV50u7tTYZcRs7YY2sM+vqdpzkq/DL6qCgC8YKAkky81NSPerK3gk2E/har4+xx/KrLWsXq9CPGYwHTYX0/3m5MjP3FhvGbUQQurM/YL1UbfSy12q0Oev3Eez0lnGPKRw9CQqsoAvGCgJJMvNSUiFnMZpoIi6iKKizYj/gyW3DtwkI+4ZWLSqI3HRYojl3eXCgJzafHlGQoHz1UFIF4QkBJJk5qi5j0hBkmXkhFE0IpT1uSa+KgNO6daizlmBJ0khQWlUjNp8dJwrPfxIRoKJfy0UNFEYgnBJRk4qC2iH/Cm/2KlMNxoG1gKrJhkefBQzTTYQMn7JANW4862lNB6tYDcmfDMVKj9VJbiYbyKJfyVRSBeENAScbhNYYvLQiG9YmKIFt3F8nwKbtNSOTOQ7eYkMhWpM54eL5F32+XOxuMkbvbLLOFaCiH8r79fns8wKM6KgLnIaAkcx4kzjmBWxR8bbEfJJ4l/8gpmbwoR1r0TZOn+6QaB5V5BaU3JcbT8y36LkuqNhwj1VsuiirRkD8EQ3kqikC8IqAk49Cag1ggmOTM2Ec7zMjIkLS0NDl58qQbLY6toGd79uwx1/PzXREhSXTy9FmzfkREzkdf2ijvfLZL0rJi/yzuBwjzS1L6PqneZKxUS5wZdGCz8izMTECyxJkmf8pRUQTiGQElGQfWHjHamSJL2x77Thm3Kw0bNjQhmdetW2fQ2rp1qzmeO3euOR4zZow57tGjh2TsOCaDP99liIWQwN+uO+QO++xAqMNS6WBBkbTsPUeqNZ4QsVENoxfyI1/yV1EE4h0BJRmH1eDKlMOGYHBF7xRJSkoyJLJgwQKj0nvvvWeOIRdk8tS5ctFFF0m9F+YYP2qTFuYIU2SVRVinebDVRKnedKLZtFmr7ZqgptFIX63FAnM/+ZCfiiJQURBQknFQTS5ZX2DMlLOynfUGm5uba0hl4sSJxvrrpptukquvvlpatO1pYrRc9/f68o/q9UyMFuAsLi6WkpLzTW3PnDkjR4/G9/qSv+bCJtFvVm2XxH/PlNufGiV3N58sVRPnSPWW30rN51fKPe1WS612a80nx5znOulI36LHLHN/LDab+nsmPa8IRAIBJZlIoBiBPAiqxUbLQL0HR6DIgLM4d+6cXHzxxTJs2DB5c+g4uTWhnvznddXl9nuby2ezkuTyyy+XnTt3ysqVK+WWW26Rxx57TP74xz9Ku3btTBmQ1D333CO1atWSGjVqyB133CGFhYUBlx9vCQsKT8hXy7dJ72FLBBcwNZp+bIikSr0R5pNjznOddKRXUQQqKgJKMg6oWdzU4ypm94FiB2hzvgp780vkit/+Tv73ruby66v+KEPHrZAmzZpL/fr1pU6dOjJgwABz07Jly+Sjjz4y31mvYQrt7NmzkpOTI127djWjoFOnTskvf/lLmTNnzvkF6RlFQBGocAgoycS4SmcuzzPOLunInSRWjJaOQ7aYNaI//vlWM5rp2bOnUZNF/ksuuURuv/12YRrMU44fPy7NmjWTxo0be5423zEa4L5t27add01PKAKKQMVDIG5IhrnqTdvzZfz0jdJ54AKp126yJDQZW2oagmPOc510pHfyHPfUxQeMu/6cQz+ZBseyiYEVptP9x20/L0YLU2DXXXedWW9Bx+HDh8ull14qkIanfPvtt3LttdfKVVddJV999ZXnJTly5IiZThs1alSp83qgCCgCFRcBx5NMRlae9Hl/mdzVcIwkNJ0gCYnTJSHxK7m71RKp+fwqrwXVVeY81026phPMfdxPPk4SLLDYmEjgsVjLrgPnx2gpPHG6lFoFBQVy+PBPbm3YI+O5L6ZUYhEZMWKE/PznPxfWYxDSJyQkGHLyTqvHioAiUHERcCzJrEnZK41f+tK4NmfDW83nVgZlFmpteOM+7sdFOvmRb6xl/Lx9Zvf7wcLSHbmdekEinjFaPpq5RyCbcGTx4sWyaNEikwV7ai644ALZu9eFd4MGDaRfv35mXWbp0qVimUOHU57eqwgoAs5HwHEkcyD/uLTtM89EI6zW4itj9mkRRjifmI+SH1EOyZ9yYiFEdWw5IF2OHCu9jmGHLlaMltfGZJnpsEjHaJkwYYL8/ve/l2eeeUYwc37zzTfNY61YscIQDqRj/Y0fP96OR9YyFAFFIMYIOIpkFizf5vIJ1XxWxF11WASFy47qzWeZcijPTvlg+h5p82aGeE9FRVuH7fvsi9GCuTPTangKUFEEFAFFwBEkw4JznxGsu4yTmq3t8W5LOZRHudE2DqDDHfrFbhM++HixPSMYzxgthGoeO2evOM2CTX9+ioAiUPERiDnJnD59Vtq/MV8Smn5mFvGtEYcdn+zCplzKR49oCAT29sSd0nHIZjkR5QiP8RijJRqYa56KgCLgHARiSjJ0wHTw1ZtNjtr0WHlkZabPmk02ekR6RHP23A8yYPwO6TJsq0QzZsqW7ONxHaPFOT8H1UQRUAQijUBMSYapKkYSdPTlkUE0r1M+eqBPpIRF9j5js6T7B5nG7X2k8rXysWK0JPZLk2avpcon8/dLrkP221g66qcioAgoAjEjGXw2sSbClFU0CSTQvNEDfSJhDMC0Va9R26TnyEw5eeZcxFrZqdPnTIwWiKsixmiJGFCakSKgCDgGgZiQDObDRBa0a5E/UKJBH/QKx7wZIugxIlNeHZ0lZ85GhmAIwVxZYrQ45pehiigCikBEEIgJybBPBTPiQDt/O9OhF/qFIiWnzhrX933HbRemy8IRwhN/+tV+MxXGlNikbypXjJZwsNN7FQFFwDkI2E4y7LhnQ2Ss12H8ERd6oV+wngFY2O/87hZ589OdIZtEkwcu/zEUePTljcbs2UnBy5zTbFUTRUARiBcEbCeZxt2mm533/jp5J5zHMwB6BipFJWekw+DNZkor2E2IpE/OPGbI6eFuG6THyExZtrFAWNdRUQQUAUUg3hGwlWRwUokPMVy8OIFM/OmAfugZiFPNY0Vn5Pm3NsnwqdlB7XLfn18i4+buM27+cTODR2Y2UKooAoqAIlCRELCVZPCGjLNKf527k86jJ/qWJYXHzkjrgRkyckZ2Wcnc1xjxEKCMjZmP/ztZPpiWLVn7NCqiGyD9oggoAhUOAdtIho2OuOsP1Zuy3QSEnujrb4Mmo45n+qUZdy1ltQruX7e50GeMlrLu02uKgCKgCFQEBGwjGQKIEQ/GbrIIpzz0RW9vYSNk8zfSjPWX9zXr2DtGy8xlebY7xrR00U9FQBFQBGKFgG0kQ6RKAomF0+nbfS/6orenYFrMDvvJi3I8T5vvVoyWtm9tkvo9UyQSMVrOK0RPKAKKgCIQRwjYRjKERCZipd1EEU556IveluQcPCmNe6fK9KWuaI+cZz/M92lHxDtGC37LVBQBRUARqOwI2EYy9dpNNqGRw+n07b6XEM/ojezNLZEGvVJk9krX9Jl3jJY5q/Kk6IQ9bvwre6PV51cEFIH4QcA2kkloPEZqPr8qrJHMSyMyZcbyfHlldJbJJ7F/ukxdkisDJ+4MK19/5IW+CU3Gyu6cYjP99eWSXJm2JNdYlGmMlvhp5KqpIqAIxA4B20jmtic/klphOMPsOSrLTE2xSbGo5Kw065smhcddI4cpi3OjQjLoe0fDT6Vetw3y3KBNJmTxwAk7ZMPWo0HtiYld9WrJioAioAjEFgHbSKZKvRFhbcLsNHyL9Pl4uxnJABl+wnbsLxbIp+GrqVEimbVSo9ViebZ/uixcfTCqMWFi2wy0dEVAEVAEooOAbSTDSCYSbv1fHbPdIFF44kwpcqn9wjpp89Ym+feH2+TpvukRIR30vf2pUdFBXnNVBBQBRaASIGAbybC2Ee6aDGsngyfvNtWyJPlwKSI5cvyM5BackvQdx831tz/fXeq6v3WXss5bazKVoB3oIyoCioAiEBUEbCOZSFiXPTsgQ06UuNZhIBQI4sFOSeZz4dpDZp2Gc8eLz8qCNYfCJhlP67KooK+ZKgKKgCJQwRGwjWTC2SdTv1eKbN5dZBb6k7OOyaKkAlMts1flC9Nm1mikeb90+XzRAePBuPuHme7z1vVgP733yVTwtqCPpwgoAopAxBGwjWTC2fHfe3SWeXBctTzeI1n+/dE2Y91F/JV+43e4yQRDAHyFQTzPvbXZfT5YcrHS+9rxH/Ea0AwVAUVAEajACNhGMuH4Lru/Y5I82TNZWNy3CKDeyxvlwc6uqTLrHJ+QUFHxWdm47ag7ref1YL77811WgduDPpoioAgoAhFFwDaSiaYXZoJ9Zew8Lo1eTRUIKe/IKVmzuTAskinPC3NEa0EzUwQUAUWggiJgG8mAX7TiyUAyGAKw4M8oJju3WFoN2hQWyQQST6aCtgl9LEVAEVAEIoaArSQT7ciYdbusl4e6bgiLXJhOCyYyZsRqQjNSBBQBRaACImAryYBf427TpVoLZ3tjRj/0VFEEFAFFQBEIDwHbSWZNyl65q/E4qdV+TdgjjmAW8QNNi17oh54qioAioAgoAuEhYDvJoG7bPvOkevNZjiQZ9EI/FUVAEVAEFIHwEYgJyRzIPy5VG46Rmq2XOYpo0Ae90E9FEVAEFAFFIHwEYkIyqP3V8m1yV8NxEXGaGehUWFnpcIaJPguWbwsfVc1BEVAEFAFFwCAQM5Kh9D4jlklC089ivj7DOgx6oI+KIqAIKAKKQOQQiCnJsEGz/RvzpXqzyTEjGgiG8tEDfVQUAUVAEVAEIodATEmGxzh9+qzp4BlJRCLeTFlTYt7XKI9yIRj0UFEEFAFFQBGILAIxJxkehxEEU1WsidhlDEA5lEe5OoKJbKPS3BQBRUARsBBwBMlYymAMgHUXZsTR2kdjpseazzLl6CK/hbx+KgKKgCIQHQQcRTI8IubD7FNhQyQ773Hx4j3NFcox+ZAf+ZK/milHp0FproqAIqAIeCLgOJKxlGPHPa5dqjb6WHBWiVfkUMiF+7iffMhPd/JbCOunIqAIKALRR8CxJGM9Ok418d58V8MxQnwXAokRsZLQyDWfXyW12q02ox0+OeY81026phPMfdxPPiqKgCKgCCgC9iLgeJKx4GBxnsBnRNgklHO9dpMloclYue3Jj6RKvRHmk2POc510pNdFfQtB/VQEFAFFwH4E4oZk7IdGS1QEFAFFQBEIFwElmXAR1PsVAUVAEVAE/CKgJOMXGr2gCCgCioAiEC4CSjLhIqj3KwKKgCKgCPhFQEnGLzR6QRFQBBQBRSBcBJRkwkVQ71cEFAFFQBHwi4CSjF9o9IIioAgoAopAuAgoyYSLoN6vCCgCioAi4BcBJRm/0OgFRUARUAQUgXARUJIJF0G9XxFQBBQBRcAvAkoyfqHRC4qAIqAIKALhIqAkEy6Cer8ioAgoAoqAXwSUZPxCoxcUAUVAEVAEwkVASSZcBPV+RUARUAQUAb8I/H8W1zhfIq2i4AAAAABJRU5ErkJggg=="
    }
   },
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### [Learning objective 8] Build more hidden layers\n",
    "* Add one hidden layer\n",
    "\n",
    "![image.png](attachment:bb80ccac-f394-4386-938a-69e6f479494a.png)\n",
    "\n",
    "* There is an activation function (sigmod)\n",
    "\n",
    "![image.png](attachment:192210c0-87eb-4236-8ecf-9ea58b870c2a.png)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['loss', 'mean_squared_error'] [0.10282432287931442, 0.10282432287931442]\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Activation \n",
    "#from keras.utils.visualize_util import plot\n",
    "dl_mod = Sequential()\n",
    "dl_mod.add(Dense(4, input_dim=6, activation='sigmoid'))\n",
    "dl_mod.add(Dense(1, activation='sigmoid'))\n",
    "dl_mod.compile(optimizer='rmsprop', loss='mse', metrics=['mean_squared_error'])\n",
    "dl_mod.fit(x_train, y_train, epochs=1000, verbose=0)\n",
    "score = dl_mod.evaluate(x_test,y_test, verbose=0)\n",
    "print( dl_mod.metrics_names, score)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['loss', 'mean_squared_error'] [0.11216876655817032, 0.11216876655817032]\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Activation \n",
    "dl_mod = Sequential()\n",
    "dl_mod.add(Dense(4, input_dim=6, activation='sigmoid')) \n",
    "dl_mod.add(Dense(2, activation='sigmoid'))\n",
    "dl_mod.add(Dense(1, activation='sigmoid'))\n",
    "dl_mod.compile(optimizer='rmsprop', loss='mse', metrics=['mean_squared_error'])\n",
    "dl_mod.fit(x_train, y_train, epochs=600, verbose=0)\n",
    "score = dl_mod.evaluate(x_test,y_test, verbose=0)\n",
    "print( dl_mod.metrics_names, score)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### [Learning objective 9] Standardize data for better modeling performance\n",
    "* For a classification model, you are advised to standardize the predictors."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['loss', 'mean_squared_error'] [0.09813830256462097, 0.09813830256462097]\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Activation \n",
    "dl_mod = Sequential()\n",
    "dl_mod.add(Dense(4, input_dim=6, activation='sigmoid')) \n",
    "dl_mod.add(Dense(2, activation='sigmoid'))\n",
    "dl_mod.add(Dense(1, activation='sigmoid'))\n",
    "dl_mod.compile(optimizer='rmsprop', loss='mse', metrics=['mean_squared_error'])\n",
    "dl_mod.fit(x_train_scaled, y_train, epochs=600, verbose=0)\n",
    "score = dl_mod.evaluate(x_test_scaled,y_test, verbose=0)\n",
    "print( dl_mod.metrics_names, score)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### [Learning objective 10] L1 and L2 Regularization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['loss', 'mean_squared_error'] [0.09434328973293304, 0.09434328973293304]\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Activation \n",
    "from keras import regularizers\n",
    "dl_mod = Sequential()\n",
    "dl_mod.add(Dense(4, input_dim=6, activation='sigmoid')) \n",
    "dl_mod.add(Dense(2, activation='sigmoid'))\n",
    "dl_mod.add(Dense(1, activation='sigmoid'))\n",
    "kernel_regularizer=regularizers.l1(0.01)\n",
    "dl_mod.compile(optimizer='rmsprop', loss='mse', metrics=['mean_squared_error'])\n",
    "dl_mod.fit(x_train_scaled, y_train, epochs=600, verbose=0)\n",
    "score = dl_mod.evaluate(x_test_scaled,y_test, verbose=0)\n",
    "print( dl_mod.metrics_names, score)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['loss', 'mean_squared_error'] [0.09841769188642502, 0.09841769188642502]\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Activation \n",
    "from keras import regularizers\n",
    "dl_mod = Sequential()\n",
    "dl_mod.add(Dense(4, input_dim=6, activation='sigmoid')) \n",
    "dl_mod.add(Dense(2, activation='sigmoid'))\n",
    "dl_mod.add(Dense(1, activation='sigmoid'))\n",
    "kernel_regularizer=regularizers.l2(0.01)\n",
    "dl_mod.compile(optimizer='rmsprop', loss='mse', metrics=['mean_squared_error'])\n",
    "dl_mod.fit(x_train_scaled, y_train, epochs=600, verbose=0)\n",
    "score = dl_mod.evaluate(x_test_scaled,y_test, verbose=0)\n",
    "print( dl_mod.metrics_names, score)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### [Learning objective 11] Dropout Regularization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['loss', 'mean_squared_error'] [0.0974554643034935, 0.0974554643034935]\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Activation , Dropout\n",
    "from keras import regularizers\n",
    "dl_mod = Sequential()\n",
    "dl_mod.add(Dense(4, input_dim=6, activation='sigmoid')) \n",
    "dl_mod.add(Dropout(0.25))\n",
    "dl_mod.add(Dense(2, activation='sigmoid'))\n",
    "dl_mod.add(Dropout(0.25))\n",
    "dl_mod.add(Dense(1, activation='sigmoid'))\n",
    "dl_mod.compile(optimizer='rmsprop', loss='mse', metrics=['mean_squared_error'])\n",
    "dl_mod.fit(x_train_scaled, y_train, epochs=600, verbose=0)\n",
    "score = dl_mod.evaluate(x_test_scaled,y_test, verbose=0)\n",
    "print( dl_mod.metrics_names, score)"
   ]
  }
 ],
 "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": "3.8.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}