17_TensorFlow_Keras_LSTM_Tendencias_Politicas_Twitter.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 17 - TensorFlow - Keras: Clasificación de tendencias politicas en tweets con LSTM (Long Short-Term Memory)\n",
    "\n",
    "* En este notebook vamos a clasificar una serie de tweets etiquetados por un partido político (psoe, pp, vox, podemos y ciudadanos) que corresponden a la tendencia politica de la cuenta que publicó ese tweet,\n",
    "\n",
    "\n",
    "* El ejercicio que vamos a realizar a continuación es muy similar al realizado en el notebook *13_PoC_Tendencias_Politicas_Twitter_Generacion_Exportacion_Modelos.ipynb* y *16_TensorFlow_Keras_MLP_Tendencias_Politicas_Twitter.ipynb* y se pueden comparar los resultados que vamos a obtener con este tipo de ***Redes Neuronales Recurrentes, las LSTM***\n",
    "\n",
    "\n",
    "* Los pasos que vamos a dar son los siguientes:\n",
    "\n",
    "    1. Cargar los datos\n",
    "    2. Normalizar los tweets (igual que en el notebooks anteriormente descritos)\n",
    "    3. Particionado de datos (Train y Test)\n",
    "    4. One-Hot Encoding y codificación del target\n",
    "    5. Creacción de las secuencias de palabras\n",
    "    5. Definición de la Red Neuronal\n",
    "    6. Entrenamiento de la Red Neuronal\n",
    "    7. Evaluación del modelo\n",
    "    \n",
    "    \n",
    "<hr>\n",
    "\n",
    "\n",
    "\n",
    "## Carga de Datos\n",
    "\n",
    "\n",
    "* El fichero que contiene los tweets lo podemos leer como un '*csv*' con pandas pasandole como separador '***::::***'.\n",
    "\n",
    "\n",
    "* Este fichero esta estructurado de la siguiente manera\n",
    "    - **Cuenta**: Cuenta de twitter\n",
    "    - **Partido**: Partido político al que pertenece (ciudadanos, podemos, pp, psoe)\n",
    "    - **Timestamp**: Instante en el que se publicó el tweet\n",
    "    - **Tweet**: Tweet.\n",
    "    \n",
    "    \n",
    "* Leemos los datos y filtramos los tweets; quedamos con los tweets a partir de una fecha dada."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Número de Tweets Cargados: 175309\n",
      "Número de Tweets a procesar: 54296\n"
     ]
    },
    {
     "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>cuenta</th>\n",
       "      <th>partido</th>\n",
       "      <th>timestamp</th>\n",
       "      <th>tweet</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>156636</th>\n",
       "      <td>vox_es</td>\n",
       "      <td>vox</td>\n",
       "      <td>1,642,346,589.00</td>\n",
       "      <td>📸 Sobre cuatro principios, como las cuatro col...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>145350</th>\n",
       "      <td>ALevySoler</td>\n",
       "      <td>pp</td>\n",
       "      <td>1,638,088,782.00</td>\n",
       "      <td>Madrid se está convirtiendo en un destino impr...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>161183</th>\n",
       "      <td>PSOE</td>\n",
       "      <td>psoe</td>\n",
       "      <td>1,643,994,682.00</td>\n",
       "      <td>La buena política es la que desarrolla el Gob....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>132070</th>\n",
       "      <td>Yolanda_Diaz_</td>\n",
       "      <td>podemos</td>\n",
       "      <td>1,633,843,724.00</td>\n",
       "      <td>Despedir a través de inteligencia artificial n...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>168722</th>\n",
       "      <td>SaraGimnez</td>\n",
       "      <td>ciudadanos</td>\n",
       "      <td>1,646,215,820.00</td>\n",
       "      <td>Es tan lamentable ver cómo ante el sufrimiento...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>171515</th>\n",
       "      <td>populares</td>\n",
       "      <td>pp</td>\n",
       "      <td>1,647,441,180.00</td>\n",
       "      <td>🎙️ A partir de las 17:00h, @FeijooGalicia part...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>173486</th>\n",
       "      <td>MazelLilith</td>\n",
       "      <td>podemos</td>\n",
       "      <td>1,648,063,213.00</td>\n",
       "      <td>Estos días he tenido la ocasión de hablar con ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>151294</th>\n",
       "      <td>Yolanda_Diaz_</td>\n",
       "      <td>podemos</td>\n",
       "      <td>1,640,170,238.00</td>\n",
       "      <td>@Claudia__Delso @subirats9 Amore! Vémonos este...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>172593</th>\n",
       "      <td>ALevySoler</td>\n",
       "      <td>pp</td>\n",
       "      <td>1,647,952,807.00</td>\n",
       "      <td>@C_Cervera_M A veces se me olvida que en twitt...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>132991</th>\n",
       "      <td>J_Sanchez_Serna</td>\n",
       "      <td>podemos</td>\n",
       "      <td>1,634,106,124.00</td>\n",
       "      <td>Hoy Teo venía preocupado al Congreso por el in...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 cuenta     partido        timestamp  \\\n",
       "156636           vox_es         vox 1,642,346,589.00   \n",
       "145350       ALevySoler          pp 1,638,088,782.00   \n",
       "161183             PSOE        psoe 1,643,994,682.00   \n",
       "132070    Yolanda_Diaz_     podemos 1,633,843,724.00   \n",
       "168722       SaraGimnez  ciudadanos 1,646,215,820.00   \n",
       "171515        populares          pp 1,647,441,180.00   \n",
       "173486      MazelLilith     podemos 1,648,063,213.00   \n",
       "151294    Yolanda_Diaz_     podemos 1,640,170,238.00   \n",
       "172593       ALevySoler          pp 1,647,952,807.00   \n",
       "132991  J_Sanchez_Serna     podemos 1,634,106,124.00   \n",
       "\n",
       "                                                    tweet  \n",
       "156636  📸 Sobre cuatro principios, como las cuatro col...  \n",
       "145350  Madrid se está convirtiendo en un destino impr...  \n",
       "161183  La buena política es la que desarrolla el Gob....  \n",
       "132070  Despedir a través de inteligencia artificial n...  \n",
       "168722  Es tan lamentable ver cómo ante el sufrimiento...  \n",
       "171515  🎙️ A partir de las 17:00h, @FeijooGalicia part...  \n",
       "173486  Estos días he tenido la ocasión de hablar con ...  \n",
       "151294  @Claudia__Delso @subirats9 Amore! Vémonos este...  \n",
       "172593  @C_Cervera_M A veces se me olvida que en twitt...  \n",
       "132991  Hoy Teo venía preocupado al Congreso por el in...  "
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import datetime\n",
    "import numpy as np\n",
    "import time\n",
    "import pandas as pd\n",
    "pd.options.display.float_format = '{:,.2f}'.format\n",
    "\n",
    "# Leemos el fichero y lo pasamos a un DataFrame\n",
    "TWEETS_FILE = './data/tweets_politica.csv'\n",
    "df = pd.read_csv(TWEETS_FILE, sep='::::', engine='python')\n",
    "\n",
    "# Eliminamos los tweets que tengan algún valor a nulo\n",
    "df = df.dropna()\n",
    "print('Número de Tweets Cargados: {num}'.format(num=df.shape[0]))\n",
    "\n",
    "# Los ordenamos por fecha\n",
    "df = df.sort_values(by='timestamp')\n",
    "\n",
    "# Filtramos los tweets a partir de una fecha\n",
    "DATE = \"01/09/2021\"\n",
    "timestamp = time.mktime(datetime.datetime.strptime(DATE, \"%d/%m/%Y\").timetuple())\n",
    "df = df[df.timestamp >= timestamp]\n",
    "\n",
    "# Mostramos un sample y el número de tweets a procesar\n",
    "print('Número de Tweets a procesar: {num}'.format(num=df.shape[0]))\n",
    "df.sample(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Tweets publicados por partido político"
   ]
  },
  {
   "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>partido</th>\n",
       "      <th>tweet</th>\n",
       "      <th>PCT</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>psoe</td>\n",
       "      <td>14125</td>\n",
       "      <td>26.01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>vox</td>\n",
       "      <td>12595</td>\n",
       "      <td>23.20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>podemos</td>\n",
       "      <td>10289</td>\n",
       "      <td>18.95</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>pp</td>\n",
       "      <td>9503</td>\n",
       "      <td>17.50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>ciudadanos</td>\n",
       "      <td>7784</td>\n",
       "      <td>14.34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      partido  tweet   PCT\n",
       "3        psoe  14125 26.01\n",
       "4         vox  12595 23.20\n",
       "1     podemos  10289 18.95\n",
       "2          pp   9503 17.50\n",
       "0  ciudadanos   7784 14.34"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_count = df.groupby('partido')['tweet'].count().reset_index().sort_values(by='tweet', ascending=False)\n",
    "df_count['PCT'] = (df_count['tweet'] / df_count['tweet'].sum())*100\n",
    "df_count"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Nos quedamos solo con el nombre del partido y el tweet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Número de Tweets Cargados: 54296\n"
     ]
    }
   ],
   "source": [
    "tweets = [list(x) for x in df[['tweet', 'partido']].values]\n",
    "print('Número de Tweets Cargados: {num}'.format(num=len(tweets)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<hr>\n",
    "\n",
    "\n",
    "## Normalización\n",
    "\n",
    "\n",
    "* Utilizamos ***spaCy*** para la tokenización y normalización.\n",
    "\n",
    "\n",
    "* Tras realizar un análisis del contenido de los tweets pasamos a realizar las siguientes acciones para ***normalizar*** los tweets:\n",
    "    1. Pasamos las frases a minúsculas.\n",
    "    2. Sustituimos los puntos por espacios ya que hay muchas palabras unidas por un punto\n",
    "    3. Quitamos la almuhadilla de los hashtags para considerarlos como palabras.\n",
    "    4. Eliminamos los signos de puntuación.\n",
    "    5. Eliminamos los enlaces(http) y las menciones (@)\n",
    "    6. Pasamos la palabra a su lema\n",
    "\n",
    "\n",
    "* Todos estos pasos los vamos a realizar en una misma función.\n",
    "\n",
    "* Descartaremos aquellos tweets que tras ser normalizados no tengan por lo menos 3 palabras.\n",
    "\n",
    "\n",
    "* ***NOTA***: Cuando trabajamos con Redes Neuronales Recurrentes y en especial con las LSTM aplicado al Procesamiento del Lenguaje Natural, no solo nos interesa saber las palabras que componen un tweet (o documento), si no tambien la secuencia de esas palabras dentro de la frase, por ese motivo ***vamos a tratar los tweets como secuencias de palabras y no como bolsas de palabras*** (como haciamos en los ejemplos anteriores). Por tanto no vamos a eliminar las palabras con menos de 3 caracteres ni las stop-words.\n",
    "\n",
    "\n",
    "#### CUIDADO - IMPORTANTE:\n",
    "\n",
    "* Dado que los procesos de normalización de textos son muy pesados y tardan mucho, se ha implementado despues de la normalización de los tweets, un proceso de guardado de los tweets ya normalizados. Por tanto:\n",
    "    - Si es la primera vez que se ejecuta este notebook, se puede ejecutar completo sabiendo que se guardarán en un fichero binario los tweets normalizados. Este guardado se realiza [AQUI](#Escritura).\n",
    "    - En caso de haberse ejecutado el proceso de normalización de tweets y haberse guardado este en un fichero binario, no será necesario ejecutar las dos siguientes celdas de código y bastaría con ejecutar la celda de código que lee el fichero binario con los tweets normalizados. Esto se hace en la siguiente [CELDA](#Lectura)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████| 54296/54296 [06:07<00:00, 147.74it/s]\n"
     ]
    }
   ],
   "source": [
    "import spacy\n",
    "\n",
    "from tqdm import tqdm\n",
    "\n",
    "nlp = spacy.load('es_core_news_sm')\n",
    "\n",
    "def normalize(tweets, min_words=5):\n",
    "    \"\"\"\n",
    "    Función que dada una lista de tweets ([tweet, clase]), normaliza los tweets y devuelve una lista\n",
    "    con los tweets normalizados, descartando aquellos tweets que tras la normalización tengan menos de\n",
    "    \"min_words\" palabras en el tweet.\n",
    "    \n",
    "    :param tweets:       Lista de Tweets ([tweet, clase]) con el tweet y la clase a la que pertenece\n",
    "    :param min_words:    Número minimo de palabras que tiene que tener un tweet tras la normalización\n",
    "    :return:             Lista de Tweets ([tweet, clase]) normalizados\n",
    "    \"\"\"\n",
    "    tweets_list = []\n",
    "    for tweet in tqdm(tweets):\n",
    "        # Tokenizamos el tweets realizando los puntos 1,2 y 3.\n",
    "        tw = nlp(tweet[0].lower().replace('.', ' ').replace('#', ' ').strip())\n",
    "        \n",
    "        # Normalizamos Puntos 4,5y 6\n",
    "        tw = ([word.lemma_ for word in tw if (not word.is_punct)\n",
    "               and (not word.text.startswith('@'))\n",
    "               and (not word.text.startswith('http'))\n",
    "               and (not ':' in word.text)])\n",
    "        \n",
    "        # Eliminamos los tweets que tras la normalización tengan menos de \"min_words\" palabras\n",
    "        if len(tw) >= min_words:\n",
    "            tweets_list.append([' '.join(tw), tweet[1]])\n",
    "    return tweets_list\n",
    "\n",
    "# Normalizamos las frases\n",
    "X_norm = normalize(tweets)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### <a name=\"Escritura\">Guardado de los tweets normalizados en un fichero binario</a>\n",
    "\n",
    "* Se guarda una lista de listas, donde en cada una de las listas se tiene:\n",
    "    - [0]: El Tweet normalizado\n",
    "    - [1]: La clase a la que pertenece el Tweet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle\n",
    "\n",
    "filename = './models/17_normalized_tweets_LSTM.pickle'\n",
    "save_list = open(filename,\"wb\")\n",
    "pickle.dump(X_norm, save_list)\n",
    "save_list.close()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### <a name=\"Lectura\">Lectura de los tweets normalizados de un fichero binario</a>\n",
    "\n",
    "* Lectura de una lista con la siguiente estructura:\n",
    "    - [0]: El Tweet normalizado\n",
    "    - [1]: La clase a la que pertenece el Tweet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle\n",
    "\n",
    "filename = './models/17_normalized_tweets_LSTM.pickle'\n",
    "X_norm = pickle.load(open(filename, 'rb'))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<hr>\n",
    "\n",
    "\n",
    "## Particionado de Datos (Train y Test)\n",
    "\n",
    "* Dado que las temáticas de los tweets varian en función del tiempo (tienen caracter temporal), vamos a particionar los datos temporalmente.\n",
    "\n",
    "\n",
    "* Seleccionaremos como tweets de entrenamiento aquellos que se han escrito hasta una determinada fecha y como tweets de test los últimos tweets escritos.\n",
    "\n",
    "\n",
    "* Seleccionaremos un 2% de tweets de test, siendo estos los últimos tweets registrados en el dataset (los tweets están escritos de manera cronológica).\n",
    "\n",
    "\n",
    "* Por tanto nos vamos a quedar con:\n",
    "    - 98% de datos de entrenamiento\n",
    "    - 2% de datos de test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Corte en el tweet número 50762 de los 51797 tweets del Dataset.\n",
      "Tweets de entrenamiento: 50762\n",
      "Tweets de Test: 1035\n"
     ]
    }
   ],
   "source": [
    "# Divido los datos en dos listas \n",
    "#     X: los tweets\n",
    "#     y: target (polaridad)\n",
    "\n",
    "X = [doc[0] for doc in X_norm]\n",
    "y = np.array([doc[1] for doc in X_norm])\n",
    "\n",
    "# Particionamos los datos en Entrenamiento y Test\n",
    "\n",
    "PCT_TEST = 0.02\n",
    "n_tail = len(X) - int(len(X) * PCT_TEST)\n",
    "print('Corte en el tweet número {} de los {} tweets del Dataset.'.format(n_tail, len(X)))\n",
    "\n",
    "X_train = X[:n_tail]\n",
    "y_train = y[:n_tail]\n",
    "X_test = X[n_tail:]\n",
    "y_test = y[n_tail:]\n",
    "\n",
    "print('Tweets de entrenamiento: {}'.format(len(X_train)))\n",
    "print('Tweets de Test: {}'.format(len(X_test)))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<hr>\n",
    "\n",
    "\n",
    "## One-Hot Encoding y codificación del target\n",
    "\n",
    "\n",
    "* Con las Redes Neuronales Recurrentes vamos a tratar los tweets como secuencias de palabras y por tanto tenemos que realizar una codificación de todas palabras que aparecen en los tweets.\n",
    "\n",
    "\n",
    "* Para ello TensorFlow nos proporciona una función llamada \"one_hot\" que dado un texto y el tamaño que tendrá nuestro vocabulario, nos codifica el texto por palabras. En el siguiente enlace esta la documentación de esta función:<br>\n",
    "https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/text/one_hot\n",
    "\n",
    "\n",
    "* Veamos a continuación un ejemplo de como codificamos los textos:\n",
    "\n",
    "\n",
    "```python\n",
    "one_hot(\"Esto es un ejemplo de onehot encoding\", 100)\n",
    "```\n",
    "\n",
    "```ssh\n",
    ">> [6, 88, 30, 81, 31, 52, 22]\n",
    "```\n",
    "\n",
    "* Si realizamos una codificación de un string que continene palabras identicas, nos lo codifica con el mismo identificador como es de esperar:\n",
    "\n",
    "```python\n",
    "one_hot(\"Esto es otro ejemplo\", 100)\n",
    "```\n",
    "\n",
    "```ssh\n",
    ">> [6, 88, 79, 81]\n",
    "```\n",
    "\n",
    "\n",
    "* Por tanto vemos como utiliza identificadores numéricos del 1 al 100 para codificar las palabras de los strings:\n",
    "\n",
    "    + \"Esto\" = 6\n",
    "    + \"es\" = 88\n",
    "    + \"un\" = 30\n",
    "    + \"ejemplo\" = 81\n",
    "    + \"de\" = 31\n",
    "    + \"onehot\" = 52\n",
    "    + \"encoding\" = 22\n",
    "    + \"otro\" = 79\n",
    "    \n",
    "    \n",
    "* Los identificadores que asigna a las palabras los obtiene aplicando una función \"*hash*\" a las palabras, de ahí que no sea necesario crear un \"modelo\" de \"one-hot Encode\" (como en scikit) para codificar otros strings (strings a predecir). \n",
    "\n",
    "\n",
    "* Lo único que hay que respetar para que nos codifique una misma palabra con el mismo identificador es el ***tamaño del vocabulario 'n'*** que se le pasa a la función \"*one_hot(text, n)*\" que en el ejemplo puesto anteriormente tiene valor *100*.\n",
    "\n",
    "\n",
    "* Este valor '*n*' del vocabulario debe de tener (a priori) un valor lo suficientemente grande como para poder codificar todas las posibles palabras que podamos utilizar en una lengua (Español, Inglés, Alemán, etc.).\n",
    "\n",
    "\n",
    "* El último paso que vamos a realizar es la codificación del nombre del partido político utilizando la clase \"LabelEncoder()\" de scikit.\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import LabelEncoder\n",
    "from tensorflow.keras.preprocessing.text import one_hot\n",
    "\n",
    "# Hacemos un one-hot encoder del texto\n",
    "VOCAB_SIZE = 100000 # Poner un valor muy alto\n",
    "X_train = [one_hot(doc, VOCAB_SIZE) for doc in X_train]\n",
    "X_test = [one_hot(doc, VOCAB_SIZE) for doc in X_test]\n",
    "\n",
    "# Codificación del Target\n",
    "encoder = LabelEncoder()\n",
    "encoder.fit(y_train)\n",
    "y_train = encoder.transform(y_train)\n",
    "y_test = encoder.transform(y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<hr>\n",
    "\n",
    "\n",
    "## Creacción de las secuencias de palabras\n",
    "\n",
    "\n",
    "* Las Redes Neuronales necesitan un tamaño fijo de entrada y los tweets tienen un número variable de palabras, por lo que hay que hay que transformar los tweets a secuencias de palabras de tamaño fijo.\n",
    "\n",
    "\n",
    "* Para realizar esta labor, TensorFlow nos proporciona la función \"*pad_sequences(sequences, maxlen)*\" que dada una lista de secuencias nos devuelve esa secuencia con el tapaño que le hayamos indicado como parámetro, rellenando en caso de que la secuencia sea de mayor tamaño con algún valor por defecto (en principio a cero). En el siguiente enlace esta la documentación de esta función:<br>\n",
    "https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/sequence/pad_sequences\n",
    "\n",
    "\n",
    "* Un ejemplo de esta función sería la siguiente, en el que indicamos que nos devuelva las secuencias que le pasamos como parámetro con un tamaño de '10' posiciones y rellenando las posiciones faltantes con ceros (value=0) añadiendolos al final de la secuencia (padding='post'):\n",
    "\n",
    "```python\n",
    "sequences = [[6, 88, 30, 81, 31, 52, 22], [6, 88, 79, 81]]\n",
    "pad_sequences(sequences=sequences, maxlen=10, padding='post', value=0)\n",
    "```\n",
    "\n",
    "\n",
    "```ssh\n",
    "array([[ 6, 88, 30, 81, 31, 52, 22,  0,  0,  0],\n",
    "       [ 6, 88, 79, 81,  0,  0,  0,  0,  0,  0]], dtype=int32)\n",
    "```\n",
    "\n",
    "\n",
    "* A continuación pasamos a crear las secuencias de los tweets (50 elementos) tanto de entrenamiento como de test:\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
    "\n",
    "MAX_WORDS_TWEET = 50\n",
    "pad_corpus_train = pad_sequences(X_train, maxlen=MAX_WORDS_TWEET, padding='post')\n",
    "pad_corpus_test = pad_sequences(X_test, maxlen=MAX_WORDS_TWEET, padding='post')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<hr>\n",
    "\n",
    "\n",
    "# Definición de la Red Neuronal\n",
    "\n",
    "\n",
    "* La arquitectura de la red para la clasificación de las tendencias políticas es la siguiente:\n",
    "<span></span><br><br>\n",
    "    - ***Capa Embedding***: Capa de Embedding que soportará un tamaño de 100.000 palabras (VOCAB_SIZE) y cada palabra estará representada por 32 características (EMBEDDING_SIZE). -*ver Bonus Track II*-\n",
    "<span></span><br><br>\n",
    "    - ***Capa LSTM***: Capa LSTM con 32 celdas de memoria. Son 32 por el tamaño de representación de la palabra (EMBEDDING_SIZE).\n",
    "<span></span><br><br>\n",
    "    - ***Capa 4***: Capa de salida con 5 neurona (una por tendencia política) y una función de activación 'softmax' para que nos \"normalice la salida\"\n",
    "    \n",
    "    \n",
    "* Para la capa de salida aplicamos un dropout del 20% para que nuestra red no se sobreajuste y generalize mejor\n",
    "\n",
    "\n",
    "* Por último vamos a compilar la red:\n",
    "<span></span><br><br>\n",
    "    - ***Función de perdida: 'categorical_crossentropy'*** (Funciones de Perdida: https://www.tensorflow.org/api_docs/python/tf/keras/losses). En realidad usaremos la '*sparse_categorical_crossentropy*' ya que la salida no la hemos codificado con un one-hot encoding \\[\\[1,0,0,0,0\\], \\[0,1,0,0,0\\], ..., \\[0,0,0,0,1\\],\\] si no con un label encode \\[0,1,2,3,4\\].\n",
    "<span></span><br><br>\n",
    "    - ***Optimizador: 'adam'*** (Optimizadores: https://www.tensorflow.org/api_docs/python/tf/keras/optimizers)\n",
    "<span></span><br><br>\n",
    "    - Metricas: en cada epoch pediremos que nos muestre el accuracy."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding (Embedding)        (None, None, 32)          3200000   \n",
      "_________________________________________________________________\n",
      "lstm (LSTM)                  (None, 32)                8320      \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 32)                0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 5)                 165       \n",
      "=================================================================\n",
      "Total params: 3,208,485\n",
      "Trainable params: 3,208,485\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from tensorflow.keras.layers import Dense, Dropout, Flatten, Input, LSTM\n",
    "from tensorflow.keras.layers import Embedding\n",
    "from tensorflow.keras.models import Sequential\n",
    "\n",
    "EMBEDDING_SIZE = 32\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Embedding(VOCAB_SIZE, EMBEDDING_SIZE))\n",
    "model.add(LSTM(EMBEDDING_SIZE))\n",
    "model.add(Dropout(0.1))\n",
    "model.add(Dense(5, activation='softmax'))\n",
    "model.compile(loss='sparse_categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "\n",
    "# Imprimimos la arquitectura de la red\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Parámetros a calcular en la red:\n",
    "    - Capa de Embedding: -> 100000 palabras * 32 caracterísiticas = 3200000\n",
    "    - Capa LSTM: 4*(32 * 32 + 32<sup>2</sup> + 32) = 8320\n",
    "    - Conexiones \"LSTM -> Capa salida\" = (32 * 5) + 5 = 165\n",
    "    \n",
    "    \n",
    "<hr>\n",
    "\n",
    "\n",
    "# Entrenamiento de la Red Neuronal\n",
    "\n",
    "\n",
    "* Vamos a entrenar la red con los siguientes parámetros:\n",
    "<span></span><br><br>\n",
    "    - *batch_size = 128* : Cada 128 tweets se calculan los nuevos pesos de la red\n",
    "<span></span><br><br>\n",
    "    - *epochs = 10*\n",
    "<span></span><br><br>\n",
    "    - *validation_split = datos de test* \n",
    "    \n",
    "\n",
    "* Dado que puede ser un entrenamiento muy pesado y que puede tardar varios minútos, vamos a utilizar la herramienta del \"Tensorboard\" para monitorizar la evolución de la función de perdida y el Accuracy.\n",
    "\n",
    "\n",
    "* Para más información sobre esta herramienta ver el siguiente enlace:<br>https://www.tensorflow.org/tensorboard\n",
    "\n",
    "\n",
    "* Básicamente lo que hacer Tensorboard es leer una serie de logs y representar esos logs por medio de una interfaz web.\n",
    "\n",
    "\n",
    "* Para representar la evolución de la función de perdida y el Accuracy, tenemos que escribir estos valores en un log que debemos guardarlos en una ruta conocida ('tensorboard_logs/FECHA_EJECUCION') y que generaremos  en el siguiente código.\n",
    "\n",
    "\n",
    "* Para instanciar el Tensorboard lo hacemos de la siguiente manera por consola, pasandole como parámetro la ruta donde se encuentran los logs: \n",
    "```ssh\n",
    ">> tensorboard --logdir ./tensorboard_logs/\n",
    "```\n",
    "\n",
    "* Una vez instanciado el Tensorboard nos indicará que abramos en el navegador el siguiente enlace:<br> http://localhost:6006/"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Se fuerza a usar la CPU y no la GPU en caso de tener GPU en el ordenador\n",
    "import os \n",
    "os.environ['CUDA_VISIBLE_DEVICES'] = '-1'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/15\n",
      "397/397 - 8s - loss: 1.3226 - accuracy: 0.4239 - val_loss: 1.0124 - val_accuracy: 0.5932\n",
      "INFO:tensorflow:Assets written to: ./models\\LSTM_Best_Model\\assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: ./models\\LSTM_Best_Model\\assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 2/15\n",
      "397/397 - 3s - loss: 0.7561 - accuracy: 0.7304 - val_loss: 0.8729 - val_accuracy: 0.6841\n",
      "INFO:tensorflow:Assets written to: ./models\\LSTM_Best_Model\\assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: ./models\\LSTM_Best_Model\\assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 3/15\n",
      "397/397 - 3s - loss: 0.4519 - accuracy: 0.8502 - val_loss: 0.8894 - val_accuracy: 0.7024\n",
      "INFO:tensorflow:Assets written to: ./models\\LSTM_Best_Model\\assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: ./models\\LSTM_Best_Model\\assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 4/15\n",
      "397/397 - 3s - loss: 0.2889 - accuracy: 0.9113 - val_loss: 0.9594 - val_accuracy: 0.6734\n",
      "Epoch 5/15\n",
      "397/397 - 3s - loss: 0.1959 - accuracy: 0.9415 - val_loss: 1.1339 - val_accuracy: 0.6976\n",
      "Epoch 6/15\n",
      "397/397 - 3s - loss: 0.1314 - accuracy: 0.9622 - val_loss: 1.2598 - val_accuracy: 0.6802\n",
      "Epoch 7/15\n",
      "397/397 - 3s - loss: 0.0998 - accuracy: 0.9715 - val_loss: 1.3390 - val_accuracy: 0.6812\n",
      "Epoch 8/15\n",
      "397/397 - 3s - loss: 0.0814 - accuracy: 0.9770 - val_loss: 1.4931 - val_accuracy: 0.6821\n",
      "Epoch 9/15\n",
      "397/397 - 3s - loss: 0.0671 - accuracy: 0.9810 - val_loss: 1.5881 - val_accuracy: 0.6763\n",
      "Epoch 10/15\n",
      "397/397 - 3s - loss: 0.0567 - accuracy: 0.9840 - val_loss: 1.4866 - val_accuracy: 0.6879\n",
      "Epoch 11/15\n",
      "397/397 - 3s - loss: 0.0470 - accuracy: 0.9866 - val_loss: 1.6962 - val_accuracy: 0.6647\n",
      "Epoch 12/15\n",
      "397/397 - 3s - loss: 0.0427 - accuracy: 0.9880 - val_loss: 1.7118 - val_accuracy: 0.6696\n",
      "Epoch 13/15\n",
      "397/397 - 3s - loss: 0.0361 - accuracy: 0.9901 - val_loss: 1.8102 - val_accuracy: 0.6589\n",
      "Epoch 14/15\n",
      "397/397 - 3s - loss: 0.0361 - accuracy: 0.9898 - val_loss: 1.7003 - val_accuracy: 0.6551\n",
      "Epoch 15/15\n",
      "397/397 - 3s - loss: 0.0329 - accuracy: 0.9902 - val_loss: 1.9590 - val_accuracy: 0.6454\n"
     ]
    }
   ],
   "source": [
    "import absl.logging\n",
    "absl.logging.set_verbosity(absl.logging.ERROR)\n",
    "\n",
    "from tensorflow.keras.callbacks import ModelCheckpoint, TensorBoard\n",
    "\n",
    "# Configuramos TensorBoard\n",
    "now = datetime.datetime.now().strftime(\"%Y%m%d-%H%M%S\")\n",
    "os.makedirs('tensorboard_logs/{}'.format(now))\n",
    "tensorboard_path = os.path.join('tensorboard_logs', now)\n",
    "tensorboard = TensorBoard(log_dir=tensorboard_path,\n",
    "                          histogram_freq=2,\n",
    "                          write_graph=True,\n",
    "                          write_images=True)\n",
    "\n",
    "# Creamos el checkpoint para exportar el modelo que mejor accuracy tiene con los datos de test\n",
    "model_path ='./models/LSTM_Best_Model'\n",
    "checkpoint = ModelCheckpoint(model_path,\n",
    "                             monitor='val_accuracy',\n",
    "                             save_best_only = True,\n",
    "                             save_weights_only = False,\n",
    "                             mode='max',\n",
    "                             verbose=0)\n",
    "\n",
    "history=model.fit(pad_corpus_train, \n",
    "                  y_train, \n",
    "                  batch_size=128, \n",
    "                  epochs=15, \n",
    "                  validation_data=(pad_corpus_test, y_test), \n",
    "                  verbose=2,\n",
    "                  callbacks=[tensorboard, checkpoint])\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Pintamos como evoluciona la función de perdida y el Accuracy por epoch tanto en entrenamiento como en test."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "# Pintamos las métricas por epoch\n",
    "def plot_metric(history, name, remove_first=0):\n",
    "    metric_train = np.array(history.history[name])[remove_first:]\n",
    "    metric_test = np.array(history.history['val_{}'.format(name)])[remove_first:]\n",
    "    acum_avg_metric_train = (np.cumsum(metric_train) / (np.arange(metric_train.shape[-1]) + 1))[remove_first:]\n",
    "    acum_avg_metric_test = (np.cumsum(metric_test) / (np.arange(metric_test.shape[-1]) + 1))[remove_first:]\n",
    "    plt.figure(figsize=(20, 8))\n",
    "    plt.subplot(1, 2, 1)\n",
    "    plt.title('{} - Epochs'.format(name))\n",
    "    plt.plot(metric_train, label='{} Train'.format(name))\n",
    "    plt.plot(metric_test, label='{} Test'.format(name))\n",
    "    plt.grid()\n",
    "    plt.legend(loc='upper center')\n",
    "    plt.subplot(1, 2, 2)\n",
    "    plt.title('AVG ACCUMULATIVE {} - Epochs'.format(name))\n",
    "    plt.plot(acum_avg_metric_train, label='{} Train'.format(name))\n",
    "    plt.plot(acum_avg_metric_test, label='{} Test'.format(name))\n",
    "    plt.grid()\n",
    "    plt.legend(loc='upper center')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABIcAAAHiCAYAAABycKzVAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8/fFQqAAAACXBIWXMAAAsTAAALEwEAmpwYAACm8klEQVR4nOzdd3hUVf7H8fdJBxJ6EkLvndB7C6B0RQULKiiCiGJZ19W17q7dn1h2FRQBERtNxQKIgCV0qdJ7770lQELK+f1xBw0QSIBMbpL5vJ4nj5m5d+585hjI4XtPMdZaRERERERERETEN/m5HUBERERERERERNyj4pCIiIiIiIiIiA9TcUhERERERERExIepOCQiIiIiIiIi4sNUHBIRERERERER8WEqDomIiIiIiIiI+DAVh0R8iDFmuzHmOrdzuMUY8x9jzBdu5xARERHxFcYYa4yp7HYOtxhjxhhjXnE7h0hGVBwSEVcYY+41xqQYY+Iv+CrpdjYRERHxHmNMrDHmmDEm2PO4uTHmlDEmLJ1z/zDGPOz5PsgY8y9jzAbP+XuMMdOMMR0zeD9jjNlqjFl7iWOPGmNWe6652xjzlTGmTppzmhhjfjTGHDfGHDXGLDLG9PMcu9cYMzed6/55Q85THLDGmBsvOOe/nufv9TxO9yZW2uKKp+0GZPB5L2zf4Wn6WWeNMUlpHk8zxpT3vEeAMeYZY8zsdK5Z3PPa2rm5D+dp46QLch93O5dITqDikIi4aYG1NvSCr71uhxIRERHvMMaUB1oDFrgRwFq7ANgN9Lzg3NpATWCc56mvgR5AX6AIUAH4H9Atg7dtA0QAFY0xjS849j/gMeBRoChQFfju3DWNMc2BX4FZQGWgGPAg0CWzn9ljI3BPms8WANwKbLnC61zWJdp30Ll+FvAaMCFNv+vCz/E50MIYU+GC5+8AVllrV3se5+Y+3IQLchd2O5BITqDikIiPMsYEe+5Y7fV8/TfNHabixpgpae6QzTHG+HmO/dNzpy7Oc+eug5fybffcvVrrufv1iTEmJM3x+40xmz35fkh7t8oYU8sYM9Nz7IAx5tk0lw4yxnzmyb/GGNMozeuy5bOJiIj4sL7A78AY0hRLgE89xy48d6q19ohnFM71QA9r7UJr7VnP10/W2scyeM97gO+BHzm/QFMFGAz0ttb+aq1NtNaettZ+aa19w3PaEOBTa+3/WWsPW8dSa+1tV/i5JwMtjTFFPI87AyuB/Vd4nYxcqn0zxVq7G6cY1ied6356reGMMYU8/bBDxpgdxpjn0/QxKxtjZhljThhjDhtjJnieN8aYd40xBz3HVnoKh1nOM4LqUeOMNDtsjBmSJp+fJ+8OT5bPjDGF0ry2lTFmvqf/vMt4RoR5FDHGTPX0MRcaYypl92cTyYiKQyK+6zmgGVAPqAs0AZ73HHsC5w5eOBAJPAtYY0w14GGgsbU2DOgEbPdixrs871EJ507e8wDGmPbA68BtQBSwAxjvORYG/Az8BJTEucv3S5pr3ug5tzDwAzDU87rs/mwiIiK+qC/wpeerkzEm0vP850BrY0xZcP4hDtwJfOY5fh2w0FO8yDRjTH6gV5r3vMMYE+Q53AHYba1ddJnXNscZsXStEnD6HXd4Hvflr8+WlS7VvlfiU9IUhzx9pHr8NYLrWrwPFAIqAm1x8vbzHHsZmIEzKqy051yAjjijv6ri9N9uB45kQZZLuRloBDTAGal2n+f5ez1f7Tz5Q/mrH1kWmObJHI7TXsvTXLM38CLOZ9sMvOp5Prs/m8glqTgk4rvuAl6y1h601h7C+YV1riOQhFN0KWetTbLWzrHWWiAFCAZqGmMCrbXbrbXXMhy6mefuyrmvC6811Fq7y1p7FOeXaO802Udba5dZaxOBZ4DmnqHU3YH91tq3rbUJ1to4a+3CNNeca6390VqbgtMRret5Pqs/m4iIiKRhjGkFlAMmWmuX4kypuhPAWrsLZ+rW3Z7TOwAhwFTP4+KkGWVjjCnq6TucMMYkXOZtbwEScYoOU4AA/pqGVgzYd5nXFsH599LlzrkSnwF9PaNN2uJMX8syl2vfK/QtEGmMaeF53BeY5ukvnpNRHy69fP44xY9nPP2z7cDbnN//LAeU9PTh5qZ5PgyoDhhr7Tpr7bX8P7ntguy/XXD8/6y1R621O4H/cn7/8x1r7VZrbTxO//MO40wRvAv42Vo7ztN3PmKtXZ7mmpOstYustck4hbt6XvpsIldNxSER31USZ8TNOTs8z4EzhHozMMMzrPZpAGvtZuBvwH+Ag8aY8SadxQeNMWVNmoX+LpPhd2tt4TRflS44vusS+c7L7vkFfQQoBZTh8vP30w7fPg2EGGMCMvvZRERE5KrdA8yw1h72PB7LpaeW9QHGWmuTPI+P4Ny4AsDzj/fCQEOcmzuXe8+J1tpkzw2lSWne87xrpuMYkJrBOclAYDrPB+L8w/9PnmJHOM5I6CnW2jMZXcsYc+5xEhnLqH0zxVp7GvgKp5BlcAofF04py6gPl57iQBAX9z9Leb5/CjDAIs/U//s8eX7FGaEzDDhgjBlhjCl44cWNMa3T9D/XXCbHxAuyt7vgeKb6n57vA3BG2V9p/zP0Sj6bSHZQcUjEd+3FuTtzTlnPc3ju5jxhra0I3AD8/dz6O9basdbac3emLPB/F17YWrsz7UJ/15CxTHr5LsxujCmAc/dvD84v9Mx0UC6Smc8mIiIiV84Ykw9nOnhbY8x+Y8x+4HGgrjHm3CjeSUApY0w7nBE/aadd/QI0NsaUvoL3LA20B+5O8569gK7GmOKea5Y2adYfTMtTJFnABQtlX2AnUNZTRDn3vvlxFsDekc75X+BM309vStlOoPwFz1XAGd285zIZMtu+V+JTz/WuxxnZMuUqrnGhw/w1Ouicsng+m7V2v7X2fmttSeAB4APj2aXNWvuetbYhUAtnCtaTF17cM9L9XP+z1jXkzFT/03MsGTjAtfU/M/xsItlBxSER3zUOeN4YE+7pIP0Lp8OCMaa7cRYFNMBJnE5JijGmmjGmvXEWrk4AzniOectgY0xpY0xRnHWPJnieHwv0M8bU82R5DWcdgu04nZcSxpi/GWfR7TBjTNOM3siFzyYiIuJLbsL5vVoTZ0pNPaAGMAfPaCFr7Smc9X0+AXZYa5ece7G1dgbwG/CdMaapcba1D8RZP/FS+uDsEnZuzZx6OP/43o2zCPUm4ANgnDEmxnPNEGPMHedGTeOMZrnXGPOkMaYYgDGmrjFmvOf4Qpx+w9Oe1xYA3gCWkH5x6D2cgstF28XjrJdYzRjTxxgT6On/vAZ87ZmOdE6A573OfQVmpn2v0BzgODACGG+tPXsV1ziPZ0r/ROBVT/+sHPB3/up/3pqm+HcM50ZdijGmsef/eSBwCqe9vdlHe9IYU8QYUwZnJ7tz/c9xwOPGmArGmLQ7v52bKnadMeY2Y0yAMaaYMaZeRm/kwmcTuSQVh0R81ys4HZeVwCpgmec5gCo4izrH49wx+8BaG4szbPsNnDs/+3Huij3L1WuedvqZ5yvtFrNjcdYI2Or5egXAWvsL8ALwDc46AJXwLPBorY3D6XTd4Mm4CWfhwIxk9WcTERGRv9wDfOIZXbz/3BfOlJq7POu2gDNipRzpj6y5Becm0Bc4hYttOFOeOl/mPT9I+36e9xzOX9OtHuWvaT3HcaYG3YyzuxjW2vk4o4/aA1uNMUdxCiY/eo4n4qxhFINTdNqKM/3oNs96jefxTIf75RLHDgJdcUbNHARWAyeABy849UOcm1jnvj4h8+2bKZ58n3Hp/xcZ9eEu5RGcIshWYC5OX2+051hjYKFnSYIfgMestduAgsBInILRDpzpgG9dyee5wO3pZI9Ic/x7YCnOgtJTgY89z4/GWa9yNs7PXoLn8+BZn6grzqiwo57XZmbEVlZ/NpGrZtL5e0lExHXGmO3AAGvtz25nEREREZG8zxhjgSqetShFfIpGDomIiIiIiIiI+DAVh0REREREREREfJimlYmIiIiIiIiI+DCNHBIRERERERER8WEqDomIiIiIiIiI+LAr2tIwuxQvXtyWL1/eK9c+deoUBQoU8Mq1cxO1g9rgHLWDQ+3gUDs41A4Ob7bD0qVLD1trw71ycbkq6n95n9rBoXZwqB0cageH2sGhdnC40QfLkcWh8uXLs2TJEq9cOzY2lpiYGK9cOzdRO6gNzlE7ONQODrWDQ+3g8GY7GGN2eOXCctXU//I+tYND7eBQOzjUDg61g0Pt4HCjD6ZpZSIiIiIiIiIiPkzFIRERERERERERH6bikIiIiIiIiIiID8uRaw6JSNZJSkpi9+7dJCQkpHu8UKFCrFu3LptT5TyXa4eQkBBKly5NYGBgNqcSERGR3Ep9sMxRH0wkZ1BxSCSP2717N2FhYZQvXx5jzEXH4+LiCAsLcyFZznKpdrDWcuTIEXbv3k2FChVcSCYiIiK5kfpgmaM+mEjOoGllInlcQkICxYoVS7dTIhkzxlCsWLFL3vUTERERSY/6YNdGfTCR7JVhccgYU8YY85sxZp0xZo0x5rF0zjHGmPeMMZuNMSuNMQ3SHOtsjNngOfZ0Vn8AEcmYOiXXRu0nIiIiV0N9iGuj9hPJPpkZOZQMPGGtrQE0AwYbY2pecE4XoIrnayDwIYAxxh8Y5jleE+idzmtFJI8LDQ3N8mu++uqr1KtXj3r16uHv7//n9++9916mXj9gwADWrl2b5blEREREcgr1wUQkszJcc8hauw/Y5/k+zhizDigFpP0T3QP4zFprgd+NMYWNMVFAeWCztXYrgDFmvOdc/W0gItfkueee47nnngOcjs/y5cvPO26txVqLn1/6NfBRo0Z5O6KIiIhInqM+mEjedEULUhtjygP1gYUXHCoF7ErzeLfnufSeb3qJaw/EGXVEZGQksbGxVxIt0+Lj47127dxE7eA7bVCoUCHi4uIueTwlJeWyx7NKXFwc1lpeeOEFZs6ciTGGJ598kp49e7J//37uvfde4uLiSE5O5t1336Vp06YMHjyYP/74A2MMd999Nw8//PBlr79jxw569uxJ69atWbx4MWPHjuXdd99l2bJlnDlzhh49evzZmenatSuvvPIKDRo0ICoqigceeIAZM2YQEhLC+PHjiYiIOO/6CQkJPvHz4it/LjKidnCoHURE8gZrLU899RTTpk3DGMPzzz/P7bffzr59+7j99ts5efIkycnJfPjhh7Ro0YL+/fuzZMkSjDHcd999PP7445e9/vbt2+nSpQvt2rVjwYIFfPfdd7zxxhssXryYM2fO0KtXL1588UUAYmJieOutt2jUqBGhoaE8+OCDzJgxg3z58vH9998TGRmZHU0iIhfIdHHIGBMKfAP8zVp78sLD6bzEXub5i5+0dgQwAqBRo0Y2JiYms9GuSGxsLN66dm6idvCdNli3bt2fO0C8OHkNa/ee/8c3JSUFf3//q75+zZIF+fcNtTI8LywsjG+++Ya1a9eyatUqDh8+TOPGjenUqRM//PADXbt25bnnniMlJYXTp0+zceNGDh48+Oew4+PHj192R4+wsDBCQ0PZtGkTn3766Z93pd58802KFi1KSkoKHTp0YNu2bURHR+Pv70+BAgUICwvj1KlTNG3alLfffpunnnqKcePG8fzzz593/ZCQEOrXr3/V7ZRb+Mqfi4yoHRxqBxGRrOFmHwxg0qRJLF++nBUrVvzZB2vTpg1jx46lU6dO5/XBli9fzp49e1i9ejXg9MEyY8OGDXzyySd88MEHgDP9LG0fbOXKlURHR5/3mlOnTtG4cWOGDBnCU089xciRIy/qg4lI9sjUbmXGmECcwtCX1tpJ6ZyyGyiT5nFpYO9lnhcRHzR37lx69+6Nv78/kZGRtG3blsWLF9O4cWM++eQT/vOf/7Bq1SrCwsKoWLEiW7du5ZFHHuGnn36iYMGCmXqPcuXK0axZsz8fT5w4kQYNGlC/fn3WrFmT7hz3oKAgOnfuDEDDhg3Zvn17lnxeERERkZxAfTARyUiGI4eMs0T8x8A6a+07lzjtB+Bhz5pCTYET1tp9xphDQBVjTAVgD3AHcGfWRBeRK5Xe3aW4uLjLjsjJSs6yZBdr06YNs2fPZurUqfTp04cnn3ySvn37smLFCqZPn86wYcOYOHEio0ePzvA9ChQo8Of327Zt46233mLx4sUUKVKEe++9N93tUAMDA//cDcPf35/k5OSr/IQiIt5njBkNdAcOWmtrX+a8xsDvwO3W2q+zK5+IXEx9MPXBRHK6zIwcagn0AdobY5Z7vroaYwYZYwZ5zvkR2ApsBkYCDwFYa5OBh4HpwDpgorV2TVZ/CBHJHdq0acOECRNISUnh0KFDzJ49myZNmrBjxw4iIiK4//776d+/P8uWLePw4cOkpqbSs2dPXn75ZZYtW3bF73fy5EkKFChAoUKFOHDgANOmTfPCpxIRyXZjgM6XO8GzY+z/4fTBRMTHqQ8mIhnJzG5lc0l/7aC051hg8CWO/YhTPBIRH3fzzTezYMEC6tatizGGN998kxIlSvDpp58yZMgQAgMDCQ0N5bPPPmPPnj3069eP1NRUAF5//fUrfr+6detSv359atWqRcWKFWnZsmVWfyQRkWxnrZ3t2STkch7BWRKgsfcTiUhOpz6YiGTkinYrExG5GvHx8QAYYxgyZAhDhgw57/g999zDPffcc9HrMnun6tz1y5cv/+fiieeMGTMm3dek3YEpPj7+zx3bevXqRa9evTL1viIiOZExphRwM9AeFYdEfJr6YCKSWSoOiYiIyMWOboNLrFEhOd5/gX9aa1POreVxKcaYgcBAgMjIyPP+0ZaV4uPjvXbt3ETt4PCVdihUqNCfhY/0pKSkXPa4r8ioHRISEnzi58VX/lxkRO0AwQkHiU/Kl+3toOKQiIiInO/0URh1HZWLNIV27dxOI1euETDeUxgqDnQ1xiRba7+78ERr7QhgBECjRo1sTEyMVwLFxsbirWvnJmoHh6+0w7p16y674HR2Lkidk2XUDiEhIdSvXz8bE7nDV/5cZMRn2yE1Fbb+CotGwsbpLK/7MvViHsnWCCoOiYiIyPl+/g+cOca+mtdT2u0scsWstRXOfW+MGQNMSa8wJCIiIi47cxyWj4XFI+HoVigQAW2f4vTZ7O+BqTgkIiIif9n5Oyz7FJo/zKngChmfL9nOGDMOiAGKG2N2A/8GAgGstcNdjCYiIiKZsX+1UxBaORGSTkOZZtDuOahxIwQEcdaFqXUqDomIiIgjJQmmPA4FS0PMM7BgiduJJB3W2t5XcO69XowiIiIimZWSBOsmw+JRsGMeBIRAnVuhyf0QVdftdCoOiYiIiMeCYXBwLdwxFoJD3U4jIiIikvvFHYClY2DpJxC3D4qUh46vQL27IH9Rt9P9yc/tACKS94WGZv0/Ml999VXq1atHvXr18Pf3//P79957L9PXeO2117I8l0iudWwHxL4B1bpB9W5upxERkSygPpiIS6x1pup/fR+8WwtiX4PI2nDnRHhkGbR4JEcVhkAjh0Qkl3ruued47rnnAKfjs3z58iu+xmuvvcazzz6bxclEciFr4ccnwfhB1zfdTiMiIjmY+mAil3H2NKz6ytl17MAqCC4ETQZC4/5QrJLb6S5LI4dEJNtYa3nyySepXbs2derUYcKECQDs27ePNm3aUK9ePWrXrs2cOXNISUnh3nvv/fPcd999N8Prp6Sk8OSTT9K4cWOio6P56KOPLnn9p59+mjNnzlCvXj3uuusur35ukRxv3Q+waTq0exYKaX8yEZG8Rn0wES87uhWmPwfvVIfJjwIWbvgfPLEOOr+W4wtDoJFDIr5l2tOwf9V5T+VLSQb/a/iroEQd6PJGpk6dNGkSy5cvZ8WKFRw+fJjGjRvTpk0bxo4dS6dOnXjuuedISUnh9OnTLF++nD179rB69WoAjh8/nuH1P/74YwoVKsTixYtJTEykZcuWdOzYkUmTJl10/datWzN06NA/73bFxcVdbQuI5G4JJ2HaPyGyDjQd5HYaEZG8SX0w9cEk70lNhS2/wKIRsGkm+Pk7u401uR/KNgdj3E54RVQcEpFsM3fuXHr37o2/vz+RkZG0bduWxYsX07hxY+677z6SkpK46aabqFevHhUrVmTr1q088sgjdOvWjY4dO2Z4/RkzZrBy5Uq+/vprAE6cOMGmTZvSvb6IePz2KsTth9u/uLZ/pIiISI6lPphIFjpzDP740tl17Ng2CI2Etv+EhvdCwSi301019QJFfEk6d5fOxMURFhaWLW9vrU33+TZt2jB79mymTp1Knz59ePLJJ+nbty8rVqxg+vTpDBs2jIkTJzJ69OgMr//+++/TqVOni46ld30Rn7f3D+duV6P7oHQjt9OIiORd6oOpDya53/5VTr9p5VeQfMYZHdThBah+AwQEuZ3ummnNIRHJNm3atGHChAmkpKRw6NAhZs+eTZMmTdixYwcRERHcf//99O/fn2XLlnH48GFSU1Pp2bMnL7/8MsuWLcvw+p06deLDDz8kKSkJgI0bN3Lq1Kl0rw8QGBj457kiPic1BaY8DgXCocO/3E4jIiJepD6YyFVKPgurv4HRnWF4K6cwFH0rPDAH7vsJavfME4Uh0MghEclGN998MwsWLKBu3boYY3jzzTcpUaIEn376KUOGDCEwMJDQ0FA+++wz9uzZQ79+/UhNTQXg9ddfz/D6AwYMYPv27TRo0ABrLeHh4Xz33XfExsZedH2AgQMHEh0dTYMGDRg+fLhXP7tIjrN4lDNyqOfHkK+w22lERMSL1AcTuUIn98HSMbD0E4g/AEXKQ8dXof5dkK+I2+m8wlxqiKGbGjVqZJcsWeKVa8fGxhITE+OVa+cmagffaYN169ZRo0aNSx6Py8YhzTlZRu2QUTvmFb7y5yIjeb4dTu6FoU2gTGO4e9IlF0z0ZjsYY5ZaazWXLQdR/8v71A4OX2kH9cEyR30wh6/8uciIq+2QmgI7Fzg30NZNdh5Xud7Zir5SB/DLvolXbvTBNHJIRETE1/z0NKSchW5v57qdNERERESyzMm9sOVX2PwzbI11FpsOKeTs4NrovlyxBX1WUXFIRETEl2ycAWu/h3bPQ9GKbqcRERERyT5JCbBzPmz+xSkKHVzrPB8aCVW7QOUOUK0rBOV3N6cLVBwSERHxFWdPw49PQPGq0PJRt9OIiIiIeJe1cHgTbPnFKQhtn+vsNOYf5Ow2dv1LzpSxyFo+P5paxSERH2Ctxfj4X3bXIieuzSZyVWa/Ccd3wr1TISDY7TQiInme+mDXRn0wuSpnjsO2WX+NDjqxy3m+WGVo0BcqXwflW0JQAVdj5jQqDonkcSEhIRw5coRixYqpc3IVrLUcOXKEkJAQt6OIXJsDa2H++1DvLijfyu00IiJ5nvpg10Z9MMm01BTYu/yv0UG7F4NNgeCCUKENtP67MzqoSDm3k+ZoKg6J5HGlS5dm9+7dHDp0KN3jCQkJ+qXL5dshJCSE0qVLZ3MikSyUmgpTHnc6Sde/7HYaERGfoD5Y5qgPJlfl5L6/ikFbf3MWksZAyXrQ6nFndFDpRuAf6HbSXEPFIZE8LjAwkAoVKlzyeGxsLPXr18/GRDmT2kHytD8+h12/Q49hUKCY22lERHyC+mCZo3aQTElKcLaZ3/ILbP4VDq5xnk+7kHTFdurnXAMVh0RERPKy+EMw819QtoUzpUxEREQkp7MWjmx2RgZt/vmChaSbaSFpL1BxSEREJC+b+QKcPQXd31XnSURERHKuhBMUP7QAJn/rjA46sdN5/s+FpDs46yZqIWmvUHFIREQkr9o2G1aMg9ZPQER1t9OIiIiInM9a2LMUFo2ANd9SO+UsBIVBxbbQ6m9OQahIebdT+gQVh0RERPKi5ERnEeoi5aHNk26nER+2fv9JJm44S5s2Fj8/jV4TEREg6QysnuQUhfYtdwpCDfvxx9ly1L9hoBaSdoGKQyIiInnR3P86c/Xv/gYC87mdRnzY+n1x/LgtiWU7j9GofFG344iIiJuObYfFHzubZZw5BuE1oNvbEH07BIdxIjZWhSGXqDgkIiKS1xzZAnPehlq3OFu5irjoupqRBPrBlJX7VBwSEfFFqamw9VdYNBI2TgfjBzW6Q+P7nTWEtCZijqDikIiISF5irTOdLCAYOr/udhoRQoMDiA7358dV+3ihe038NbVMRMQ3nDkOy8fC4pFwdCsUiHCmuje8FwqVcjudXEDFIRERkbxk1VewbRZ0fQvCSridRgSAJiUCWLoikcXbj9KsYjG344iIiDftX+WMElr1FSSdhjJNod1zUONGCAhyO51cgopDIiIiecWZYzD9WSjZABrd53YakT/VC/cnX6A/U1buVXFIRCQvSkmCdT/AolGwcz4E5IPoW52pY1HRbqeTTFBxSERE5JzUVFj1FYWP7Qdi3E5z5X5+EU4fcRah9vN3O43In4IDDO1rRPDT6v3854ZaBPj7uR1JRESywsl9sHSM8xW/39klteOrUP8uyFfE5XByJVQcEhERAdi3Aqb8HfYsoR5AwDbo+AoEh7ocLJN2LYKln0CzwRBV1+00Ihe5ITqKqSv3sXDbUVpWLu52HBERuVrWwo75zlpC6yZDagpUuR6aDIVKHcBPNwByIxWHRETEtyWchN9eg0UfQf5icNNwdi6dTtmlY2Drb3DTcCjX3O2Ul5eSBJP/BgVLQbtn3U4jkq6YahEUCHKmlqk4JCKSCyXGw6qJztSxg2sgpDA0HQSN+0PRim6nk2uk4pCIiPgma2HNJPjpWYg/4HRs2r8A+Qqz9XgUZa8bCN8Ogk+6QItHnIUUA0PcTp2+3z9wOmm3f5l7RjqJzwkJ9Oe6mpH8tHo/L/WoTaCmlomI5A6HN8PiUc7OY4knoEQduPF9qN0LgvK7nU6yiIpDIiLie45sgR//AVt+hah60HsslGp4/jnlWsCD82DG8zD/Pdg0E275KOdN2Tq+E2LfgKpdoHo3t9OIXFb36JJ8v3wv87ccoW3VcLfjiIjIpaSmwMbpztSxLb+CXyDUuslZYLpMEzDG7YSSxVQcEhER35GUAHPfdb4CgqHLEGfE0KUWbw4Ogxv+B9W7w/cPw8j20PZpaPU4+OeAX6HWwo9POt93fVMdNcnx2lQtTlhwAFNW7FVxSEQkJzp1BP74DBaPhhM7IawktHseGt4DoRFupxMvygE9WxERkWyw+RdntNDRrc4w6E6vQliJzL22yvXw0AKnEPPbK7BxGtz8ERSv4t3MGVk/BTb+BNe/DIXLuptFJBOCA/y5vlYk09fs59Wb6xAUoKllIiI5wp5lsGgkrP4GUhKhfGvo9ApU6wr+gW6nk2yg4pCIiORtJ/fB9GdgzbdQrDL0+Q4qtbvy6+QvCr0+dqZuTf07DG8F170ITQa6sytHYhz8+BRE1oZmD2b/+4tcpRuiSzJp2R7mbj5E++qRbscREfFdifGw6itnt9N9KyCwANS/G5rcDxE13E4n2UzFIRERyZtSkp158r++CilnnQWlWz7mTCe7FrVvcdYj+uFR+Omfzuidmz7I/pE7v70Gcfvgts90R09ylZaVi1MoXyBTVuxTcUhExA37VsCST5zC0Nl4iKgFXd+C6NsgpJDb6cQlKg6JiEjes3sJTPkb7F8Fla+DrkOydovVsBJw5wT443P46Rn4oAV0eQPq3ZU96/7sXQ4Lh0OjflCmsfffTyQLBQX40alWJNNW7SchKYWQwEus+SUiIlnn7ClYPckZJbRnKQSEQK1bnL5E6cZat1AyLg4ZY0YD3YGD1tra6Rx/ErgrzfVqAOHW2qPGmO1AHJACJFtrG2VVcBERkYucOQY/vwhLx0BYlDOqpsaN3unwGAMN+kKFNvDdYPh+MKyf6ixg7c0FG1NTnMJX/mLQ4V/eex8RL+oeXZKJS3Yze+MhOtbK5NpfIiJy5Q6scUYJrZwAiSchvDp0/j+oezvkK+J2OslBMjNyaAwwFPgsvYPW2iHAEABjzA3A49bao2lOaWetPXyNOUVERC7NWlgxDma84BSImg+GmKed3ca8rUh5uGcyLPzQKUx90Ay6vws1e3jn/RZ/DHv/gFtGqVMnuVbzSsUokj+QKSv3qTgkIpLVks44ay0u+QR2LwL/YKdf0ug+KNtMo4QkXRkWh6y1s40x5TN5vd7AuGtKJCIiciUOrncWiN4xD0o3ge7vQIk62ZvBz88pSFW+Dr59ACb2hTq3OdvLZ2UB5+Q++OUlqBgDdXpl3XVFslmgvx+da0fx/fI9nDmbQr4gTS0TEblmB9c708ZWjIOEE85GHB1fhXp3OhtriFxGlq05ZIzJD3QGHk7ztAVmGGMs8JG1dkRWvZ+IiPi4s6dg1puwYKgzQuiG96B+H3d2DjsnvBr0nwlz3obZQ2D7XOgxFCp3yJrrT3/GWVy72zu66ye53g3RUYxbtJPYDQfpUifK7TgiIrlTUgKs+8EZJbRzPvgFQs0boWE/KN9K/QXJtKxckPoGYN4FU8paWmv3GmMigJnGmPXW2tnpvdgYMxAYCBAZGUlsbGwWRvtLfHy8166dm6gd1AbnqB0cagdHbmmHYocXUmXTKEISD7KvRAe2VryHpLhCMDvdXzFX7NrboRmh9f+PGuv+S4EvbmFPyS5sqXQvqf4hV33FokeWEb3mW7aVv5Mdq3YBu64hX+bklp8HyZ2aVChK8dAgpqzcp+KQiMiVOrzJWWNx+Vg4c9TZeOP6l5zNMQoUdzud5EJZWRy6gwumlFlr93r+e9AY8y3QBEi35+4ZVTQCoFGjRjYmJiYLo/0lNjYWb107N1E7qA3OUTs41A6OHN8Ox3fCtH/Chh8hvAZ0/4yocs3J6n9WZk07xEDnu+HXlym1YBilzqyHm4c7c/2v1NnT8MGjUKwKFe7+LxUCgq8xW+bk+J8HydUC/P3oUjuKr5bu4lRiMgWCtYmuiMhlJSfCuslOUWj7HPALgOrdnFFCFdq6O3pacr0s+ekxxhQC2gLfp3mugDEm7Nz3QEdgdVa8n4iI+JjkszD3XRjWFLbGOnfGBs2Bcs3dTnZ5gSHQ6VW4dyrYFPikC8z8t9O5uxKzh8DxHc56StlUGBLJDt2jo0hISuXX9QfdjiIiknMd2QIz/wXv1IRv+jt9gg7/gsfXOjuzVmqnwpBcs8xsZT8OiAGKG2N2A/8GAgGstcM9p90MzLDWnkrz0kjgW+PMcQwAxlprf8q66CIi4hO2z3MWnD60Hqp3h85vQOEybqe6MuVbwoPzYfpzMO+/sGmmM4ooKjrj1x5cB/Pfg7q9oUIbr0cVyU6NyhclIiyYKSv3ckPdkm7HERHJOVKSYP1UZ4HprbFg/KFaF2jUDyq2VzFIslxmdivrnYlzxuBseZ/2ua1A3asNJiIiPu7UYWdr+hVjoXBZ6D0BqnV2O9XVCw6DG99zClw/PAIj20PMP6Hl4+B/iV/Hqakw5XHntR1fyd68ItnA38/QtU4UYxftJC4hibCQQLcjiYi469h2WPop/PEFnDoIBUtDu+eh/t1QUOuzifdocreIiOQsqamwbAz8/KKzI1nrJ6D1PyAov9vJskbVjvDQApj6BPz6Cmz4CW7+CIpXvvjc5V/CzgVw4/taXFLyrBvqRjFm/nZ+WXeQm+qXcjuOiEj2S0mm+KEF8Pl7sOVXZ4exKp2cUUKVrwM/f7cTig9QcUhERNyVmuKMEorfDyd2w5x3YM8SKN8aur3tbA+f1+QvCrd+AjW6O0Wi4a3g+heh8f1/DRM/dRhmvgBlm0O9u93NK+JF9csUIapQCFNW7lVxSER8S2Kcs7j0gg+oHbcXwkpC239Cgz5QqLTb6cTHqDgkIiLekZwI8Qcg7oBT+Inb73l8wX9PHQKb+tfrCoTDzSMg+jbnzlleVrsnlGvpTDOb9hSsnwI9PnDWVJrxgtNp7P6u1hWQPM3Pz9CtThSfLtjOiTNJFMqnqWUiksfFH4KFw2HxSEg4AeVbs6rcvdS5+YlLTzUX8TL95ImIyJVJjLt0wSdtMejMsYtfa/yc4k9oJISVgKi6zn/PPQ4tARE1IDg0+z+XW8JKwJ0TYdmnzoLVH7ZwhpGvGAut/u60h0ge171uSUbN3cbMtQfo1VB3y0Ukjzq2HeYPhT8+d26i1ejurD1YuiFHYmNVGBJX6adPREQcKUkUiN/uzHX/s/iTzn+TTl38Wv8gp7ATFgnFKjm7c4VGpin6eP5bIFzz5tNjDDS8FyrGwHcPwbz/QeFy0OZJt5OJZIu6pQtRukg+pqzcq+KQiOQ9+1c7u5WunuTcKKt7O7R4DMKrup1M5E8qDomIiHP3akw3Gu9eDEvSPB8UBqERnlE+9aBqiYsLPqGRkK9I3p8Clh2KlId7psDKCVCiTt5ZhFskA8YYukVH8fGcbRw7dZYiBYLcjiQicm2shR3zYe67sHkmBIVCsweh+WAoWNLtdCIXUXFIRERg+rOwezGbK/Wjcpvb/hr140vTu3IKPz+o19vtFCLZ7oboknw0ayvT1+znjiZl3Y4jInJ1UlNh4zSnKLR7MeQvDu2fh8YDnJtpIjmUikMiIr5u5VeweBS0eITdQddRuVwLtxOJiA+qVbIg5YvlZ+qqfSoOiUjuk3wWVn3lTA0/vAEKl4Wub0H9uyEwn9vpRDKk4pCIiC87uB4mP+Zsl97h3zBnntuJRMRHnZtaNnzWVo7EJ1IsNNjtSCIiGUuMdzaVWDAMTu6ByNrQ82OoeZMWmJZcRXvjioj4qsR4mNjXWdem1yfgr+2jRcRd3aNLkpJqmbZ6v9tRREQu79Rh+PVVeLeWMz2/SAW462sYNBfq9FJhSHId/cSKiPgia50RQ0c2QZ9voWCU24lERKheIoxK4QWYunIfdzcr53YcEZGLHd/pbEe/7DNIPgPVu0PLv0GZxm4nE7kmKg6JiPiiJR/D6q+dBRIrxridRkQEODe1rCRDf93EwbgEIsJC3I4kIuI4sMZZT2jV184OrdG3Q8vHILya28lEsoSmlYmI+Jo9S+GnZ6Dy9dDqCbfTiIic54boKFItTFulqWUikgPsWABf3gYftoB1U6DpIHhsBdz0gQpDkqdo5JCIiC85fRQm3utsU3/LCGfbdBGRHKRKZBjVIsOYunIf97Qo73YcEfFFqamwabqzHf2uhZC/GLR7ztmOPn9Rt9OJeIWKQyIiviI1Fb4dBHH74L7p6tyISI7VLTqKd3/eyP4TCZQopKllIpJNUpKcaWPz/guH1kOhstBliLMdfVB+t9OJeJVuGYuI+Ip57zp3wTq/DqUbup1GRK6SMWa0MeagMWb1JY7fZYxZ6fmab4ypm90Zr1X36Cishamr9rkdRUR8wdlT8PuH8L968N0gMH5wy0h4dBk0HajCkPgEjRwSEfEF22bDr69A7Z7OkGgRyc3GAEOBzy5xfBvQ1lp7zBjTBRgBNM2mbFmiYngoNaMKMnXlXvq3quB2HBHJi6yFg2th9TewZDScOQZlW0D3d6HK9c6i0yI+RMUhEZG87uQ++Po+KFYZbvifOjsiuZy1drYxpvxljs9P8/B3oLTXQ3lBt+gohkzfwJ7jZyhVOJ/bcUQkL7AW9q+Etd87X0c2O6OEqnZ2tqMvm6vq6CJZSsUhEZG8LCXZKQydPQX3TIbgMLcTiUj26g9Mu9RBY8xAYCBAZGQksbGxXgkRHx9/xdcOP5MKwHvfzqVLhUAvpMp+V9MOeZHawaF2cHi9HawlLG4z4YfmE35oPvkS9mPx41iROhyq+iCHizcjKagwbD0DW72YIwP6eXCoHRxutIOKQyIiedmvL8HO+c68+YgabqcRkWxkjGmHUxxqdalzrLUjcKad0ahRIxsTE+OVLLGxsVzNtb/YOpd1p+D/Yi75EXKVq22HvEbt4FA7OLzSDqmpsGfJXyOETuwCvwCoGAM1n8NU60bRAsUoCuSUzej18+BQOzjcaAcVh0RE8qr1U2He/6DRfRB9m9tpRCQbGWOigVFAF2vtEbfzXK1udaJ4fdp6dh45TdliWhBWRC4jNcXZdn7t97D2B4jbC/5BUKk9tHsWqnWBfEXcTimSY6k4JCKSFx3dBt8+CFH1oNPrbqcRkWxkjCkLTAL6WGs3up3nWnSLdopDU1bt5aGYym7HEZGcJiXZGSG99ntYNxniD0BACFS+Dmq+BFU7QUhBt1OK5AoqDomI5DVJCTCxLxjgtk8hMMTtRCKShYwx44AYoLgxZjfwbyAQwFo7HPgXUAz4wDgL0Cdbaxu5k/balC6Sn/plCzNlxT4Vh0TEkZIE2+d4CkJT4PRhCMzv7DBW8yao0hGCQ91OKZLrqDgkIpLX/PRPZyeO3uOhSHm304hIFrPW9s7g+ABgQDbF8bpudaJ4Zeo6th6Kp2K4/sEn4pOSz8K2WbD2O2fa/JljEBTq7DJWs4czUihIU09FroWKQyIiecnycbB0DLR63JlbLyKSy3WLdopDU1fu45EOVdyOIyLZJSkBtv7mjBBa/yMknoDgglCtq1MQqtReo6NFspCKQyIiecWBNTDlcSjXCto973YaEZEsEVUoH43LF2GKikMied/Z07D5Z1j3A2z4Cc7GQUhhqHGDUxCq2BYCgt1OKZInqTgkIpIXJMY56wyFFIReo8Fff72LSN7RrU4U/5m8lk0H4qgSGeZ2HBHJQv7JZ2D1JGeE0KYZkHQa8heD2rc4BaEKbcA/0O2YInme/vUgIr7t+E6Y8TwNd66Gih9A2WZuJ7py1sIPj8DRrXDPZAiLdDuRiEiW6lonihenrGXKyn08fr2KQyJ5wpEtEPs6LdZ8D6lnoUAE1O3tFITKtdSNLpFspj9xIuKbks7AvPdg7jtg/Agy+WB0J2h0H1z3Hwgp5HbCzFs0AtZ86+Qu38rtNCIiWS6iYAhNKxRlysq9/O26Knh2YROR3Cj+IMz6P2eNRP9g9pe4jlIdH3Zu0Pn5u51OxGf5uR1ARCRbWevscjGsCcS+5iza/PBiFjUZBs0GOx2VoU2coc3Wup02Y7sWw/TnoGoXaPGY22lERLyme3RJthw6xYYDcW5HEZGrkRgPsW/Ae/VhySfQ4B549A82VX0AyrdUYUjEZSoOiYjvOLwJvugJ4++EwALOFKxbx0Ch0qQE5IPOr8H9v0JohLN+z/g74cRut1Nf2qkj8NW9ULAk3Pwh+OmvdBHJuzrXLoGfgSkr9rkdRUSuREoSLBoJ79WD2NedXcYGL4Lu72gqvEgOon9JiEjelxgHM/8FHzSH3Yuh8xswaI6zwOGFStaH+3+Djq/A1lgY1hQWfgSpKdke+7JSU2HS/XDqENz2GeQr4nYiERGvKh4aTItKxZmyci82N4zsFPF11sKa75y+1I//gGJVoP/PcPvnULyy2+lE5AIqDolI3mUtrPwKhjaGef+D6NvhkaXQ7MHL73rhHwAtHoGHFkCZpjDtKfj4eti/OvuyZ2TOW7DlF+jyf1CynttpRESyRffoKLYfOc2avSfdjiIil7N9Hoy6Dr66B/yDoPcE6PcjlGnsdjIRuQQVh0Qkb9q/GsZ0g0kDIDTSuVN10zBnylhmFSkPd38DPT+GYztgRFv4+T/OYtZu2vIb/PYaRN8BDe91N4uISDbqVKsEAX6GKSs1tUwkRzqwFsbeDmO6wsm9cONQeHAeVOsMWkheJEfTbmUikrecOeYUThaPgpDCcMP/oH6fq1/k0Bio08uZHz/zBZj7rjNEuvu7UKldVibPnBN74Jv+EF7dmauvjpaI+JAiBYJoWdmZWvbPztW0a5lITnFij7PRx/KxEBQKHf4NTQdBUH63k4lIJmnkkIjkDampsPRTeL+hUxhq1N+ZQtbw3qzZ/SJ/UegxzFnE2vjB5zfBpAfg1OFrv3ZmpSTB1/0gOdGZrx9UIPveW0Qkh+geHcXuY2dYsfuE21FE5MxxZ1T1+w1g5URo+iA8tgJa/12FIZFcRiOHRCT3273UWehw7zIo2xy6DoESdbzzXhXawIPzYc7bziiiTTOg02tQ9w7vj+L5+T+wayH0Gg3Fq3j3vUREcqiONUvwrP8qpq7cS70yhd2OI+KbkhOdHcjmvOWM2q5zG7R/HoqUczuZiFwljRwSkdwr/hB8PxhGtXfmtd8yCvpN815h6JzAEGj/nLPjWfEq8N0g+KwHHNnivfdc+wMsGApNHoDaPb33PiIiOVyh/IG0qRLO1JX7SE3VrmUi2So1FVZMgPcbwYznnF1eH5gNPUeqMCSSy6k4JCK5T0oy/D7cmUK2YgK0eBQeWQLRt2bvGjwRNaDfT9Dtbdj7B3zYAua840z/ykpHtjhFsFKNoOMrWXttEZFcqHvdKPaeSOCPXcfcjiLiOzb/AiPawLcDIV9h6PMd9PkWouq6nUxEsoCmlYlI7rJtjrO1/MG1ziLRnf8Pwqu6l8fPDxoPgGrdnFy/vAirvoYb34PSja79+klnYGJfZ92kW8dAQNC1X1NEJJe7rkYkQQF+TFm5j4blirodRyRv27scfv43bI2FwmWdkdq1ezp9IBHJM/QnWkRyhxN74Kt+8Gl3OBsPt38Jd09ytzCUVsEoZ5HoO8Y6c+9HXQc/PgWJcdd23R//AQfWOB2xwmWyJquISC4XFhJITNVwflylqWUiXnNsO3zdH0a0hX0rodPr8LBnpLYKQyJ5jkYOiUjOlpzorLUz+y2wqRDzDLR8DALzuZ0sfdW7QfnW8OsrsGgErJ8CXd+C6l2v/FrLPoc/voA2T0GV67I+q4hILta9bklmrD3A4u1HaVqxmNtxRPKOU0dg9hBn91e/AGj9hNP3CinkdjIR8aIMS77GmNHGmIPGmNWXOB5jjDlhjFnu+fpXmmOdjTEbjDGbjTFPZ2VwEfEBG2fAB83gl5ecKWSDF0HM0zm3MHROSEHo+iYM+BlCCsP43jDhbji5L/PX2L/KGTVUoa3zmUVE5DwdqkcQEujH1FVX8HeriFza2dPOzbj36sGij6Beb3h0GXT4lwpDIj4gM+MBxwCdMzhnjrW2nufrJQBjjD8wDOgC1AR6G2NqXktYEfERR7fC2Dtg7K1g/J3pY3d8mft2wSjdCB6YBR3+DZtmwrAmzl241NTLvy7hhLPOUL4i0PNjZ70hERE5T4HgANpXj+DHVftJ0dQykauXkgxLx8B79eHXl6F8K3hwAdz4PhQs6XY6EckmGRaHrLWzgaNXce0mwGZr7VZr7VlgPNDjKq4jIr7i7GlnOtawZrB9Dlz/Ejw4Hyp3cDvZ1fMPhNZ/h4cWQKkGMPUJ+KQzHFyX/vnWOjuTHdvhLEAdGp6tcUVEcpPu0SU5HJ/Iwq1H3I4ikvtYC+unOrutTn7MWWy630/QexxEVHc7nYhks6xaSay5MWaFMWaaMaaW57lSwK405+z2PCcicj5rYc13MLSxM8e9Zg9nwcOWj+Wd3bmKVnS2fL35Izi8CYZ71iVKSjj/vN8/gHWTncJY2WauRBURyS3aVYsgf5A/UzS1TOTK7F4Cn3SB8Xc6azre/gX0nwHlmrudTERckhULUi8Dyllr440xXYHvgCqASefcS475NcYMBAYCREZGEhsbmwXRLhYfH++1a+cmage1wTlut0PImX1U2/ABRY6vJL5ABTbVe40ThWvBsg3AhmzLkX3tUILA+v+j0pbRlJg9hNOLv2Rj1Yc4XqQOBU+so97yFzhSvBlrEmuBC/9f3P55yCnUDg61g+R0+YL86VAjkp9W7+elG2sR4K8dlEQu69h2Zy3H1d9AgXDo9g40uAf8tU+RiK+75r8FrLUn03z/ozHmA2NMcZyRQmn3XS4N7L3MdUYAIwAaNWpkY2JirjVaumJjY/HWtXMTtYPa4BxX2+HAWvjsfkhJhK5vEdqwH/Vd6pxkfzvcCFt+I/+Uv1FvxfNQ7y7Y8hsULkv4gInEuLTwo/5cONQODrWD5Abdo6OYvGIv87ccoU1VTcUVSdeZ4zDnLVj4kbOeY5snnRHawWFuJxORHOKa/xVmjCkBHLDWWmNME5ypakeA40AVY0wFYA9wB3Dntb6fiOQRe5fD5zdDQDDcOxPCq7mdKPtVaucs+Dj7TZj3nrM+Uf+Z2hFEROQKtK0aTmhwAFNW7lVxSORCyWdhyWiY9YZTIKp3J7R7DgpptQ8ROV+GxSFjzDggBihujNkN/BsIBLDWDgd6AQ8aY5KBM8Ad1loLJBtjHgamA/7AaGvtGq98ChHJXXYthi96Olu+3/ODsx6PrwrKD9f9B+r2hrOnICra7UQiIrlKSKA/19eMZPqaA7xyUypBAZpaJuIsNj0FZv7L2QW2Qhvo+ApE1XU7mYjkUBkWh6y1vTM4PhQYeoljPwI/Xl00EcmTts+FsbdDaAT0/QEKl8n4Nb7AF0dOiYhkke7RUXz7xx7mbT5Mu+oRbscRcdfupTDjOdi5AIpXgzu/girXg0lvSVgREYdWHhOR7LP5Zxh/FxQpD32/h7ASbicSEZE8oFWV4oSFBDB55V4Vh8R3HdsBv7z412LT3d+F+n212LSIZIr+phCR7LF+Knx1rzNCps93UKC424lERCSPCA7wp1OtEkxfvZ/E5BSCA/zdjiSSfc4chzlvw8LhWmxaRK6aJmWLiPet/gYm9oUSdeCeySoMiYhIluseHUVcYjKzNx52O4pI9khJcnYfe68+zH8faveCR5ZC++dVGBKRK6aRQyLiXcvHwveDoUwzuHOCswi1iIhIFmtZuTiF8wcyZeVerq8Z6XYcEe/5c7Hpf8PRLVpsWkSyhIpDIuI9i0fB1CegYju4Y6yzM5eIiIgXBPr70blWCSav2EtCUgohgZpaJnnQnqUw/XnYOd+z2PREqNJRi02LyDXTtDIR8Y75Q53CUNUu0Hu8CkMiIuJ13aNLcupsCrEbDrodRSRrHdsBX/eHke3hyCZnsekH50PVTioMiUiW0MghEcl6s4bAb69AzZvglpEQEOR2IhER8QHNKhalWIEgJq/cR+faUW7HEbl2Z47D3Hfg9+FOEaj1P5zFpjVNX0SymIpDIpJ1rIVfXnI6MdF3QI9h2j5VRESyTYC/H51rl2DSsj2cPptM/iD9DpJcKiUJlnwCsa/DmWNQ9w5noelCpd1OJiJ5lKaViUjWsBZ+esYpDDXsBzd9qMKQiIhku+7RJTmTlMKv6zW1THIha2HdFBjWFKY9CZG1YGAs3DxchSER8Sr9y01Erl1qKkx9HJaOgWYPQafXNP9dRERc0aRCUcLDgpmyYh/do0u6HUck8/YshRkvwI55ULwq9J6gNYVEJNuoOCQi1yYl2dmqfuV4aP0EtH9BnRgREXGNv5+ha+0SjF+8i/jEZEKD1d2VHO74Tmda/qqvIH9x6PYONLhHI7BFJFtpWpmIXL3ks/DNfU5hqP3z0OFfKgyJiIjrutctSWJyKr+sO+B2FJFL8k8+BTP/Be83gnWTnZtsj/4BjfurMCQi2U5/64jI1UlKgK/ugY0/OdPImg92O5GIiAgADcsWoUTBECav2EePeqXcjiNyvrOnYOmnNF34OiSddDbx6PCC1hQSEVepOCQiV+7sKRh/J2yNdYY+N+7vdiIREZE/+fkZutaJ4ovfd3DiTBKF8gW6HUnE2XVs0Uj4/UM4c5RThesQdNtQKFnP7WQiIppWJiJXKOEkfNELts12diRTYUhERHKg7nWjOJuSys9rNbVMXBa3H2Y8D+/Wht9ehdKN4b7prKj3igpDIpJjaOSQiGTe6aPwZS/YtwJ6fgy1b3E7kYiISLrqlylMqcL5mLJyLz0barqOuODoVpj3Hiz/ElKTodYt0OpxKFHbOb411tV4IiJpqTgkIplz6jB8dhMc3gC3fQ7Vu7qdSERE5JKMMXSLjmL03G0cP32WwvmD3I4kvmL/apj7LqyZBH4BUO9OaPEoFKvkdjIRkUtScUhEMnZyH3zWw9lqtfd4qNzB7UQiIiIZ6h4dxYjZW5mx5gC3NS7jdhzJ63b+DnPegU3TISjU2ayj2WAoGOV2MhGRDKk4JCKXd3wnfHojnDoEd38N5Vu5nUhERCRT6pQqRNmi+Zm8cq+KQ+Id1sLmX2DO27BzPuQrCu2egyb3Q74ibqcTEck0FYdE5NKObHFGDCWchD7fQZnGbicSERHJtHNTy0bM3sqR+ESKhQa7HUnyitQUWPu9M31s/0ooWAo6vwEN+kJQAbfTiYhcMe1WJiLpO7QBPunqbFt/zw8qDImISK7UPTqKlFTL9DXatUyyQHIiLP0UhjaGr/tB0mnoMQweXQ7NHlRhSERyLY0cEpGL7V/lLD5t/KDfjxBRw+1EIiIiV6VmVEEqFi/AlJV7ubNpWbfjSG6VGA9Lx8CCYRC3F6Lqwq2fQo0bwM/f7XQiItdMxSEROd/upfDFzc5Cin1/gOKV3U4kIiJy1YwxdI+OYuhvm9l55DRli+V3O5LkJqePwqIRsHA4nDkG5VtDj6FQqT0Y43Y6EZEso2llIvKXHfOdNYZCCkO/aSoMiYhInnBXs3IE+PvxQexmt6NIbnFyL0x/Dt6tDbGvQ9nm0P9nuHeKs2urCkMiksdo5JCIOLb8BuPvdBZUvOcHKFjS7UQiIiJZIrJgCLc3KsP4xTt5pEMVShXO53YkyamObIF5/4MV45xFp2v3hFaPQ2RNt5OJiHiVRg6JCGycDmNvhyIVnDWGVBgSEZE8ZlBMJQCGx25xOYnkSPtWwlf9YGgjWDEe6veBR5dBz5EqDImIT9DIIREfF35wHsx+B0rUgbsnQf6ibkcSERHJcqUK56NXw9JMWLyLwe0qU6JQiNuRJCfYMR/mvAObZ0JQGLR4FJo9BGGRbicTEclWGjkk4qtOH4VfX6Hm2regVEPo+70KQyIikqc92LYyKdby0WyNHvJp1sLGGTC6M3zSBfYug/YvwOOr4foXVRgSEZ+kkUMivubUEfh9GCwcAWfjOBjRmsi7x0NwqNvJREREvKpssfzcXL8UYxfu5MGYSkSEafSQT7EW1k+FWW/A/lVQqAx0edOZQhakXexExLepOCTiK04dhvnvw6KRkHQaat0EbZ5k3bpDRKowJCIiPmJwu8pMWrabUXO28WzXGm7HkeyQmgrrp8CsN+HAKihaEXp8ANG3gX+g2+lERHIEnykOHT99lm//2EP+U6luRxHJXvEHnV03loyG5ARn143W/4CI6s7xdbGuxhMRkbzPP/mU2xH+VKF4AW6oW5LPF+zggTYVKRYa7HYk8ZbUVFg/2VMUWg3FKsPNH0HtXuDvM/8MEhHJFJ9ZcygxOZUXJ69l6YFkt6OIZI+4/fDTM/DfaPj9A6hxIwxeBD1H/VUYEhER8bbV39B04YPOGi85xMPtKpOQnMLHc7e5HUW8ITUV1nwHw1vBxL6QnAi3jHT6QXXvUGFIRCQdPvM3Y2TBEGpEFWTl4Xi3o4h414k9zkihpWMgNdnpBLV+AopVcjuZiIhkAWPMaKA7cNBaWzud4wb4H9AVOA3ca61dlr0p04iszdmgIgSNvRWaDYbr/g0B7o7WqRIZRtfaUXy2YAcD21SkcP4gV/NIFklNhbXfwewhcHAtFK8Kt4yC2reAn7/b6UREcjSfGTkE0LZqOJuOpRKfqNFDkgcd3wVT/g7v1YMlHzvz6B9ZCjd9oMKQiEjeMgbofJnjXYAqnq+BwIfZkOnSwquxrMEQaDLQ2RBh1HVweLOrkQAebl+Z+MRkPpm33e0ocq1SU2D1N/BhC/i6n/O458fw0O8QfasKQyIimeBTxaGYauGkWJi3+bDbUUSyzrEdMPkxeK8+LPsM6t0FjyyDHkOhaAW304mISBaz1s4Gjl7mlB7AZ9bxO1DYGBOVPenSl+ofBF2HwB3j4MRu+KgN/PGls3uUS2pEFaRjzUhGz9vGyYQk13LINUhNgVVfwwfN4ev7AAu9RsNDC6BOLxWFRESugE8VhxqWK0KIP8zaeMjtKCLX7ug2+P5heL8BLB8LDe+Bx5bDDf+FIuXcTiciIu4pBexK83i35zn3Ve8KD86DUg3g+4fgmwGQcNK1OI+0r0JcQjKfzd/uWga5CqkpsPIr+KAZfNMfjB/0+gQeXOBsvKGikIjIFfOZNYcAAv39qFXcn1kbDmGtxZmSL5LLHNkCc96GFePBLwAa9YdWf4OCJd1OJiIiOUN6HZx0h+gYYwbiTD0jMjKS2NhYrwSKj48//9rlHqesKUuF1eNI2DyHtTWfIK5gNa+8d0bqhvvz4W8bqZS6m3wB3u0bXtQOPuqq28GmEHFwDuW3TyT/mT3EFyjHjppPcSi8ORz2g9mzszyrN+nnwaF2cKgdHGoHhxvt4FPFIYA6xf1ZuuYMWw7FUzkizO04Ipl3aCPMeQtWfQX+wdB0ELR8FMJKuJ1MRERylt1AmTSPSwN70zvRWjsCGAHQqFEjGxMT45VAsbGxXHztDrDzHvJ9M4CGy5+F9s9Di8fAL3sHtheqeIybP5jPjsCyDGrr3TX60m8H33PF7ZCS7KwpNPtNOLIZImrBDa8TWv0GamXzz0tW0s+DQ+3gUDs41A4ON9rBJ4tDALEbDqk4JLnDwfXOrhurv4HAfNB8MLR4FEIj3E4mIiI50w/Aw8aY8UBT4IS1dp/LmdJXtikMmuOsnffzf2BrLNz8Ubbe+KhftgitqxRn5Oyt3NO8PPmCNCUpx0hJdm6KzR4CR7dAZB247XOo3j3bi4giInmdz/2tWiyfH1UiQrXukOR8B9bAxHuc+fQbf3Kmjv1tFXR8RYUhEREfZowZBywAqhljdhtj+htjBhljBnlO+RHYCmwGRgIPuRQ1c/IVhlvHwI3vw86Fzo5TG2dka4RHO1ThyKmzfLlwR7a+r1xCSrKznuKwxvDdIAjKD7d/CQ/Mhpo3qjAkIuIFPjdyCJxdyz6dv4PTZ5PJH+STTSA52b6VzrDpdZMhKAxaP+GMFspf1O1kIiKSA1hre2dw3AKDsylO1jAGGvSFMk2dXafG3grNBsN1/4aAYK+/fePyRWlWsSgjZm/l7mblCAnU6CFXpCTDygnOSKFj26BENNwxFqp1dX5GRETEa3yy7N62agRnU1JZsOWI21FE/rL3Dxh3J3zUGrbOhrb/hL+thA4vqDAkIiK+IbwaDPgFmjwAvw+DUdfB4c3Z8taPdqjCwbhEJi7ZlfHJkrVSkmDZ5zC0obOLXUhBuGOcM1KoejcVhkREskGGxSFjzGhjzEFjzOpLHL/LGLPS8zXfGFM3zbHtxphVxpjlxpglWRn8WjSuUIT8Qf6aWiY5w+6l8OVtMCIGdsyDds85RaF2z6ooJCIivicwBLq+Cb3Hw4nd8FEb+ONLsOluuJZlmlcsRqNyRfgwdguJySlefS/xSEmCZZ/B+w3hh4chpDD0ngADZ0F1jRYSEclOmRk5NAbofJnj24C21tpo4GU8O16k0c5aW89a2+jqIma94AB/WlQqRqxnS3sRVxzeBF/0hFHtYfciaP+Cs6ZQ26ec9RdERER8WbUu8OA8KNXAGU3yzQBIOOG1tzPG8EiHKuw7kcA3S/d47X0ETGoSLP0U3m8APzwC+YvBnRNhYCxU66yikIiICzJccMdaO9sYU/4yx+enefg7znapOV7bquH8vO4g24+cpkLxAm7HEV+z9gf47iHwD4Dr/gONB0Cwds8TERE5T8GS0Pd7mPsO/PY67F4MvUZDae/cc2xTpTh1yxTmg9jN3NqoNIH+PrkCg/cknIQV42m6cAgkHoRSDaHbO1D5OhWERERcltW/8foD09I8tsAMY8xSY8zALH6va9K2qrPbU+yGgy4nEZ+Skgwz/w0T+0B4VRg0F1o9rsKQiIjIpfj5Q5snod80Z2rZ6E4w5x1ITc3ytzLG8Gj7yuw+doZv/9DooSxzcB1M+Tu8UwOmPUlicBG46xtnfakq16swJCKSA2TZVl3GmHY4xaFWaZ5uaa3da4yJAGYaY9Zba2df4vUDgYEAkZGRxMbGZlW088THx/957RL5Dd8u2ECFJN/btjRtO/iq7G6DwLMnqLn2LYocX8neqE5sqnQ/9o/NODsNu0c/Cw61g0Pt4FA7ONQOkqOUbQqD5sDkx+CXF2FrLNwyAsJKZOnbtK8eQa2SBfngt83cUr8UARo9dHVSkmD9FFg0CnbMBf9gqNMLGg/gj00niakS43ZCERFJI0uKQ8aYaGAU0MVa++cWYNbavZ7/HjTGfAs0AdItDllrR+BZr6hRo0Y2JiYmK6JdJDY2lnPX7hK3hrELd9KsZWuf27I0bTv4qmxtgz1LYcJgOHUIegyjZP27KZk975wh/Sw41A4OtYND7eBQO0iOk68w3DoG/vgcfnwKPmwBNw2Hqh2z7C2MMTzSvgqDvljKlJX7uKl+qSy7tk+I2w9LxzhfcfugcDm4/iWo3+evjTY2xboYUERE0nPNt0KMMWWBSUAfa+3GNM8XMMaEnfse6Aiku+OZW9pWDScxOZWF2466HUXysqVjYHRnMH7QfzrUv9vtRCIiIrmXMdCgLzwwC8KiYOyt8NMzkJyYZW/RsWYk1SLDeP/XTaSkavOSDFkL2+fBV/fCu7Ug9nWIrO0sMv3oH9DyMe3AKiKSw2U4csgYMw6IAYobY3YD/wYCAay1w4F/AcWAD4wzXzjZszNZJPCt57kAYKy19icvfIar1qxiMYID/IjdcJC2VcPdjiN5TVIC/PgP5+5mxXbOAprqGImIiGSN8GrOmjUz/wW/fwDb50CvT6B4lWu+tJ+f4ZEOlXl47B9MW72P7tE5ZbxvDpMYDyvHw+KP4eBaZyv6poOgcX8oWtHtdCIicgUys1tZ7wyODwAGpPP8VqDu1UfzvpBAf5pVLMasjYfcjiJ5zfGdMLEv7P0DWv8D2j3rLKgpIiIiWScwBLq+CZXaObuAftQGug6Bendd8yLHXWpHUSl8I+//spmutaPw89OiyX86tBEWj4Tl4+BsHETVhRuHQu2eEJTf7XQiInIVfH6FvZhq4Ww9dIpdR0+7HUXyii2/wUdt4cgWuGMsdHhBhSERERFvqtYFHpznbI3+/WD4ZgAknLimS/r7GR5uX5kNB+KYsfZAFgXNxVKSYe0P8OmNMKyxM22+elfo/zMMnAUN+qgwJCKSi/l8cejcdLJYjR6Sa2Wts7XuF7dAaCTc/xtU7+Z2KhEREd9QsCT0/R7aPw9rvoXhrWH3kmu65A3RJSlfLD/v/7oJa3107aH4gzBrCPwvGib2gaNbocO/4PG1zm5xZRprK3oRkTzA54tDFYoXoGzR/MzacNDtKJKbJZyECXc7W+vWvAkG/AzFK7udSkRExLf4+UObJ6HfNOemzehOzo2b1NSrulyAvx8PtavMmr0n+XW9D/UVrYWdv8PX/eGdmvDbK1C8qjMi+tHl0PoJCNV6nSIieUmWbGWfmxljaFs1nG+W7SYxOYXgAE3/kSt0cD1MuAuOboNOr0OzB3UHTURExE1lm8KgOTD5MefGzdZYZ5RLWIkrvtTN9Uvx3i+beO/XzbSvHoHJy7/jz56CVV/B4lGwfxUEF4LGA5wFprNgoW8REcm5fL44BM66Q5//voMl24/RsnJxt+NIbrJ6Enz/MAQVgHsmQ/mWbicSERERgHyF4dYxzq6hPz4F/6sHJepAidrONusloiGypvM7/DIC/f14KKYyz367ijmbDtMmL+5we2SLUxD640tIPAERtaD7fyH6tgzbR0RE8gYVh4DmlYoR5O9saa/ikGRKSjL8/G9YMBRKN4HbPnXWOhAREZGcwxho0BfKNIMlo53RMKu+cb53ToBilTzFonMFo9rO7/Q0I4R6NizF+79u4r1fNtG6SvG8MXooNQU2Tnd2HdvyK/gFQM0ezkihss01ClpExMeoOATkDwqgSYWizNp4iOe0frBkJP4gfNUPdsyFxvdDp9cgIMjtVCIiInIp4VWhyxvO99bCiV1OoWj/ajiwCvYth7Xf/XV+vqKeEUZ1oEQdgkvUZnCbsjw/eSMLth6hRaVcfDPx1GFY9hks+QRO7ISwKGj3HDS4B8Ii3U4nIiIuUXHIo23VcF79cR17j5+hZOF8bseRnGrXYmenjjPH4eaPoO4dbicSERGRK2EMFC7rfKXdVTThJBxYAwdWewpHq2DJx5CcAMBdfoE0CSnFvq8rQ5v2ntFGdSB/UZc+yAVSUyHhuHMT69RBz38Pnf84/iAcXAspZ6F8a+j0ClTrCv6BbqcXERGXqTjkEVPNKQ7N2niI3k3Kuh1HchprnQ7itKehUCkYMNPpEIqIiEjeEFIQyjV3vs5JSYajW2D/Ksz+VeRbv5jqh5fA9J//Oqdgqb8KRedGGxWtCH5ZsClwagqcPnqJYs+h858/dQhSky++hl8AFAh3vkIjnGljDfpCRI1rzyciInmGikMelSNCKVkohNgNB1UckvMlnYEpf4cVY6FKR2e3k3xF3E4lIiIi3uYfAOHVnK86vSjeNoXWb/5K04gUhnUI+mtq2v5VsPlnsCnO6wILOItdl6jzV+EooqZzLCUZTh+5uOATfyBN8cfz39OHwaZenMsv0Cn0FAh3dmArEe1sLV8gwnk+NOKv70MKZ02hSkRE8jQVhzyMMbStFsHkFXtJSkkl0F+/RAU4th0m3O10/GKegTZPqYMlIiLio/IF+TOgdUXemLaeAZ3rU79l+78OJiXAofVOoejc1LQLFr9uGRAKsfGAvfjiASGegk44FCoDpRr8VeA5N+rn3PGQwlowWkREspSKQ2nEVAtn3KKdLN1xjGYVi7kdR9y26Wf4pj9g4c6JULWj24lERETEZX2aleOjWVt4/9fNjL638V8HAkOgZD3n65wLFr8+uHEZpSpHXzy6p0A4BIep4CMiIq5RcSiNFpWKEeBnmLXxkIpDviw1Fea8Bb+9BpG14PbPnbUDRERExOcVCA6gf6sKvDVjI6v3nKB2qUKXPvmCxa83EUupmJhsyyoiIpJZmh+TRlhIII3KFyF2wyG3o4hbzhyH8b3ht1ch+jboP1OFIRERETlP3xblKRgSwHu/bHI7ioiISJZQcegCbatGsG7fSQ6cTHA7imS3/athRIyzoGSXIc5W9UH53U4lIiIiOUzBkED6tazAjLUHWLfvpNtxRERErpmKQxeIqRYOwKyNGj3kU1ZOhFHXOTuT3TsVmg7UvH8RERG5pPtaViA0OIChv252O4qIiMg1U3HoAtVLhBFZMFjFIV+RkgTT/gmT7oeS9eGB2VC2mdupREREJIcrlD+Qe1qU48fV+9h8MM7tOCIiItdExaELGGNoWzWcORsPkZyS6nYc8aKgxKMwpjssHA7NHoJ7foCwSLdjiYiISC7Rv1VF8gX6a/SQiIjkeioOpaNt1QhOJiSzYvdxt6OIt+xbScOlf4f9K6Hnx9D5dfAPdDuViIiI5CJFCwTRp1k5flixl22HT7kdR0RE5KqpOJSOVlWK4+9ntGtZXnVkC3xxC9b4w4CfoU4vtxOJiIhILjWgdUWCAvwY9ptGD4mISO6l4lA6CuULpH6ZwioO5UUn98HnN4FNZWX0ixBZy+1EIiIikouFhwXTu0lZvv1jD7uOnnY7joiIyFVRcegSYqqFs2rPCQ7HJ7odRbLKmWPwxS1w+ijc9TWnC5R2O5GIiIjkAYPaVsLfz/BBrEYPiYhI7qTi0CW0rRoBwGztWpY3nD0NY2+HI5vhji+hVAO3E4mIiEgeEVkwhNsbleHrpbvZc/yM23FERESumIpDl1CrZEGKhwZpS/u8ICUJJvaF3Yuh5yioGON2IhEREcljBsVUAmB47BaXk4iIiFw5FYcuwc/P0KZKOLM3HiIl1bodR65Waip89xBsngnd34WaPdxOJCIiInlQqcL56NWwNBOW7OLAyQS344iIiFwRFYcuo221cI6dTmLVnhNuR5GrYS1MfwZWTYQO/4KG97qdSERERPKwB9tWJiXV8tGsrW5HERERuSIqDl1G6yrhGAOxGw66HUWuxuy3YOFwaDYYWv3d7TQiIiKSx5Utlp+b65fiy4U7OBSnTU1ERCT3UHHoMooWCKJu6cJadyg3WjwKfnsF6vaGjq+AMW4nEhERER8wuF1lklJSGTVHo4dERCT3UHEoA22rhrN813GOnTrrdhTJrNWTYOo/oGpnuPF98NOPuYiIiGSPCsULcEPdknz++w6Oqv8oIiK5hP7VnIGYauFYC3M2H3Y7imTG5l9g0kAo2xxuHQP+gW4nEhERER/zcLvKnElK4eO5Gj0kIiK5g4pDGYguXZgi+QO17lBusHsJTOgD4dWh9zgIzOd2IhEREfFBVSLD6Fo7ik/n7+DE6SS344iIiGRIxaEM+PsZWlcJZ/bGw6RqS/uc6+B6+LIXhEbA3d9AvsJuJxIREREf9nD7ysQnJvPJ/G1uRxEREcmQikOZEFMtnMPxiazdd9LtKJKe4zvh85vBPwj6fAthkW4nEhERER9XI6ogHWtGMnruNuISNHpIRERyNhWHMqF1lXAA7VqWE5067BSGzp6CuydB0QpuJxIREREB4JH2VTiZkMxnC3a4HUVEROSyVBzKhPCwYOqUKqR1h3KaxDhnKtmJ3XDnBChR2+1EIiIiIn+qU7oQ7atHMGrOVk4lJrsdR0RE5JJUHMqktlXDWbbzOCfOaFhwjpCcCOPvhH0r4bbPoFxztxOJiIiIXOSR9pU5djqJL37X6CEREcm5VBzKpJhq4aSkWuZpS3v3pabAN/1h22y46QOo2sntRCIiIiLpql+2CK2rFGfknK0kpmhzExERyZlUHMqkemUKUzAkQFPL3GYtTHkc1k2GTq9D3TvcTiQiIiJyWY92qMLh+LPE7tLUMhERyZlUHMqkAH8/WlcJZ9bGQ1iruz6u+eUlWPYptP4HNH/I7TQiIiIiGWpcvijNKhblx21JWqJARERyJBWHrkDbquEcOJnI+v1xbkfxTfOHwtx3oGE/aP+822lEREREMu2ZLjWIO2v5zw9r3I4iIiJyERWHrkDbatrS3jXLx8GM56BmD+j2NhjjdiIRERGRTKtbpjA3VAzk2z/28OOqfW7HEREROY+KQ1cgsmAI1UuEad2h7LZhGnw/GCrGwC0jwc/f7UQiIiIiV+yGSoFEly7Es9+u4uDJBLfjiIiI/Mm3ikOpqdd8iZhqESzZfoz4RC0omC22z4Ov7oWounD7FxAQ7HYiERERkasS4Gd457Z6nDmbwlPfrNQ6liIikmNkWBwyxow2xhw0xqy+xHFjjHnPGLPZGLPSGNMgzbHOxpgNnmNPZ2XwK3bqMIyMocjR5dd0mbZVw0nWlvbZY99KGHcHFC4Ld30NwWFuJxIRERG5JpUjQnmmS3ViNxxi7KKdbscREREBMjdyaAzQ+TLHuwBVPF8DgQ8BjDH+wDDP8ZpAb2NMzWsJe01SzkJqCnVWvQTLPr/qyzQsV4TQ4ACtO+RtR7fCFz0huCD0+RYKFHM7kYiIiEiW6Nu8PK0qF+eVKevYfviU23FEREQyLg5Za2cDRy9zSg/gM+v4HShsjIkCmgCbrbVbrbVngfGec91RsCT0m8bxwnXgh4fh11fgKobyBgX40aJSMWZt0Jb2XhO3Hz67CVKTncJQodJuJxIRERHJMn5+hiG3RhPob/j7xOUkp1z70gciIiLXIivWHCoF7ErzeLfnuUs9756Qgqyq8wI06Auzh8CkgZCceMWXiakWwZ7jZ9hyKN4LIX3cmWPw+S3ONMC7vobwqm4nEhEREclyUYXy8fJNtVm28zgfzd7qdhwREfFxAVlwjfT2FLeXeT79ixgzEGdaGpGRkcTGxmZBtIvFn04gNuwWylZIpeKqLzi2ay1raj1DcmBopq8RfMa5uzNq6u90rhDolZzeFh8f77U2vlp+KYnUXfFvwuI2sarOCxzbHAebY732fjmxDdygdnCoHRxqB4fawaF2EPGuG+uWZMbaA7w7cyNtq4ZTu1QhtyOJiIiPyori0G6gTJrHpYG9QNAlnk+XtXYEMAKgUaNGNiYmJguiXSw2Nhbn2u1gZQxFvn+IVhtehLu+giLlM32d4etmsSc1hJiYpl7J6W1/tUMOkZIE4++Ek+vh1jHUrXWT198yx7WBS9QODrWDQ+3gUDs41A4i3mWM4dWbarN421Een7CcyY+0IiTQ3+1YIiLig7JiWtkPQF/PrmXNgBPW2n3AYqCKMaaCMSYIuMNzbs4RfSv0+Q7iD8Ko62DP0ky/NKZaOAu3HuX0WW1pf81SU+H7wbBpBnR/B7KhMCQiIiKSExTOH8SbvaLZdDCet6ZvcDuOiIj4qMxsZT8OWABUM8bsNsb0N8YMMsYM8pzyI7AV2AyMBB4CsNYmAw8D04F1wERr7RovfIZrU74l9J8Jgfnhk26wfmqmXta2agRnU1L5fesRLwfM46yF6c/CygnQ/nlodJ/biURERHI8Y0xnY8wGY8xmY8zT6RwvZIyZbIxZYYxZY4zp50ZOyZyYahHc3awsH8/bxoIt6luKiEj2y8xuZb2ttVHW2kBrbWlr7cfW2uHW2uGe49ZaO9haW8laW8dauyTNa3+01lb1HHvVmx/kmoRXhQG/QGRNGH8X/D48w5c0rlCEfIH+xG7QlvbXZM5bsPBDaPYQtP6H22lERERyPGOMPzAM6ALUBHobY2pecNpgYK21ti4QA7ztGcktOdSzXWtQvlgB/vHVCk4mJLkdR0REfExWTCvLG0LD4Z4pUL0b/PRP+OkZSE255OnBAf60qFSMWG1pf/WWjIZfX4Ho26Hjq2DSW8NcRERELtAE2Gyt3WqtPQuMB3pccI4FwowxBggFjgKaC5+D5Q8K4O3b6rLvxBle/GGt23FERMTHqDiUVlB+uO0zZxTL7x/AxL5w9vQlT4+pFs7Oo6fZfuTS58glrPkWpvwdqnSCHsPATz+KIiIimVQK2JXm8W7Pc2kNBWrgbAayCnjMWpuaPfHkajUoW4TB7SrzzbLd/LR6n9txRETEh2TFbmV5i58/dH4dCpeDn56GT7tD7wnOyKILtK0aAawhdsNBKhSvkP1Zc6NjO2DRCFj4EZRtBreOAf9At1OJiIjkJukNtb1wGHMnYDnQHqgEzDTGzLHWnjzvQsYMBAYCREZGEhsbm+VhAeLj47127dwkM+1QN8BSvqAf/5iwjIQ9+SgcnPduoOnnwaF2cKgdHGoHh9rB4UY7qDh0Kc0GQeEy8HV/GNUB7vraWZsojbLF8lOxeAFmbTxEv5YqDl2StbBzgTMaa/1UwDg7knV72xmtJSIiIldiN1AmzePSOCOE0uoHvGGdue+bjTHbgOrAorQnWWtHACMAGjVqZGNiYrwSODY2Fm9dOzfJbDuUrx1Ht/fm8sO+MD6+pxEmj02918+DQ+3gUDs41A4OtYPDjXbIe7cislL1bnDvVEg6DR9fD9vnXXRKm6rhLNhyhISkS69P5LOSE2HFeBjRFj7pAtvmQMvH4G8roddoyFfE7YQiIiK50WKgijGmgmeR6TuAHy44ZyfQAcAYEwlUw9ldVnKByhFh/LNzdX5df5Dxi3dl/AIREZFrpOJQRko3hAE/Q2gEfH4TrPzqvMMx1cJJTE5l4baj7uTLieIPQez/wX/rwLcPQFICdH8X/r4OrvsPFCrtdkIREZFcy1qbDDwMTAfWAROttWuMMYOMMYM8p70MtDDGrAJ+Af5prT3sTmK5Gve2KE/LysV4ecpadhw55XYcERHJ4zStLDOKlIf+M2D83TBpABzfAa2fAGNoVrEYwQF+xG44SNuqF69L5FP2r4Lfh8OqiZByFipfD80ehErttROZiIhIFrLW/gj8eMFzw9N8vxfomN25JOv4+RmG9KpLp//O5omJK5jwQHP8/dSfEhER79DIoczKVwT6TII6t8GvL8PkRyEliZBAf5pVLMasjYfcTuiO1BRnHaEx3WF4K1gzCRr0hcGL4e6voXIHFYZERERErkLJwvl4qUctluw4xkezt7gdR0RE8jCNHLoSAcFwywgoUg5mD4ETe+DWMbStGs5LU9ay6+hpyhT1kQWWE07CH1/Aoo/g2HYoWBquf8kpDGktIREREZEscVO9Usxce4B3Z26kbdVwapUs5HYkERHJgzRy6EoZA+2fhxuHwrZZ8EkXOpRKAiDWF0YPHd0K056Gd2rC9GcgtATc+ik8tsJZbFqFIREREZEsY4zh1ZvqUDh/EH+fsEKboIiIiFeoOHS1GvSBOyfCsR2UnXQj7QrvZ9aGg26n8g5rYdtsGHcnvNcAFo+Eal3g/l+h/3RnW3p/DUITERER8YYiBYJ4s1c0Gw7E8c7MjW7HERGRPEjFoWtRuQP0n44xfgw/+xxmyy8kJuehuzlJCc7UseGt4NMbYNfv0OYf8LfV0HMklGrodkIRERERn9CuWgR3Ni3LyDlb+X3rEbfjiIhIHqPi0LWKrAUDfuZswfJ8aP6PXTM/dDvRtYvbD7++Cu/Wgu8HOyOHbnwfHl/jTKkrGOV2QhERERGf81zXGpQtmp8nJq4gLiHJ7TgiIpKHqDiUFQqWxO++acy3dai88Dn4+UVITXU71ZXb+wdMegDere0suF26MfT9AR6c5yw0HZjP7YQiIiIiPqtAcADv3FaPfSfO8NLktW7HERGRPEQLxWSRAgWLMLLM68Qd+h/d5r4Dx3fCTR84O5zlZCnJsGEq/P4h7FwAQaHQuD80GQjFKrmdTkRERETSaFiuCA/GVGLYb1u4vmYkHWuVcDuSiIjkASoOZaE21aIYvKUvra9rTMG5r8DJvXDHl5C/qNvRLhKQFA/z34eFI+DETihcFjq9BvXvhhBtkSoiIiKSUz3WoSqxGw7xzKRV1C9bhPCwHH4zUkREcjxNK8tCMdXCAcPUgrdDr9GwZwl83BGObnMvVEqyU6TavRTWTYFFI2Hy32i+oD/MeN4pCt3+JTy6HJoPVmFIREREJIcLCvDj3dvrEZeYzDOTVmKtdTuSiIjkcho5lIUqR4RSslAIsRsO0rtPTwgrCeN7w6jroPd4KNM4697MWjh9BOL2OQtIx+2Dk/vOfxy3D+IPAhd0GPyDORTekhI3/Qei6mZdJhERERHJFlUjw3iqUzVembqOiUt2cXvjsm5HEhGRXEzFoSxkjKFttQgmr9hLUkoqgeWaQ/+f4cue8Gl36DkKatyQ8YUSTp5f4LmoALQf4vdDytmLX5u/OIRFQVgJKFEHCpZ0vj/3XFhJKFCc9bPnUEKFIREREZFc676WFfhl3UFemryW5hWLU7ZYfrcjiYhILqXiUBZrWzWccYt2snTHMZpVLAbFK8OAX2DcHTChD1z/IpRucvEInz+/3w9n4y++cHDBv4o85Vr89X3BqL8KP6GROX8BbBERERHJEn5+hrduq0vnd2fzxFfLGT+wOf5+xu1YIiKSC6k4lMVaVi5GgJ9h1sZDTnEIoEBxuGcyTBoIM/91/gv8g/8q8JSoA1U6nT/Sp2BJp+gTHJr9H0ZEREREcrRShfPxnxtr8cRXKxg5ZyuD2mq3WRERuXIqDmWxsJBAGpYrQuyGQ/yzc/W/DgTmg1s/hc0/g5//X8WffEXA6A6PiIiIiFydWxqUYubaA7wzYyNtq4ZTI6qg25FERCSX0W5lXhBTLYJ1+05y4GTC+Qf8/KBqR6jcASJrOlvcqzAkIiIiItfAGMNrt9ShYL5AHp+wnMTkFLcjiYhILqPikBe0rRoOwKyNh1xOIiIiIiK+oGiBIN7sVYf1++N4Z+ZGt+OIiEguo+KQF9SICiMiLFjFIRERERHJNu2rR9K7SRlGzN7Kom1H3Y4jIiK5iIpDXmCMoW3VcOZsPERySqrbcURERETERzzfrSZliuTnia+WE5+Y7HYcERHJJVQc8pKYahGcTEhmxe7jbkcRERERER9RIDiAd26ry55jZ3h58lq344iISC6h4pCXtKpcHD8DsRs0tUxEREREsk+j8kV5oG0lJizZxcy1B9yOIyIiuYCKQ15SKH8gDcoW0bpDIiIiIpLtHr+uKjWiCvLMpJUciU90O46IiORwKg55Uduq4azcfYLD+oUsIiIiItkoKMCP/95ej5Nnknlm0iqstW5HEhGRHEzFIS+KqRYBwJxNGj0kIiIiItmrWokwnuxUjRlrD/DV0t1uxxERkRxMxSEvqlWyIMVDg7TukIiIiIi4on+rCjStUJSXJq9l19HTbscREZEcSsUhL/LzM7SpEs7sjYdISdVQXhERERHJXn5+hrdvqwvAE1+tUJ9URETSpeKQl7WtFs6x00ms2nPC7SgiIiIi4oNKF8nPv2+oyaJtR/l47la344iISA6k4pCXta4SjjEQu+Gg21FERERExEf1aliaTrUi+b+fNvCztrcXEZELqDjkZUULBFG3dGFtaS8iIiIirjHG8NatdaldsiAPjV3G/M2H3Y4kIiI5iIpD2aBt1XCW7zrOsVNn3Y4iIiIiIj4qLCSQMf2aUL5YfgZ8toQ/dh5zO5KIiOQQKg5lg5hq4VgLc3SHRkRERERcVKRAEF/0b0rx0GDu/WQx6/addDuSiIjkACoOZYPo0oUpkj9Q6w6JiIiIiOsiCobw5YCm5Av0p8/Hi9h2+JTbkURExGUqDmUDfz9D6yrhzN54mFRtHyoiIiIiLitTND9fDGhCqrXcPWohe46fcTuSiIi4SMWhbNK2ajiH4xNZq6G7IiIiIpIDVI4I47P7mnDyTBJ9Ri3kUFyi25FERMQlKg5lkzZVwwG0a5mIiIiI5Bi1SxVidL/G7D1xhr6jF3HidJLbkURExAWZKg4ZYzobYzYYYzYbY55O5/iTxpjlnq/VxpgUY0xRz7HtxphVnmNLsvoD5BbhYcHULlVQ6w6JiIiISI7SuHxRPurTiM0H4+g3ZhGnEpPdjiQiItksw+KQMcYfGAZ0AWoCvY0xNdOeY60dYq2tZ62tBzwDzLLWHk1zSjvP8UZZFz33iakawbKdxzlxRndkRERERCTnaFs1nPd712f5ruMM/HwJCUkpbkcSEZFslJmRQ02Azdbardbas8B4oMdlzu8NjMuKcHlNTLVwUlItP63e53YUEREREZHzdK4dxZu96jJv8xEeGfcHSSmpbkcSEZFskpniUClgV5rHuz3PXcQYkx/oDHyT5mkLzDDGLDXGDLzaoHlBw3JFiC5diPd+2Uxisu7GiIiIiEjO0qthaV68sRYz1x7gqa9XaqddEREfEZCJc0w6z13qt8QNwLwLppS1tNbuNcZEADONMeuttbMvehOncDQQIDIyktjY2ExEu3Lx8fFeu3ZmdCqRwpAlCbz0xa9cXz7QtRxut0NOoDZwqB0cageH2sGhdnCoHUR80z0tyhOXkMRbMzZSINifl3vUxpj0/kkgIiJ5RWaKQ7uBMmkelwb2XuLcO7hgSpm1dq/nvweNMd/iTFO7qDhkrR0BjABo1KiRjYmJyUS0KxcbG4u3rp0ZMcC8Y7/z0644nunditDgzPwvyHput0NOoDZwqB0cageH2sGhdnCoHUR81+B2lYlLSOaj2VsJCwnkn52rux1JRES8KDPTyhYDVYwxFYwxQTgFoB8uPMkYUwhoC3yf5rkCxpiwc98DHYHVWRE8N3uyUzWOnDrL6Lnb3I4iIiIiInIRYwxPd6nOXU3L8mHsFob9ttntSCIi4kUZDlux1iYbYx4GpgP+wGhr7RpjzCDP8eGeU28GZlhrT6V5eSTwrWcYagAw1lr7U1Z+gNyoftkidKwZycjZW+nTrBxFCgS5HUlERERE5DzGGF7uUZv4xGSGTN9AWEgAfZuXdzuWiIh4QabmNFlrfwR+vOC54Rc8HgOMueC5rUDda0qYR/2jUzU6/Xc2H87awrNda7gdR0RERETkIn5+hrdurcupxBT+9f0aQoMDuKVBabdjiYhIFsvMtDLxgqqRYdxcvxSfzt/O/hMJbscREREREUlXoL8fQ++sT4tKxXjy65X8tHq/25FERCSLqTjkosevq0qqtfzvl01uRxERERERuaSQQH9G9m1EnVKFeHTcH8zZdMjtSCIikoVUHHJRmaL5ubNJWSYu2cW2w6cyfoGIiIiIiEsKBAcwpl9jKoYXYOBnS1m646jbkUREJIuoOOSyh9tXIcjfj3dmbnQ7ioiIiIjIZRXOH8Tn/ZtSolAI936ymDV7T7gdSUREsoCKQy4LDwvmvlblmbxir365ioiIiEiOFx4WzBcDmhIWHEDfjxex+WC825FEROQaqTiUAwxsU4lC+QJ5a/oGt6OIiIiIiGSoVOF8fDGgKcZAn48XsvvYabcjiYjINVBxKAcolC+QQW0r8duGQyzerrnbIiIiIpLzVQwP5bP7mnIqMZm7Ry3kYJx24BURya1UHMoh7m1RnoiwYN78aT3WWrfjiIiIiIhkqGbJgoy5rwkH4xLpM2oRx0+fdTuSiIhcBRWHcoh8Qf480qEKi7cfI3aDtgYVERERkdyhQdkijOzbiG2HT3HPJ4uJT0x2O5KIiFwhFYdykNsblaFs0fy8OX0DqakaPSQiIiIiuUPLysUZemd9Vu85wYBPF5OQlOJ2JBERuQIqDuUgQQF+/P36qqzbd5Ipq/a5HUdEREREJNM61irB27fWZeG2owz+chlJKaluRxIRkUxScSiHubFuSaqXCOOdGRv0C1VEREREcpWb6pfi5R61+WX9Qf4+cQUpGg0vIpIrqDiUw/j5Gf7RsRrbj5zmqyW73Y4jIiIiInJF7m5Wjqe7VGfyir08/90qbbYiIpILqDiUA3WoEUGDsoX53y8bNV9bRERERHKdQW0rMbhdJcYt2sXr07Qbr4hITqfiUA5kjOGpztU5cDKRT+dvdzuOiIiIiMgV+0fHatzTvBwjZm9l6K+b3Y4jIiKXoeJQDtWsYjHaVg3nw1lbOJmQ5HYcEREREZErYozh3zfU4pYGpXh75kZGz93mdiQREbkEFYdysCc7VeP46SRGzt7qdhQRERHJQYwxnY0xG4wxm40xT1/inBhjzHJjzBpjzKzszigCznqab/aMplOtSF6aspaJS3a5HUlERNKh4lAOVrtUIbpFR/Hx3G0cikt0O46IiIjkAMYYf2AY0AWoCfQ2xtS84JzCwAfAjdbaWsCt2Z1T5JwAfz/e612f1lWK8/Q3K1m0P9ntSCIicgEVh3K4J66vSmJyKsN+0zxtERERAaAJsNlau9VaexYYD/S44Jw7gUnW2p0A1tqD2ZxR5DzBAf581KchDcoW4cPliYycvVWLVIuI5CABbgeQy6sYHsqtDUszduFOBrSuQOki+d2OJCIiIu4qBaSdm7MbaHrBOVWBQGNMLBAG/M9a+9mFFzLGDAQGAkRGRhIbG+uNvMTHx3vt2rmJ2gHur2JJibe8+uM6Ypdv4p5aQQT5G7djuUI/Dw61g0Pt4FA7ONxoBxWHcoHHrqvCpD/28N+fN/HWrXXdjiMiIiLuSu9f0hcOwQgAGgIdgHzAAmPM79bajee9yNoRwAiARo0a2ZiYmKxPC8TGxuKta+cmagdHkP9vrEotzTszN3LKP5QRfRoSUTDE7VjZTj8PDrWDQ+3gUDs43GgHTSvLBaIK5aNvs3JMWrabTQfi3I4jIiIi7toNlEnzuDSwN51zfrLWnrLWHgZmA7rDJDmCMYZHO1Rh+N0N2XggjhuGzmXFruNuxxIR8WkqDuUSD7WrTP6gAN6esTHjk0VERCQvWwxUMcZUMMYEAXcAP1xwzvdAa2NMgDEmP860s3XZnFPksjrXLsE3D7YgwM+PWz9awHd/7HE7koiIz1JxKJcoWiCIAa0r8NOa/bqzIiIi4sOstcnAw8B0nILPRGvtGmPMIGPMIM8564CfgJXAImCUtXa1W5lFLqVGVEF+eLgl9csU5m8TlvPGtPWkpGqhahGR7KbiUC4yoHVFihYIYsj0DW5HERERERdZa3+01la11lay1r7qeW64tXZ4mnOGWGtrWmtrW2v/61pYkQwUCw3miwFNuatpWYbP2sL9ny3hZEKS27FERHyKikO5SGhwAA/FVGLu5sPM33zY7TgiIiIiIlki0N+PV2+uw8s31Wb2xkPcPGwe2w6fcjuWiIjPUHEol7m7WTlKFgrh/6ZvwFoNuRURERGRvKNPs3J83r8pR0+dpcfQuczZdMjtSCIiPkHFoVwmJNCfx66rwopdx5mx9oDbcUREREREslTzSsX44eFWRBXKxz2jFzF67jbdFBUR8TIVh3Khng1KUzG8AG9N36AF+0REREQkzylTND/fPNSC62pE8tKUtfzzm5UkJqe4HUtEJM9ScSgXCvD344nrq7HpYLy2/BQRERGRPCk0OIDhdzfk0faVmbhkN3eOXMihuES3Y4mI5EkqDuVSXWqXoHapgrz780bdRRERERGRPMnPz/D3jtUYdmcD1uw9wY1D57J6zwm3Y4mI5DkqDuVSfn6GJztVZ/exM4xbuNPtOCIiIiIiXtMtOoqvB7XAAL2Gz2fyir1uRxIRyVNUHMrF2lQpTtMKRRn622ZOJSa7HUdERERExGtqlyrE9w+3onbJQjwy7g/emr6BVK2/KSKSJVQcysWMMTzVuTqH48/yybxtbscREREREfGq8LBgvry/Kbc3KsPQ3zbzwBdLiddNUhGRa6biUC7XsFwRrqsRwUezt3L89Fm344iIiIiIeFVwgD9v9KzDf26oya/rD3LLB/PYeeS027FERHI1FYfygH90qkZ8YjIfztridhQREREREa8zxnBvywp82q8JB04mcuOwuczfctjtWCIiuZaKQ3lA9RIF6VG3JGPmbefAyQS344iIiIiIZItWVYrz/eCWhIcG0+fjRXy+YDvWah0iEZErpeJQHvH366uRkmp575dNbkcREREREck25YsXYNJDLYipGs4L36/h2W9XczY51e1YIiK5iopDeUTZYvnp3aQsExbvYseRU27HERERERHJNmEhgYzo24iHYioxbtFO7h61kCPxiW7HEhHJNVQcykMeaV+ZAH/DOzM3uh1FRERERCRb+fs5O/n+7456rNh9nBuHzmPt3pNuxxIRyRVUHMpDIgqG0K9lBX5YsZd1+/SLUERERER8T496pfhqUHNSUi09P5zPtFX73I4kIpLjqTiUxwxqU4mw4ADemr7B7SgiIiIiIq6ILl2YHx5uSfWoMB78chnvztxIaqoWqhYRuZRMFYeMMZ2NMRuMMZuNMU+nczzGGHPCGLPc8/WvzL5Wslah/IE80LYSv6w/yJLtR92OIyIiIiLiioiCIYy7vxk9G5Tmf79s4qEvl3EqMdntWCIiOVKGxSFjjD8wDOgC1AR6G2NqpnPqHGttPc/XS1f4WslC/VqWp3hoMG9O36CtPEVERETEZ4UE+vPWrdE8360GM9bup+eH89l19LTbsUREcpzMjBxqAmy21m611p4FxgM9Mnn9a3mtXKX8QQE82qEyi7YdZdbGQ27HERERERFxjTGGAa0r8km/Juw5foYew+axYMsRt2OJiOQomSkOlQJ2pXm82/PchZobY1YYY6YZY2pd4Wsli93RuCxliuZjyPQNml8tIiIiIj6vbdVwvh/cksL5A+k98nee/24VJxOS3I4lIpIjBGTiHJPOcxdWG5YB5ay18caYrsB3QJVMvtZ5E2MGAgMBIiMjiY2NzUS0KxcfH++1a+c0nUulMHLVGd6a8AtNos7/X+1L7XApagOH2sGhdnCoHRxqB4faQUTymorhoUx+uBXvzNzIJ/O2MWPNAV7qUYtOtUpgTHr/dBER8Q2ZKQ7tBsqkeVwa2Jv2BGvtyTTf/2iM+cAYUzwzr03zuhHACIBGjRrZmJiYzOS/YrGxsXjr2jlN61TLrIOzmbbH8vhtbQj0/2ugmC+1w6WoDRxqB4fawaF2cKgdHGoHEcmLCgQH8EL3mvSoV5Knv1nFoC+WcX3NSF7qUYuoQvncjici4orMTCtbDFQxxlQwxgQBdwA/pD3BGFPCeErtxpgmnuseycxrxXv8/Qz/6FiNbYdP8fXS3W7HERERERHJMaJLF+b7h1vyTJfqzNl0iOvensWYedtI0ZIMIuKDMiwOWWuTgYeB6cA6YKK1do0xZpD5//buPD6q6u7j+OfMlsmekH0DAgmBsAsBRZAgKmhVtIsLVmurj7XV2j523/c+XdXa1lp3WxV3rYqtdQtqlVXZZQlrwmICSEKAkExynj/uJIQYJCzJTZjv+/XKa+69c2f85Sckh++ce64x14dP+zSw3BizBLgduMw6OnxtV3wj0rGzizMY3TeJP76ylvrGJrfLERERERHpMfxeD1+cPJCX/3cyY/r34SfPr+STf32b97fVHvnFIiInkc7MHMJa+6K1dpC1dqC19pfhY3daa+8Mb//ZWjvUWjvSWnuqtfbtj3utdB9jDN+cVsT22nr+8c4mt8sREREREelx8vrE8ODnS/jjZaOo3LWPC/70Fr/59yp9uCoiEaNT4ZD0bhMGpjKpMJU7ysrZozsyiIiIiIh8hDGGGaNyeOXmyVw8Ooe/lq1j2m1v8NbaHW6XJiLS5RQORYhvTiviw32N3P3mBrdLERERERHpsZJjA/zuMyN55NrxGOCz987j5scXs2tvg9uliYh0GYVDEWJEbhLnDsvk3jfXs7PugNvliIiIiIj0aBMKUvn3187gxikFPLd4K1P/UMbT71ZirRasFpGTj8KhCPL1cwaxv7GJv7y+zu1SRERERER6vKDfyzemFTH7pknkp8Zy8+NLuPLe+Wzaudft0kRETiiFQxGkID2eT52Sy0NzN7G1rtntckREREREeoWizHievH4CP79oGIsrdnPOrW/w17J1NDZpTC0iJweFQxHm5nMGkRDt47cL6tmwQ594iIiIiIh0hsdjuPLUfrxy82SmFKXzm3+v4oI/vcXiit1ulyYictwUDkWYrMRoHr72VJqaLZffNVdTYkVEREREjkJmYpA7rxzD364cw+59jVx8x3/5yXMrqDsQcrs0EZFjpnAoAhVlxvOtcdHUh5q4/K65VOza53ZJIiIiIiK9yrShmbx88xlcdWo/HnxnI2ffModXVn7gdlkiIsdE4VCEyov38NA149nb0MTld8+l8kMFRCIiIiIiRyM+6OenM4bx1JcmkBD0c+3fF/LlhxdRVVvvdmkiIkdF4VAEG5aTyEPXjKdmfyMz757H1t373S5JRERERKTXOaVvMi/cNJFvTivilfermHrLHB6au4nmZt32XkR6B4VDEW54biL/uGY8H+5tYObdc9leo085RERERESOlt/r4YYpBbz0tTMYnpPID55dziV/e4e1H+xxuzQRkSNSOCSMykviwWvGsaPOCYg0DVZERERE5Njkp8by8LXj+f1nRlJeXcd5t7/JLf9ZTX1jk9uliYgclsIhAZypsA98voTttfXMvGce1XsOuF2SiIiIiEivZIzh02NyefXmyZw/IpvbXyvnvNvfZO76nW6XJiLSIYVD0mps/z7cf3UJWz7czxX3zGVnnQIiEREREZFjlRIXxa2XjuIf14wj1GS57K65fOeppdTsa3S7NBGRQygckkOMH5DCfVeXsHnXPq64Zx679ja4XZKIiIiISK82qTCNl752BtdPHsgTiyqZeksZb21pJNTU7HZpIiKAwiHpwGkDU7j3cyVs2LGXz94zj937FBCJiIiIiByP6ICX75w7mOdvnEhOcgz3LGvg7Fvf4Ol3KxUSiYjrFA5Jh04vSOXuq8ZSXl3HZ++dp6mvIiIiIiInQHF2As98aQJfGR1F0O/l5seXcNYtc3hqkUIiEXGPwiE5rDMGpfG3z45hzfY6rrpvHrX1CohERERERI6Xx2MYk+Fj9lcm8rcrxxAT8PH1J5yQ6EmFRCLiAoVD8rGmDE7njitOYeW2Wj5333z2KCASERERETkhPB7DtKGZzL5pInddOYbYKB/feGIJU2+ZwxMLKxQSiUi3UTgkR3RWcQZ/nnkKyypruPr+BdQdCLldkoiIiIjIScMYwzlDM3nhKxO5+6qxxEX5+OaTSxUSiUi3UTgknTJtaCZ/unw0iyt284X7F7CvQQGRiIiIiMiJZIzh7OIMXvjKRO65aizxQSckOvMPc3h8YQWNColEpIsoHJJOO3d4Fn+8bBQLN+3iCw8sYH9Dk9sliYiIiIicdIwxnFWcwfM3TuTez40lMdrPt55cypl/KOPxBQqJROTEUzgkR+X8Ednceuko5m/YxbV/X0B9owIiEREREZGuYIxh6pAMnrvxdO67eizJMQG+9ZQTEj22YLNCIhE5YRQOyVGbMSqH339mJG+v28n//H2hAiIRERERkS5kjOHMwRn88wYnJOoTE+DbTy1jyu/LeHS+QiIROX4Kh+SYfPKUXH77qRG8Vb6D6x9axIGQAiIRERERka7UEhI9e8Pp3H91CSmxAb7ztBMSzZq/mYaQQiIROTYKh+SYfWZsHv938XDKVlfz5Yfe1S8jEREREZFuYIxhyuB0JyT6fAkpcVF8VyGRiBwHhUNyXC4b15dfXjyMV1dVccMj72pKq4iIiIhINzHGMKUonWe/PIEHPl9CWvzBkOiReQqJRKTzFA7JcbtifD9+NmMoL6/8gJtmvaeASERERESkGxljKC1K55kvT+DBL4wjLT6K7z3jhEQPz9ukkEhEjkjhkJwQV53Wnx+dX8y/lm/na48tJqSASERERESkWxljmDworTUkSk+I4vvPLKf0d6/z0NxNWidURA7L53YBcvL4wsR8mq3lF7Pfx+cx3HLJKLwe43ZZIiIiIiIRpSUkOqMwlTfX7uC2V9bwg2eXc8fr5XxpSgGXjM0lyud1u0wR6UEUDskJde2kAYSaLb/+1yq8xvC7z4xUQCQiIiIi4gJjDGcMSmNSYSpvle/gtlfW8sNwSPTl0oFcUpKnkEhEAIVD0gWunzyQpmbL715ajddj+M2nRuBRQCQiIiIi4gpjDJMK05hYkMp/y3dy2ytr+OE/V3DbK2v59NhcZo7rS7+UWLfLFBEXKRySLnHDlAJCTZZbX1mD12P41cXDFRCJiIiIiLjIGMPEwlROL0jhnXU7efCdjdzz5gb+Nmc9kwpTuWJ8X6YOycDv1dK0IpFG4ZB0ma+eVUhTczO3v1aO12P4xUXDMEYBkYiIyPEyxkwH/gh4gXustb8+zHklwFzgUmvtk91Yooj0YMYYJhSkMqEglQ9q63lsQQWPzt/M9Q+9S1p8FJeV5HFpSR65yTFulyoi3UThkHSp/z17EKFmyx1l6/B5DD+5cKgCIhERkeNgjPECfwHOBiqBBcaY56y1Kzs47zfAS91fpYj0FhkJQW6aWsgNUwooW13Fw/M28+fXy/nz6+VMKUpn5ri+TBmcrnVERU5yCoekSxlj+Oa0IpqaLX97Yz0ej+FH5xcrIBIRETl244Bya+16AGPMo8AMYGW7874CPAWUdG95ItIbeT2GqUMymDokg8oP9/HYggoeW1DBtX9fSHZikEtL+nJpSR6ZiUG3SxWRLqBwSLqcMYbvnDuYULPl3rc24PMYvnfeEAVEIiIixyYHqGizXwmMb3uCMSYHuBg4E4VDInKUcpNj+Po5Rdw0tZBX36/i4XmbuPWVNdz+2lqmDk7nilP7MakgVWuKipxEFA5JtzDG8INPDKGp2XL3mxvwejx8e3qRAiIREZGj19EvT9tu/zbg29bapo/7XWuMuQ64DiAjI4OysrITVOKh6urquuy9exP1waE+OHpLH4LANQPhgqxoyipCvFn+Af9Z+QFp0YbJeT4m5fhJjDr2MX1v6UNXUx8c6oPDjT4oHJJuY4zhxxcUE2pu5s4561iwcRc/uWAow3MT3S5NRESkN6kE8trs5wJb250zFng0HAylAucZY0LW2mfbnmStvQu4C2Ds2LG2tLS0SwouKyujq967N1EfHOqDozf24RLgQKiJ/6z4gIfnbeLJNbv457oQ5wzN5IpxfTltYMpRf/jbG/vQFdQHh/rgcKMPCoekWxlj+PmMYYzISeK3L63mwr+8xWfG5PKNaUWkx+v6ZRERkU5YABQaY/KBLcBlwMy2J1hr81u2jTEPAC+0D4ZERI5FlM/LBSOzuWBkNuVVdcyav5mn3q1k9tJtDEiN5fJxffn0mFySYwNulyoiR8HjdgESeYwxXFKSx+vfmMx1kwbwzHtbOPP3c7jrjXU0hJrdLk9ERKRHs9aGgBtx7kL2PvC4tXaFMeZ6Y8z17lYnIpGkID2OH55fzNzvTuXWS0fSJzbAL198n/H/9ypfe/Q95m/YhbXtr3oVkZ5IM4fENfFBP989bwiXjevLL15Yya9eXMWs+RX84BNDOHNwutYjEhEROQxr7YvAi+2O3XmYc6/ujppEJHIF/V4uHp3LxaNzWb19D4/M28TT727h2cVbKUyPY+b4vnzylFwSo/1ulyoih6GZQ+K6/NRY7r26hAc+X4LHwDUPLuRz9y+gvGqP26WJiIiIiMhRKMqM56czhjHv+1P57adGEBPw8tPnVzL+V6/wzSeW8N7mDzWbSKQH6tTMIWPMdOCPgBe4x1r763bPXwF8O7xbB3zJWrsk/NxGYA/QBISstWNPTOlysiktSuf0glT+8Y5zq8zpt73JVaf156tnFepTBhERERGRXiQm4OOSkjwuKclj+ZYaHpm/mX++t4UnFlVSnJXAzPF9uWh0jttlikjYEWcOGWO8wF+Ac4Fi4HJjTHG70zYAk621I4CfE77rRRtTrLWjFAzJkfi9Hr4wMZ+yb5RySUke97+9gSm/L+PheZtoatYnDCIiIiIivc2wnER+dfFw5n3/LH558TAAfvDscsb98hXuXnqA11dVae1REZd1ZubQOKDcWrsewBjzKDADWNlygrX27Tbnz8W5parIMUuJi+JXFw/nivF9+enzK/n+M8t5aO5mfnxBMacOSHG7PBEREREROUpxUT6uGN+PmeP6sqSyhkfmbeL5xZV8/oEFJAR9nDM0k0+MyOL0gakEfFoBRaQ7dSYcygEq2uxXAuM/5vxrgH+12bfAf4wxFvibtbb9rCIAjDHXAdcBZGRkUFZW1onSjl5dXV2XvXdv0pv68KVBlrGJUTy2ag+X3TWXkkwvlxYFSI0+vl8YvakHXUl9cKgPDvXBoT441AcREekKxhhG5SUxKi+Js/vswptdzOyl23lpxXaeXFRJQtDHtKGZnKegSKTbdCYc6uiWUR1e32OMmYITDk1sc/h0a+1WY0w68LIxZpW19o2PvKETGt0FMHbsWFtaWtqJ0o5eWVkZXfXevUlv68MU4KbGJu56Yz13lJWz9L8H+OLkgVw/eQAxgWO76V5v60FXUR8c6oNDfXCoDw71QUREuprfYygdnMGZgzM4EBrGW2t3MHvZNv69fDtPLKokMdrPtKEZnDc8i9MLUvF7FRSJdIXO/Ku6Eshrs58LbG1/kjFmBHAPcK61dmfLcWvt1vBjlTHmGZzL1D4SDokcSdDv5aaphXx6TC6//tcqbn91LU8srOA75w7mwpHZGNNRjikiIiIiIr1BlM/L1CEZTB2SwYFQkxMULd3Gv5Zt5/GFlSTF+DmnOINPjMhmwsAUBUUiJ1BnwqEFQKExJh/YAlwGzGx7gjGmL/A0cKW1dk2b47GAx1q7J7x9DvCzE1W8RKbspGhuv3w0V57Wj58+v4KvPrqYf7yziR9fMJThuYlulyciIiIiIsepfVD05hpnRtGLbYKiacXOpWcKikSO3xHDIWttyBhzI/ASzq3s77PWrjDGXB9+/k7gR0AKcEd49kbLLeszgGfCx3zAI9baf3fJdyIRp6R/H/55w0SeXFTB715azYV/eYtLxuTxjWlFpMVHuV2eiIiIiIicAFE+L2cVZ3BWcQb1jU28uXYHLy7bxuxl23hsYQVJMX6mD83kvOFZnKagSOSYdGqxFmvti8CL7Y7d2Wb7WuDaDl63Hhh5nDWKHJbXY7i0pC/nDs/iz6+Vc/9/NzB72TZumlrA1RPytXidiIiIiMhJJOj3cnZxBmeHg6I31lTz4rJtPL9kK48uqCA5xs+08F3PThuQgk9BkUinHNtKviI9TELQz/fOG8JlJXn8Yvb7/OrFVcyaX8EPzx/ClKJ0rUckIiIiInKSCfq9nDM0k3OGZrYGRbPbBUXTh4VnFCkoEvlYCofkpDIgLY77ri7h9dVV/PyFlXzhgYVMHpTGD88vpiA9zu3yRERERESkC7QPiuaEZxQ9t3grs+ZX0Cc24MwoGp7FqQP6KCgSaUfhkJyUphSlM7EglQff3sgfX13L9Nve4KrT+vPVswpJjPa7XZ6IiIiIiHSRoN/LtKGZTAsHRWWrnaDon4u3MGv+5tag6PwRWYzPV1AkAgqH5CTm93q4dtIALhqdwx/+s4b7397As4u38I1ziri0JM/t8kREREREpIsF/V6mD8tk+rCDQdHsNkFRSmyAMwenM7kojUkFaSTG6INkiUwKh+SklxoXxf99cjhXjO/Lz55fyfeeWcY/5m7iwtwmSt0uTkREREREusVHg6IqZi/bzksrtvPEoko8Bkb3TWbyoDRKi9IYlp2Ix6O1SyUyKBySiDEsJ5HHvngqs5dt41ez3+c3C+qZvfVNZo7rx4WjsomL0l8HEREREZFI4ARFWUwflkWoqZkllTXMWV3FnDXV3PrKGm55eQ19YgOcUZhKaVE6kwpTSYmLcrtskS6jfw1LRDHGcP6IbKYOzuCXs15j4YeW7z2zjF/OXsmM0TnMHNeXYTmJbpcpIiIiIiLdxOf1MKZfMmP6JXPzOUXsrDvAm2t3MGdNNW+sqebZxVsxBkbkJDJ5UBqTi9IYlZeMV7OK5CSicEgiUnTAy1n9/Pz8qkm8u3k3j8zbzFOLKnlk3mZG5iYyc3xfLhiZTUxAf0VERERERCJJSlwUF43O4aLROTQ3W5ZvrWHO6mrK1lTz59fLuf21chKj/UwsTKV0UBqTB6WRnhB0u2yR46J/+UpEM8a0fkrwo/OLefo9JyD69lPL+PkL73PR6GxmjutHcXaC26WKiIiIiEg383gMI3KTGJGbxFemFlKzr5E3y6uZs7qaOWuqmb10GwBDshIoLXKCojH9kvHrDmjSyygcEglLjPHz+dPzuXpCfxZt+pBH5m3m8YWVPDR3M6P7JnH5uL5cMCKb6IDX7VJFRERERMQFiTF+zh+RzfkjsrHW8v62PcxZU82cNVXc/cZ6/lq2jrgoH6cXpDB5kHMXtJykaLfLFjkihUMi7RhjGNu/D2P79+FHFxTz1LtbeGTeJr715FJ+/sJKPjk6h5nj+1GUGe92qSIiIiIi4hJjDMXZCRRnJ/Cl0oHsqW/k7XU7nbBodTUvrfgAgML0uPAd0NIpyU8myqcPm6XnUTgk8jGSYgJcMzGfL5zen/kbdvHI/M3Mml/Bg+9sYky/ZGaO68snRmQR9OsHvIiIiIhIJIsP+pk2NJNpQzOx1rKuuo6y8OVnf39nE/e8tYFov5fTBqa0XoLWLyXW7bJFAIVDIp1ijGH8gBTGD0jhxxc08NSiSmbN38zXn1jCz15YySdPce50Vpih2UQiIiIiIpHOGENBejwF6fFcO2kA+xpCzF2/s3WtotdWVQHQPyWG0qJ04veHGLG3gT6xAZcrl0ilcEjkKPWJDfA/Zwzg2kn5vLN+J4/M28xDczdx/383UtI/mZnj+3LuMM0mEhERERERR0zAx5mDMzhzcAYAG3fsDa9VVM2jCzZT39jMn957mYL0OMbl92Fc/z6U5PfRekXSbRQOiRwjYwwTBqYyYWAqO+sO8GR4NtH/PraEnz6/kk+dksvl4/pSkB7ndqkiIiIiItKD9E+NpX9qLJ+b0J8DoSb+/nwZjcn9WLBhF88v3soj8zYDkJMUzbj8PpT078O4/GQGpsVhjHG3eDkpKRwSOQFS4qL44uSB/M+kAa2ziR58eyP3vrWB8fl9mDm+L9OHZWrxOREREREROUSUz0thspfS0gIohaZmy6rttSzYsIv5G3fx5todPPPeFsC5iqGkf3I4LOpDcVYCPq/H3W9ATgoKh0ROII/HcHpBKqcXpFK95wBPLKpg1vzNfPXRxfSJDfDpMc5sovxULTwnIiIiIiIf5fUYhmYnMjQ7katPz8day8ad+1iwYRfzNuxiwcZdrXdCiw14OaVfcutlaKPykrS8hRwThUMiXSQtPoovlxZw/RkDeat8B4/M28y9b23grjfWM2FgCpeP68u0oZkEfEr6RURERESkY8YY8lNjyU+N5ZKSPAC219SzYOMu5ofDoj+8vAaAgNfDiNxESsLrFo3pn0xC0O9m+dJLKBwS6WIej+GMQWmcMSiNqtp6Hl9Ywaz5FXxl1nskBH2MyE1iWE4iw3ISGJ6TSN8+MbqOWEREREREDiszMcgFI7O5YGQ2ALv3NbBw44dOYLRxF3e/sZ6/lq3DGBiSmdC6blFJfjLp8UGXq5eeSOGQSDdKTwhy45mFfKm0gDfXVvPv5dtZtqWGe99aT2OTBSA+6GNYthMWOaFRIvkpsXg8CoxEREREROSjkmICnFWcwVnFzt3Q9jWEWLx5N/M3OjOLHltQwQNvbwQgPzX2kHWL9OG0gMIhEVd4PYbSonRKi9IBOBBqYs32OpZvrWHZlhpWbKnhwXc20RBqBpxriYdmJzI0PLtoWE4iA9Pi8CowEhERERGRdmICPiYUpDKhIBWAxqZmlm+pCV+K9iEvrfiAxxdWApCREEVJ/z6M7pvMyFxnraPogNYtijQKh0R6gCifl+G5iQzPTeTy8LHGpmbWfuAERsu3OF+z5m/m/kYnMIr2exmSFc/wnESG5iQyPCeRgvQ4/LpbgYiIiIiItOH3ehjdN5nRfZO57gxobrasrapzZhaF1y16Yek2wPkge1BGPCNzExmRm8SI3ESKMuP174yTnMIhkR7K7/VQnJ1AcXYCl4x1Fp4LNTWzfsdelm9pmWFUy5OLKnnwnU0ABHwehmQlMCz74AyjQRnxWvRaRERERERaeTyGosx4ijLjufLUfgBU1dazpLKGJRW7WVK5m38t386jCyoAiPJ5GJqdwIjcJEbmOaGRlr44uSgcEulFfF4PgzLiGZQRzydPyQWc1H/Dzr2ts4uWb6nluSVbeXjeZgD8XucHv7OOkfM1ODNet7gUEREREZFW6QlBzi4OcnZ43SJrLZt37WsNjJZW7j5k7aL4oI8R4dlFLbOMshKDWr+ol1I4JNLLeTyGgWlxDEyLY8aoHODgD/LlW2qdGUZba/j3ioPJv9djKEyPY3hOIr69jdQv3052UpDspGhSYgP6gS4iIiIiEuGMMfRLiaVfSiwXhu+KFmpqpry6jqUVNSyudAKju99YT6jZublOalwUo/IOXo42MjeJ5NiAm9+GdJLCIZGTUNsf5J8YkQU4gdGW3ftZvqXWmWG0tYbXV1exo66BWasWtb424POQlRgkOzGarCTnMTup7XaQ+KDfrW9NRERERERc4vN6GJyZwODMBC4pcZa+qG9s4v1ttSytrGFJ5W6WVOzm1VVVWCcvIq9PNCNykxgVDoyG5SQSG6UooqfR/xGRCGGMITc5htzkGKYPywScwOiFl8vIH3oKW3fvZ1tNPVtr9rN1dz3bdu9n3vpdbK+tpyn8SUCL+CgfWUlBssLBUXZikKwkJzjKTowmMzGoy9ZERERERCJA0O9tXey6xZ76RpZtqWFpZQ1LK3ezePNuZocXvPYYKEiPC69f5FySNjgzQeukukzhkEgEM8YQHzCtaxF1JNTUTHXdAbburg8HSPvbbNezfEsNO/c2fOR1qXEBshKjnVlI4eCoNUxKCpIeH8SrBexERERERE468UE/EwamMmFgauuxHXUHWFq5myUVTmD02qoqnlxUCUDA62FIVjxJHGBTYCNFmfEMzownKUaXpHUXhUMi8rF8Xk845IlmTL/kDs+pb2xie40TGG2tcWYdtcxA2rhzL2+v20ndgdAhr/F6DJkJQbISg2QmBkmNiyItPorUuAApsVGkhrdT46I0C0lEREREpJdLjYvizMEZnDn44ILXlR/ub51dtLSyhoUVIeY8t6L1NVmJwXBQlMDgzHgGZ8UzIDVOs4y6gMIhETluQb+X/qmx9E+NPew5tfWNbNvdEiDtP2R7xdZaduw5wJ52AVKL+CjfIWFRSvix5Sst/uC+rl8WEREREen5jDHk9Ykhr09M6zqpr7/+OsVjTuP9bbWs3r6HVeGv/5avp7HJWerC73VuyDM4M56izAQGZ8UzJDOBjIQo3VjnOOhfUSLSLRKCfhIy/RRlxh/2nPrGJnbUHWBHXQM79hxgR90Bdu5toDq8vaPuAGur6nhn/QF272vs8D2i/d5DwqO2wZHzFSAlLoq0uCgSon36BSIiIiIi0kMYY8hICJKREKS0KL31eGNTM+ur97Jqey2rtu9h9fY9zN+wi2cXb209JzHa78wuyoxncFYCRZnxFGXE68PjTlKXRKTHCPq9rYtmH0ljUzM76xpaQ6MdLdt7Du5XfriPxRUfsmtvA+3W1Aaca5tbgiQO1PP0tvdIiPaRGO0nIeh3HqP9h+wnRvuJC/q0XpKIiIiISDfxez1O2JMZz4w2x2v2NbL6gz2todGqbbU8uaiSvQ1Nref0S4mhKMMJjFrCo34psRrPt6NwSER6Jb/XQ2Z4vaIjaWq2fLivJTw6GChV1x1oDZg27bEsrdxNzf5GautDH7lDW3vxQV+bAMn3kQCpNVTqIGzSGkoiIiIiIscvMcbPuPw+jMvv03qsudmyZff+1rDIuTStllfe/6D1A+Og38OgjPjWS9OGhGcb9YmN3AWwFQ6JyEnP6zGtl5WR2fE5ZWVllJaWAs7ieHsbmqjd3+iERS2P9aFD9/c3UlvvbG/csS98TiP72nxS0ZGAzxMOjHwfmZ0UH/SREO1vDZ8O2Q4/F+Xz6HI4EREREZEOeDwH1zI6uzij9Xh9YxPlVXWHrGf02qoqHl9Y2XpOWnwUhelxDEyLY2BaLAPD21mJwZN+/K1wSESkHWMMcVE+4qJ8ZCdFH/XrG0LN7KnvOFBqCZBq9zdSu995btfeBjbs2Evt/kb21IcIHWHWkt9rOgySDj46M5biw4FSy74TMOmyOBERERGJPEG/l2E5iQzLSTzkePWeA+GwyAmNyqvreHbxFvbUH7xZTkzAy4C02HBoFP5Kj6V/SuxJc1WAwiERkRMs4POQEhdFSlzUUb/WWsv+xib21IdaZybVhrf31IeorW9sfa7tflVtXev+kWYuAcRF+drMTjoYJO3ecYCy2hUEfB4CXg8Bnwd/+DHg8xDVZjvg9eBvc15Um+NtXxcVPuZRICUiIiIiPUxafBRp8VFMLExtPWatZUddA+uq65yvqr2sq65j0aYP+WebRbCNgbzkGGeWUVocA9PjKAjPNuptl6gpHBIR6UGMMcQEfMQEfGQkHHk9pY40NjVT1y5Iqj1ssOTMYKraU8/aqkZq6pp4d0clDU3NNISaO1zI+1j5PKY1WPJ7ncCoNVBqEyoF/V6i/V6Cfi8xAS/RAWc/OuDstx73HzweHfAS4/cRDHiICfiI9ns1O0pEREREjokxpjU0OnVAyiHP7W9oYsOOvZRX17GuKhweVe/l7XU7ORBqbj0vOcZ/yCyjlu3c5Gh8Xk93f0tHpHBIROQk4/d6SI4NkHwMn1a0XXsJINTU3BoUtT623w4d4ZzDnXuYc2rrG9nf0ER9YzP7GkLsb3S2j1bA52kNkA4JlToImw49x8e6ykaqFlYc9X/zRPIag89rCHg9+LwefF6D3xN+9Bp8rdsefB7n0d/uvJZtzdoSEREROTGiA16KsxMozk445HjLQtgtYdG66jrKq+p4ddUHPLawofW8gNdD/9SYjwRHA9LiiItyL6JROCQiIoflCwcTMS7Pim1uttSHmtjX0MT+hib2N370cV/rfoj9Dc0Ht8PP1Ycf6w6EqN5z4OBrG5rY19j00TvULV/qzjfbBTzG+X8ZCIdHPo/HCZjaBkmtx5xHv9dDQaCRUreLFxEREekF2i6EXVp06HO79zW0BkYtl6mt3r6H/6z84JAxaGZCkIHpsUxNa+r2MZjCIRER6fE8noOX23WVhlBza2D01ttvc+qpp3bZf+tIrIVma2lssjQ2NRNqsjQ2O4+hpmYam8OPTZZQ+HhD+LxQc/h4UzOh5kNf3xg6/PMtx1v260IhGjVKEBERETluSTEBxvQLMKZf8iHHG0LNbNq59+Bso/Blal4XJn1r2CciIgKtax8lRvtJjfaQmxzjdkmuKysrc7sEERERkZNWwOehMCOewoz4Q467MQbreasgiYiIiIiIiIhIt+lUOGSMmW6MWW2MKTfGfKeD540x5vbw80uNMad09rUiIiIiIiIiIuKeI4ZDxhgv8BfgXKAYuNwYU9zutHOBwvDXdcBfj+K1IiIiIiIiIiLiks7MHBoHlFtr11trG4BHgRntzpkB/N065gJJxpisTr5WRERERERERERc0plwKAeoaLNfGT7WmXM681oREREREREREXFJZ+5W1tFN1Gwnz+nMa503MOY6nEvSyMjI6LLVuevq6nT3FdQHUA9aqA8O9cGhPjjUB4f6ICIiIhIZOhMOVQJ5bfZzga2dPCfQidcCYK29C7gLYOzYsba0tLQTpR29srIyuuq9exP1QT1ooT441AeH+uBQHxzqg4iIiEhk6MxlZQuAQmNMvjEmAFwGPNfunOeAq8J3LTsVqLHWbuvka0VERERERERExCVHnDlkrQ0ZY24EXgK8wH3W2hXGmOvDz98JvAicB5QD+4DPf9xru+Q7ERERERERERGRo9aZy8qw1r6IEwC1PXZnm20L3NDZ14qIiIjIsTPGTAf+iPPh2z3W2l+3e/4K4Nvh3TrgS9baJd1bpYiIiPQWnbmsTERERER6CGOMF/gLcC5QDFxujClud9oGYLK1dgTwc8LrOoqIiIh0ROGQiIiISO8yDii31q631jYAjwIz2p5grX3bWvtheHcuzk1BRERERDqkcEhERESkd8kBKtrsV4aPHc41wL+6tCIRERHp1Tq15pCIiIiI9Bimg2O2wxONmYITDk08zPPXAdcBZGRkUFZWdoJKPFRdXV2XvXdvoj441AeH+uBQHxzqg0N9cLjRB4VDIiIiIr1LJZDXZj8X2Nr+JGPMCOAe4Fxr7c6O3shaexfh9YjGjh1rS0tLT3ixAGVlZXTVe/cm6oNDfXCoDw71waE+ONQHhxt90GVlIiIiIr3LAqDQGJNvjAkAlwHPtT3BGNMXeBq40lq7xoUaRUREpBfRzCERERGRXsRaGzLG3Ai8hHMr+/ustSuMMdeHn78T+BGQAtxhjAEIWWvHulWziIiI9GwKh0RERER6GWvti8CL7Y7d2Wb7WuDa7q5LREREeidjbYfrF7rKGFMNbOqit08FdnTRe/cm6oN60EJ9cKgPDvXBoT44urIP/ay1aV303nIMNP7qFuqDQ31wqA8O9cGhPjjUB0e3j8F6ZDjUlYwxCzWtWn0A9aCF+uBQHxzqg0N9cKgPcqLoz5JDfXCoDw71waE+ONQHh/rgcKMPWpBaRERERERERCSCKRwSEREREREREYlgkRgO3eV2AT2E+qAetFAfHOqDQ31wqA8O9UFOFP1ZcqgPDvXBoT441AeH+uBQHxzd3oeIW3NIREREREREREQOisSZQyIiIiIiIiIiEhYx4ZAxZroxZrUxptwY8x2363GDMSbPGPO6MeZ9Y8wKY8xX3a7JTcYYrzHmPWPMC27X4hZjTJIx5kljzKrwn4vT3K7JDcaY/w3/nVhujJlljAm6XVN3MMbcZ4ypMsYsb3OsjzHmZWPM2vBjsps1dofD9OF34b8XS40xzxhjklwssVt01Ic2z33DGGONMalu1Ca9m8ZgGoO1pfGXQ2OwyB1/gcZgoPFXi540/oqIcMgY4wX+ApwLFAOXG2OK3a3KFSHg69baIcCpwA0R2ocWXwXed7sIl/0R+Le1djAwkgjshzEmB7gJGGutHQZ4gcvcrarbPABMb3fsO8Cr1tpC4NXw/snuAT7ah5eBYdbaEcAa4LvdXZQLHuCjfcAYkwecDWzu7oKk99MYrJXGYAdp/OWI6DFYhI+/QGMw0PirxQP0kPFXRIRDwDig3Fq73lrbADwKzHC5pm5nrd1mrX03vL0H55dQjrtVucMYkwt8ArjH7VrcYoxJAM4A7gWw1jZYa3e7WpR7fEC0McYHxABbXa6nW1hr3wB2tTs8A3gwvP0gcFF31uSGjvpgrf2PtTYU3p0L5HZ7Yd3sMH8eAG4FvgVokUI5FhqDoTFYC42/HBqDtYrI8RdoDAYaf7XoSeOvSAmHcoCKNvuVROAv5LaMMf2B0cA8l0txy204f9maXa7DTQOAauD+8PTue4wxsW4X1d2stVuA3+Ok8tuAGmvtf9ytylUZ1tpt4PxjBkh3uZ6e4AvAv9wuwg3GmAuBLdbaJW7XIr2WxmDtRPgY7DY0/gKNwTT+6pjGYIfS+Kubx1+REg6ZDo5F7Cegxpg44Cnga9baWrfr6W7GmPOBKmvtIrdrcZkPOAX4q7V2NLCXk3/66keEr+eeAeQD2UCsMeaz7lYlPYUx5vs4l4M87HYt3c0YEwN8H/iR27VIr6YxWBuRPAbT+OsQET8G0/hLPo7GX+6MvyIlHKoE8trs5xJB0xbbMsb4cQYlD1trn3a7HpecDlxojNmIM739TGPMQ+6W5IpKoNJa2/LJ5ZM4A5VIcxawwVpbba1tBJ4GJrhck5s+MMZkAYQfq1yuxzXGmM8B5wNXWGsj8R+zA3EG7UvCPy9zgXeNMZmuViW9jcZgYRqDafzVhsZgGn91RGMwNP7CxfFXpIRDC4BCY0y+MSaAs9jZcy7X1O2MMQbn2ub3rbW3uF2PW6y137XW5lpr++P8WXjNWhtxn1RYa7cDFcaYovChqcBKF0tyy2bgVGNMTPjvyFQibFHIdp4DPhfe/hzwTxdrcY0xZjrwbeBCa+0+t+txg7V2mbU23VrbP/zzshI4JfyzQ6SzNAZDYzDQ+KstjcEAjb86EvFjMI2/3B1/RUQ4FF7U6kbgJZwfOo9ba1e4W5UrTgeuxPmkZnH46zy3ixJXfQV42BizFBgF/Mrdcrpf+FO7J4F3gWU4PxfvcrWobmKMmQW8AxQZYyqNMdcAvwbONsasxblDwq/drLE7HKYPfwbigZfDPyvvdLXIbnCYPogcF43BWmkMJu1F9BgsksdfoDEYaPzVoieNv0xkztQSERERERERERGIkJlDIiIiIiIiIiLSMYVDIiIiIiIiIiIRTOGQiIiIiIiIiEgEUzgkIiIiIiIiIhLBFA6JiIiIiIiIiEQwhUMiIiIiIiIiIhFM4ZCIiIiIiIiISARTOCQiIiIiIiIiEsH+H1tNQGGxv51TAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 1440x576 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Función de perdida\n",
    "plot_metric(history=history, name='loss')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x576 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Accuracy\n",
    "plot_metric(history=history, name='accuracy')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<hr>\n",
    "\n",
    "\n",
    "## Evaluamos el (mejor) modelo\n",
    "\n",
    "* Pasamos a evaluar el modelo con los datos de test, calculando las siguientes métricas de evaluación:\n",
    "\n",
    "    1. **Accuracy**\n",
    "    2. **Precision**\n",
    "    3. **Recall**\n",
    "    4. **F1**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Importamos el mejor modelo creado"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.models import load_model\n",
    "\n",
    "model = load_model(model_path)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Obtenemos las métricas"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.7024\n",
      "F1: 0.6994\n",
      "Precision: 0.7061\n",
      "Recall: 0.7024\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "  ciudadanos       0.62      0.65      0.64       127\n",
      "     podemos       0.79      0.67      0.73       227\n",
      "          pp       0.65      0.48      0.55       148\n",
      "        psoe       0.74      0.78      0.76       320\n",
      "         vox       0.64      0.80      0.71       213\n",
      "\n",
      "    accuracy                           0.70      1035\n",
      "   macro avg       0.69      0.68      0.68      1035\n",
      "weighted avg       0.71      0.70      0.70      1035\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import accuracy_score, f1_score, precision_score, recall_score, confusion_matrix, classification_report\n",
    "\n",
    "y_true = encoder.inverse_transform(y_test.reshape(-1))\n",
    "y_pred = np.argmax(model.predict(pad_corpus_test), axis=-1)\n",
    "y_pred = encoder.inverse_transform(y_pred.reshape(-1))\n",
    "\n",
    "print('Accuracy: {acc:0.4f}'.format(acc=accuracy_score(y_true=y_true, y_pred=y_pred)))\n",
    "print('F1: {f1:0.4f}'.format(f1=f1_score(y_true=y_true, y_pred=y_pred, average='weighted')))\n",
    "print('Precision: {pre:0.4f}'.format(pre=precision_score(y_true=y_true, y_pred=y_pred, average='weighted')))\n",
    "print('Recall: {rec:0.4f}'.format(rec=recall_score(y_true=y_true, y_pred=y_pred, average='weighted')))\n",
    "print(classification_report(y_true=y_true, y_pred=y_pred))\n",
    "labels = ['psoe', 'pp', 'vox', 'podemos', 'ciudadanos']\n",
    "confusion_matrix_test = confusion_matrix(y_true=y_true,\n",
    "                                         y_pred=y_pred,\n",
    "                                         labels=labels)\n",
    "confusion_matrix_train = confusion_matrix(y_true=encoder.inverse_transform(y_train.reshape(-1)), \n",
    "                                          y_pred=encoder.inverse_transform(np.argmax(model.predict(pad_corpus_train), axis=-1)), \n",
    "                                          labels=labels)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Pintamos la matriz de confusión"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x432 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import itertools\n",
    "%matplotlib inline\n",
    "\n",
    "def plot_confusion_matrix(cm, classes, title, cmap=plt.cm.Blues):\n",
    "    \"\"\"\n",
    "    This function prints and plots the confusion matrix.\n",
    "    \"\"\"\n",
    "    plt.imshow(cm, interpolation='nearest', cmap=cmap)\n",
    "    plt.title(title)\n",
    "    plt.colorbar()\n",
    "    tick_marks = np.arange(len(classes))\n",
    "    plt.xticks(tick_marks, classes, rotation=45)\n",
    "    plt.yticks(tick_marks, classes)\n",
    "\n",
    "    thresh = cm.max() / 2.\n",
    "    for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):\n",
    "        plt.text(j, i, format(cm[i, j], 'd'), horizontalalignment=\"center\",\n",
    "                 color=\"white\" if cm[i, j] > thresh else \"black\")\n",
    "\n",
    "    plt.tight_layout()\n",
    "    plt.ylabel('True label')\n",
    "    plt.xlabel('Predicted label')\n",
    "\n",
    "\n",
    "# Pintamos las matrices de confusión\n",
    "plt.figure(figsize=(16, 6))\n",
    "plt.subplot(1, 2, 1)\n",
    "plot_confusion_matrix(confusion_matrix_train, classes=labels, title='Matriz de Confusión Datos Entrenamiento')\n",
    "plt.subplot(1, 2, 2)\n",
    "plot_confusion_matrix(confusion_matrix_test, classes=labels, title='Matriz de Confusión  Datos Test')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<hr>\n",
    "\n",
    "\n",
    "# Bonus Track I - LSTM - \n",
    "\n",
    "\n",
    "* Las ***Redes Neuronales Recurrentes*** (RNN) constituyen una herramienta muy apropiada para ***modelar series temporales y para el Procesamiento del Lenguaje Natural***.\n",
    "\n",
    "\n",
    "* Las RNN son, en principio, un modelo computacional más potente que las clásicas redes neuronales (Perceptrón). \n",
    "\n",
    "\n",
    "* Esta mayor potencia proviene del hecho de que las RNN son capaces de procesar secuencias temporales gracias a la posibilidad de ***recordar parcialmente la historia relevante*** de la secuencia por medio de una representación en forma de estado (***implementa una cierta memoria***).\n",
    "\n",
    "\n",
    "* Esta memoria no existe de manera natural en las redes no recurrentes, utilizadas principalmente para el procesamiento de datos estáticos.\n",
    "\n",
    "\n",
    "* Las LSTM (Long Short-Term Memory) son un tipo especial de unidades de memoria que conforman una red neuronal y que tienen la capacidad de recordar aquella información relevante que va aprendiendo durante el proceso de entrenamiento y olvidar aquella información que no considera relevante.\n",
    "\n",
    "\n",
    "* La arquitectura de una unidad LSTM es la siguiente:\n",
    "\n",
    "\n",
    "<img src=\"./imgs/028_lstm.png\" style=\"width: 400px;\"/>\n",
    "\n",
    "\n",
    "* Las LSTM está formadas por los siguientes elementos:\n",
    "<span></span><br><br>\n",
    "    + ***$f_t$ -> forget gate*** (Puerta de Olvido): Que parte de la información del pasado no es importante.\n",
    "<span></span><br><br>\n",
    "    + ***$i_t$ -> input gate*** (Puerta de Entrada): De la información actual, hay algo que sea importante para recordar.\n",
    "<span></span><br><br>\n",
    "    + ***$\\widetilde{C}_t$ -> candidate gate*** (Puerta Candidata): Si hay algo importante que recordar, cómo lo combinamos con la información del pasado.\n",
    "<span></span><br><br>\n",
    "    + ***$o_t$ -> output gate*** (Puerta de Salida): Que información es importante a corto plazo.\n",
    "<span></span><br><br>    \n",
    "    + ***$C_t$ -> state cell*** (Estado de la Celda): Guarda el estado de la celda actual.\n",
    "    \n",
    "    \n",
    "<img src=\"./imgs/029_lstm.png\" style=\"width: 600px;\"/>\n",
    "    \n",
    "    \n",
    "* Siendo:\n",
    "    \n",
    "$$f_t = sigmoid(W_f \\cdot [h_{t-1}, X_t])$$\n",
    "\n",
    "$$i_t = sigmoid(W_i \\cdot [h_{t-1}, X_t])$$\n",
    "\n",
    "$$\\widetilde{C}_t = tanh(W_c \\cdot [h_{t-1}, X_t])$$\n",
    "\n",
    "$$o_t = sigmoid(W_o \\cdot [h_{t-1}, X_t])$$\n",
    "\n",
    "$$C_t = f_t * C_{t-1} + i_t * \\widetilde{C}_t$$\n",
    "\n",
    "$$h_t = o_t * tanh(C_t)$$\n",
    "\n",
    "\n",
    "* Una capa con celdas LSTM estaría formada por cerlda LSTM conectadas en serie unas con otras de la siguiente manera:\n",
    "\n",
    "\n",
    "<img src=\"./imgs/030_lstm.png\" style=\"width: 700px;\"/>\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<hr>\n",
    "\n",
    "# Bonus Track II - Embedding Layers - \n",
    "\n",
    "\n",
    "* Las \"Embedding Layes\" (capas de incrustación) son un tipo de técnicas que nos permiten representar las palabras y los documentos como un vector de características latentes.\n",
    "\n",
    "\n",
    "* Para el caso particular de las palabras de un corpus, los vectores de características latentes que representan las palabras, son calculadas basándose en la ***hipótesis distribucional de Zellig Harris*** que dice: \"*elementos lingüísticos con distribuciones similares tienen significados similares*\"; es decir que ***las palabras que tienen un contexto similar, tendrán significados similares***.\n",
    "\n",
    "\n",
    "* Por tanto con las técnicas que nos permiten representar a las palabras con vectores, tendremos que la similaridad entre pares de vectores que representen a pares de palabras con significado similar, tendran una similaridad alta y no así entre pares de palabras con significados distintos que tendrán una similaridad baja.\n",
    "\n",
    "\n",
    "* Si por ejemplo esta similaridad la definimos como la distacia euclidea, tendremos que dos vectores que representan palabras con significados similares les separará una distancia pequeña frente a pares de palabras con significados dispares que les separará una distancia muy grande.\n",
    "\n",
    "\n",
    "* Dos técnicas que se suelen utilizar para realizar este tipo de cálculo de características latentes de las palabras son:\n",
    "\n",
    "    + Word2Vec\n",
    "    + GloVe\n",
    "    \n",
    "    \n",
    "* Este tipo de representación de las palabras viene muy bien para las RNN y en particular para las LSTM ya que cada palabra representada como un vector denso (y con sentido contextual frente a palabras similares entre sí) puede ser tomada como una entrada en una secuencia (que lo es) dentro de un documento.\n",
    "\n",
    "\n",
    "* Veamos a continuación un sencillo ejemplo a partir del siguiente corpus"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "corpus = ['Hola me llamo Ricardo',\n",
    "          'Hola me dicen Ricardo',\n",
    "          'Saludos desde este Notebook llamado LSTM']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Codificamos las palabras"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[58, 69, 45, 33]\n",
      "[58, 69, 34, 33]\n",
      "[36, 65, 3, 17, 49, 24]\n"
     ]
    }
   ],
   "source": [
    "from tensorflow.keras.preprocessing.text import one_hot\n",
    "\n",
    "VOCAB_SIZE = 100\n",
    "corpus_ohe = [one_hot(doc, VOCAB_SIZE) for doc in corpus]\n",
    "for i in corpus_ohe:\n",
    "    print(i)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Las pasamos a secuncias de tamaó fijo"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[58, 69, 45, 33,  0,  0],\n",
       "       [58, 69, 34, 33,  0,  0],\n",
       "       [36, 65,  3, 17, 49, 24]])"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
    "\n",
    "MAX_WORDS_DOC = 6\n",
    "pad_corpus = pad_sequences(corpus_ohe, maxlen=MAX_WORDS_DOC, padding='post')\n",
    "pad_corpus"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Resultados del Embedding\n",
    "\n",
    "* Para realizar esta representación, TensorFlow nos proporciona la clase \"Embedding()\" que tiene como importantes los siguientes parámetros:\n",
    "\n",
    "    + ***input_dim***: Número de codificaciones distintas que pueden tener las palabras de nuestro corpus.\n",
    "    + ***output_dim***: Número de características que queremos que tenga el vector que represente a las palabras.\n",
    "    + ***input_length***: Máximo número de palabras en un documento.\n",
    "    \n",
    "    \n",
    "* Para más información ver la documentación en el siguiente enlace:<br>https://www.tensorflow.org/api_docs/python/tf/keras/layers/Embedding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_1 (InputLayer)         [(None, 6)]               0         \n",
      "_________________________________________________________________\n",
      "embedding_1 (Embedding)      (None, 6, 2)              200       \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 12)                0         \n",
      "=================================================================\n",
      "Total params: 200\n",
      "Trainable params: 200\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from tensorflow.keras.layers import Input, Embedding, Flatten\n",
    "from tensorflow.keras.models import Model\n",
    "\n",
    "WORD_VECTOR_DIMENSION = 2\n",
    "\n",
    "words_input=Input(shape=(MAX_WORDS_DOC,),dtype='float64')  \n",
    "\n",
    "# Creamos el Embedding\n",
    "word_embedding=Embedding(input_dim=VOCAB_SIZE, \n",
    "                         output_dim=WORD_VECTOR_DIMENSION, \n",
    "                         input_length=MAX_WORDS_DOC)(words_input)\n",
    "\n",
    "word_vec=Flatten()(word_embedding)\n",
    "\n",
    "model = Model(inputs=[words_input],\n",
    "              outputs=word_vec) \n",
    "\n",
    "# Imprimimos el modelo\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Compilamos el modelo ya que estamos ejecutando una Red Neuronal"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['acc']) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Obtenemos los vectores de características haciendo un \"*predict()*\" sobre el corpus con el modelo generado."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "doc_embeddings=model.predict(pad_corpus)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Vamos a obtener los vectores de las palabras de los documentos del corpus, cuyos documentos tienen un tamaño fijo de 6 palabras:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Codificación de las 6 palabras del documento 1\n",
      "[[-0.02140704  0.03718336]\n",
      " [ 0.01496066 -0.04785068]\n",
      " [-0.01302377 -0.01080942]\n",
      " [ 0.00899512  0.01884394]\n",
      " [-0.04109632  0.04379017]\n",
      " [-0.04109632  0.04379017]]\n",
      "\n",
      "Codificación de las 6 palabras del documento 2\n",
      "[[-0.02140704  0.03718336]\n",
      " [ 0.01496066 -0.04785068]\n",
      " [-0.04369183  0.02071175]\n",
      " [ 0.00899512  0.01884394]\n",
      " [-0.04109632  0.04379017]\n",
      " [-0.04109632  0.04379017]]\n",
      "\n",
      "Codificación de las 6 palabras del documento 3\n",
      "[[ 0.04401072 -0.01403987]\n",
      " [-0.01878685 -0.04087279]\n",
      " [-0.03066066 -0.04682974]\n",
      " [ 0.03196997  0.02682967]\n",
      " [ 0.01063003 -0.03640258]\n",
      " [ 0.0340489  -0.04343636]]\n"
     ]
    }
   ],
   "source": [
    "word_embeddings=doc_embeddings.reshape(-1, MAX_WORDS_DOC, WORD_VECTOR_DIMENSION)\n",
    "for i, doc in enumerate(word_embeddings):\n",
    "    print(\"\\nCodificación de las 6 palabras del documento {}\".format(i+1))\n",
    "    print(doc)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* A continuación obtenemos la codificación de los documentos que no es más que la concatenación de los valores de las características de las palabras que forman el documento:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Codificación del documento 1\n",
      "[-0.02140704  0.03718336  0.01496066 -0.04785068 -0.01302377 -0.01080942\n",
      "  0.00899512  0.01884394 -0.04109632  0.04379017 -0.04109632  0.04379017]\n",
      "\n",
      "Codificación del documento 2\n",
      "[-0.02140704  0.03718336  0.01496066 -0.04785068 -0.04369183  0.02071175\n",
      "  0.00899512  0.01884394 -0.04109632  0.04379017 -0.04109632  0.04379017]\n",
      "\n",
      "Codificación del documento 3\n",
      "[ 0.04401072 -0.01403987 -0.01878685 -0.04087279 -0.03066066 -0.04682974\n",
      "  0.03196997  0.02682967  0.01063003 -0.03640258  0.0340489  -0.04343636]\n"
     ]
    }
   ],
   "source": [
    "for i, doc in enumerate(doc_embeddings):\n",
    "    print(\"\\nCodificación del documento {}\".format(i+1))\n",
    "    print(doc)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}