data_preparation + PCA.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np \n",
    "import pandas as pd \n",
    "import matplotlib.pyplot as plt \n",
    "from sklearn.decomposition import PCA\n",
    "import seaborn as sns\n",
    "from sklearn.svm import SVR \n",
    "from functools import reduce\n",
    "import xgboost\n",
    "%matplotlib inline\n",
    "from sklearn.utils.class_weight import compute_sample_weight\n",
    "from sklearn.ensemble import RandomForestRegressor\n",
    "from sklearn.metrics import accuracy_score\n",
    "from sklearn.manifold import TSNE\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dataset shape: (7803, 131)\n",
      "Labels train shape: (3739, 2)\n",
      "Labels val shape: (415, 2)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/admin/anaconda/lib/python3.5/site-packages/IPython/core/interactiveshell.py:2723: DtypeWarning: Columns (0,1,2,6,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129) have mixed types. Specify dtype option on import or set low_memory=False.\n",
      "  interactivity=interactivity, compiler=compiler, result=result)\n"
     ]
    }
   ],
   "source": [
    "data_filename = 'data/btsv01.txt'\n",
    "rois = pd.read_csv(data_filename, delim_whitespace=True)\n",
    "labels_train_filename = 'data/training_fluid_intelligenceV1.csv'\n",
    "labels_val_filename = 'data/validation_fluid_intelligenceV1.csv'\n",
    "labels_train = pd.read_csv(labels_train_filename, sep=',')\n",
    "labels_val = pd.read_csv(labels_val_filename, sep=',')\n",
    "print(\"Dataset shape: {}\".format(rois.shape))\n",
    "print(\"Labels train shape: {}\".format(labels_train.shape))\n",
    "print(\"Labels val shape: {}\".format(labels_val.shape))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "columns_description = rois.iloc[0]\n",
    "#delete the row with description of the columns from data\n",
    "rois_clean = rois.drop(rois.index[[0]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "collection_id                                                      collection_id\n",
      "btsv01_id                                                              btsv01_id\n",
      "dataset_id                                                            dataset_id\n",
      "subjectkey                     The NDAR Global Unique Identifier (GUID) for r...\n",
      "src_subject_id                        Subject ID how it's defined in lab/project\n",
      "interview_date                 Date on which the interview/genetic test/sampl...\n",
      "interview_age                  Age in months at the time of the interview/tes...\n",
      "gender                                                        Sex of the subject\n",
      "sri24precentrallgm                      left precentral gyrus gray matter volume\n",
      "sri24precentralrgm                     right precentral gyrus gray matter volume\n",
      "sri24frontalsuplgm             left superior frontal gyrus, dorsolater gray m...\n",
      "sri24frontalsuprgm             right superior frontal gyrus, dorsolater gray ...\n",
      "sri24frontalsuporblgm          left superior frontal gyrus, orbital part gray...\n",
      "sri24frontalsuporbrgm          right superior frontal gyrus, orbital part gra...\n",
      "sri24frontalmidlgm             left middle frontal gyrus, lateral part gray m...\n",
      "sri24frontalmidrgm             right middle frontal gyrus, lateral part gray ...\n",
      "sri24frontalmidorblgm          left middle frontal gyrus, orbital part gray m...\n",
      "sri24frontalmidorbrgm          right middle frontal gyrus, orbital part gray ...\n",
      "sri24frontalinfoperlgm         left opercular part of inferior frontal gyrus ...\n",
      "sri24frontalinfoperrgm         right opercular part of inferior frontal gyrus...\n",
      "sri24frontalinftrilgm                  left area triangularis gray matter volume\n",
      "sri24frontalinftrirgm                 right area triangularis gray matter volume\n",
      "sri24frontalinforblgm          left orbital part of inferior frontal gyrus gr...\n",
      "sri24frontalinforbrgm          right orbital part of inferior frontal gyrus g...\n",
      "sri24rolandicoperlgm                  left rolandic operculum gray matter volume\n",
      "sri24rolandicoperrgm                 right rolandic operculum gray matter volume\n",
      "sri24suppmotorarealgm           left supplementary motor area gray matter volume\n",
      "sri24suppmotorareargm          right supplementary motor area gray matter volume\n",
      "sri24olfactorylgm                       left olfactory cortex gray matter volume\n",
      "sri24olfactoryrgm                      right olfactory cortex gray matter volume\n",
      "                                                     ...                        \n",
      "sri24cerebelumcrus2rvolume         right crus II of cerebellar hemisphere volume\n",
      "sri24cerebelum3lvolume           left Lobule III of cerebellar hemisphere volume\n",
      "sri24cerebelum3rvolume          right Lobule III of cerebellar hemisphere volume\n",
      "sri24cerebelum45lvolume        left lobule IV, V of cerebellar hemisphere volume\n",
      "sri24cerebelum45rvolume        right lobule IV, V of cerebellar hemisphere vo...\n",
      "sri24cerebelum6lvolume            left Lobule VI of cerebellar hemisphere volume\n",
      "sri24cerebelum6rvolume           right Lobule VI of cerebellar hemisphere volume\n",
      "sri24cerebelum7blvolume         left lobule VIIB of cerebellar hemisphere volume\n",
      "sri24cerebelum7brvolume        right lobule VIIB of cerebellar hemisphere volume\n",
      "sri24cerebelum8lvolume          left lobule VIII of cerebellar hemisphere volume\n",
      "sri24cerebelum8rvolume         right lobule VIII of cerebellar hemisphere volume\n",
      "sri24cerebelum9lvolume            left lobule IX of cerebellar hemisphere volume\n",
      "sri24cerebelum9rvolume           right lobule IX of cerebellar hemisphere volume\n",
      "sri24cerebelum10lvolume        left lobule X of cerebellar hemisphere (floccu...\n",
      "sri24cerebelum10rvolume        right lobule X of cerebellar hemisphere (flocc...\n",
      "sri24vermis1gm                                anterior vermis gray matter volume\n",
      "sri24vermis2gm                               posterior vermis gray matter volume\n",
      "sri24vermis3gm                                inferior vermis gray matter volume\n",
      "sri24cblmhemiwhtlwm               cerebellum hemisphere white matter left volume\n",
      "sri24cblmhemiwhtrwm              cerebellum hemisphere white matter right volume\n",
      "sri24ponswm                                             pons white matter volume\n",
      "sri24corpuscallosumwm                        corpus callosum white matter volume\n",
      "sri24wm400wm400lwm             white matter volume within the left hemisphere...\n",
      "sri24wm400wm400rwm             white matter volume within the right hemispher...\n",
      "sri24vtlslateralvtllcsf        left lateral ventricle cerebral spinal fluid v...\n",
      "sri24vtlslateralvtlrcsf        right lateral ventricle cerebral spinal fluid ...\n",
      "sri24vtlsthirdvtllcsf          left third ventricle cerebral spinal fluid volume\n",
      "sri24vtlsthirdvtlrcsf          right third ventricle cerebral spinal fluid vo...\n",
      "sri24suptentsupratentoriumv                        suptent supratentorium volume\n",
      "collection_title                                                collection_title\n",
      "Name: 0, Length: 131, dtype: object\n"
     ]
    }
   ],
   "source": [
    "print(columns_description)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of rows: 7802\n",
      "Number of columns: 131\n"
     ]
    }
   ],
   "source": [
    "print(\"Number of rows: {}\".format(rois_clean.shape[0]))\n",
    "print(\"Number of columns: {}\".format(rois_clean.shape[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3739"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(set(labels_train['subject']).intersection(set(rois_clean['subjectkey'])))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now, we found that our labels with residual fluid intelligence scores are a subset of values in a column 'subjectkey'.\n",
    "We are constructing a training set, therefore in the training set we will only consider those entries that also have labels. <br>\n",
    "Furthermore, we are eliminating the following columns: \n",
    "- btsv01_id\n",
    "- interview_date\n",
    "- collection_id\n",
    "- dataset_id\n",
    "- collection_title\n",
    "- src_subject_id (which is the repetition of subjectkey, but with lowercase letters (for some mysterious reason))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "rois_clean = rois_clean.drop(['btsv01_id', 'interview_date', 'collection_id', \\\n",
    "                              'dataset_id', 'collection_title', 'src_subject_id'], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#now we are changing the string values of gender to number\n",
    "gender_to_int = {\n",
    "    'M': 0,\n",
    "    'F': 1\n",
    "}\n",
    "\n",
    "rois_clean['gender'] = rois_clean['gender'].apply(lambda x: gender_to_int[x])\n",
    "# rois_clean = pd.get_dummies(rois_clean, columns=['gender'])\n",
    "#skip gender in this analysis\n",
    "rois_clean = rois_clean.drop(['gender'], axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create training and validation dataset "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_data = rois_clean[rois_clean['subjectkey'].isin(labels_train['subject'])]\n",
    "val_data = rois_clean[rois_clean['subjectkey'].isin(labels_val['subject'])]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_data_labels = pd.merge(train_data, labels_train, left_on='subjectkey', right_on='subject').drop(['subject'], axis=1)\n",
    "val_data_labels = pd.merge(val_data, labels_val, left_on='subjectkey', right_on='subject').drop(['subject'], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#load features from the images\n",
    "means_train = pd.read_csv('data/training_data_means.csv', header=None, prefix='m', index_col=0).fillna(0)\n",
    "means_val = pd.read_csv('data/validation_data_means.csv', header=None, prefix='m', index_col=0).fillna(0)\n",
    "std_train = pd.read_csv('data/training_data_stdevs.csv', header=None, prefix='std', index_col=0).fillna(0)\n",
    "std_val = pd.read_csv('data/validation_data_stdevs.csv', header=None, prefix='std', index_col=0).fillna(0)\n",
    "entropy_train = pd.read_csv('data/training_data_entropy.csv', header=None, prefix='e', index_col=0).fillna(0)\n",
    "entropy_val = pd.read_csv('data/validation_data_entropy.csv', header=None, prefix='e', index_col=0).fillna(0)\n",
    "\n",
    "extra_dfs = [means_train, means_val, std_train, std_val, entropy_train, entropy_val]\n",
    "for df in extra_dfs:\n",
    "    df.index.name = 'subjectkey'\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_additional_features = reduce(lambda left,right: pd.merge(left,right,on='subjectkey'), [means_train, std_train, entropy_train])\n",
    "val_additional_features =  reduce(lambda left,right: pd.merge(left,right,on='subjectkey'), [means_val, std_val, entropy_val])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#here we will construct a matrix only with data - without labels\n",
    "#the first column is identification - it is a string and we want to delete it. Other columns are identified as string,\n",
    "#but we want to convert it to floats and to the matrix\n",
    "\n",
    "def generate_dataset(df_train, df_val, include_additional=True):\n",
    "    #the input is the dataframe with both data and labels in the last column\n",
    "    train_labels = df_train['residual_fluid_intelligence_score'].astype(np.float).values\n",
    "    val_labels = df_val['residual_fluid_intelligence_score'].astype(np.float).values\n",
    "    drop_columns = ['subjectkey', 'residual_fluid_intelligence_score']\n",
    "    if include_additional: \n",
    "        train_data = pd.merge(df_train, train_additional_features, on='subjectkey').drop(drop_columns, axis=1).astype(np.float)\n",
    "        val_data = pd.merge(df_val, val_additional_features, on='subjectkey').drop(drop_columns, axis=1).astype(np.float)\n",
    "    else:\n",
    "        train_data = df_train.drop(drop_columns, axis=1).values.astype(np.float)\n",
    "        val_data = df_val.drop(drop_columns, axis=1).values.astype(np.float)\n",
    "    return train_data, train_labels, val_data, val_labels\n",
    "\n",
    "train_data, train_labels, val_data, val_labels = generate_dataset(train_data_labels, val_data_labels, include_additional=False)\n",
    "train_data_extra, train_labels_extra, val_data_extra, val_labels_extra = generate_dataset(train_data_labels, val_data_labels, include_additional=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#now for the final form - train_data/val_data will be our matrices; train_labels, val_labels will be our residuals\n",
    "dataset_columns = train_data_labels.columns[1:-1] \n",
    "#important: the last two columns are now binarized wrt to genders"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Experiment on the initial features"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Distribution of residual values (what are our labels?)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAAEACAYAAAC9Gb03AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAEopJREFUeJzt3X+sXGd95/H3xzHxkrK4ptQeFKc4VdLgUFRAi1mJXWm0\npGlgJTtSqzStWiWk/QsV+KNqYwOtL6vVQvgHIe1mpdVSZK3CZl0QjdmliRPCqGqlJHRJIMXGe1fU\nxrj4Zpemu8qmQgn59o85yQ6+zr1zf4xn5sn7JY1y5pnznPkeZ+ZzzzxzzjOpKiRJ7doy7QIkSZNl\n0EtS4wx6SWqcQS9JjTPoJalxBr0kNW6soE+yPckfJzmZ5FtJ3plkR5LjSU4leSDJ9pH1DyVZ7Na/\ncXLlS5JWM+4R/aeBL1fVXuAXgG8DB4GHquo64GHgEECS64FbgL3Ae4C7k2SzC5ckjWfVoE/yWuCf\nV9VnAarq+ar6P8AB4Ei32hHg5m55P3Bvt95pYBHYt9mFS5LGM84R/dXA/07y2SRfT/IfklwB7Kqq\nJYCqOg/s7Na/Ejg70v9c1yZJmoJxgn4r8Hbg31XV24H/x3DY5sK5E5xLQZJm0NYx1vkecLaq/rK7\n/wWGQb+UZFdVLSXpAU91j58Drhrpv7tr+zFJ/MMgSetQVWv63nPVI/pueOZskp/rmt4NfAs4Btze\ntd0G3NctHwNuTXJ5kquBa4DHXmbbc3s7fPjw1Guw/unX8Uqsf55rb6H+9RjniB7gg8A9SV4FfAd4\nH3AZcDTJHcAZhmfaUFUnkhwFTgDPAe+v9VYnSdqwsYK+qr4BvOMiD93wMut/HPj4BuqSJG0Sr4xd\np36/P+0SNsT6p2ue65/n2mH+61+PTGtUJYkjOpK0Rkmozf4yVpI03wx6SWqcQS9JjTPoJalxBr0k\nNc6gl6TGGfSS1DiDXpIaZ9BLUuMMeklqnEEvSY0z6CWpcQa9JDXOoJc2Ua+3hyTrvvV6e6a9C2qQ\n0xRLmygJsJHXddb9c3F6ZXCaYknSMga9NFO2OfSjTefQjbSJNmPoxqEfrcShG0nSMga9JDXOoJek\nxhn0ktQ4g16SGmfQS1LjDHpJapxBL0mNGyvok5xO8o0kjyd5rGvbkeR4klNJHkiyfWT9Q0kWk5xM\ncuOkipckrW7cI/oXgH5Vva2q9nVtB4GHquo64GHgEECS64FbgL3Ae4C7M7xcUJI0BeMGfS6y7gHg\nSLd8BLi5W94P3FtVz1fVaWAR2IckaSrGDfoCHkzytSS/3bXtqqolgKo6D+zs2q8Ezo70Pde1SZKm\nYOuY672rqr6f5KeB40lOsXzmJWdSkqQZNFbQV9X3u//+ryR/wnAoZinJrqpaStIDnupWPwdcNdJ9\nd9e2zMLCwkvL/X6ffr+/1volqWmDwYDBYLChbaw6TXGSK4AtVfVMkp8AjgMfA94N/G1V3ZXkTmBH\nVR3svoy9B3gnwyGbB4FrL5yT2GmKNYt6vT0sLZ3Z4FacpliTs55pisc5ot8FfDFJdevfU1XHk/wl\ncDTJHcAZhmfaUFUnkhwFTgDPAe830TUvhiG/0aCWZos/PCKNmIUfDvGIXivxh0ckScsY9JLUOINe\nkhpn0EtS4wx6SWqcQS9JjTPoJalxBr0kNc6gl6TGGfSS1DiDXpIaZ9BLUuMMeklqnEEvSY0z6CWp\ncQa9JDXOoJekxhn0ktQ4g16SGmfQS1LjDHqpKdtIsu5br7dn2jugCci0fjE+Sflr9Zo1SYCNvC7n\nv7/vy9mWhKrKWvp4RC9JjTPoJalxBr0kNc6gl6TGGfSS1DiDXpIaZ9BLUuPGDvokW5J8Pcmx7v6O\nJMeTnEryQJLtI+seSrKY5GSSGydRuCRpPGs5ov8QcGLk/kHgoaq6DngYOASQ5HrgFmAv8B7g7gyv\nQpEkTcFYQZ9kN/Be4D+ONB8AjnTLR4Cbu+X9wL1V9XxVnQYWgX2bUq0kac3GPaL/FPB7/Pi11buq\nagmgqs4DO7v2K4GzI+ud69okSVOwdbUVkvxLYKmqnkjSX2HVNU+QsbCw8NJyv9+n319p85L0yjMY\nDBgMBhvaxqqTmiX5N8BvAM8Drwb+MfBF4J8A/apaStIDvlpVe5McBKqq7ur63w8crqpHL9iuk5pp\n5jipmZOazbqJTGpWVR+uqp+pqp8FbgUerqrfBL4E3N6tdhtwX7d8DLg1yeVJrgauAR5bS1GSpM2z\n6tDNCj4BHE1yB3CG4Zk2VNWJJEcZnqHzHPB+D90laXqcj14a4dCNQzezzvnoJUnLGPSS1DiDXpIa\nZ9BLUuMMeklqnEEvSY0z6CWpcQa9mtLr7SHJum9Si7xgSk3xgicvmGqdF0xJkpYx6CWpcQa9JDXO\noJekxhn0ktQ4g16SGmfQS1LjDHpJapxBL0mNM+glqXEGvSQ1zqCXNGLbhiaF6/X2THsHdBFOaqam\nOKnZ9Pv7vp4sJzWTJC1j0EtS4wx6SWqcQS9JjTPoJalxBr0kNW7VoE+yLcmjSR5P8mSSw137jiTH\nk5xK8kCS7SN9DiVZTHIyyY2T3AFJ0srGOo8+yRVV9WySy4C/AD4I/DLwg6r6ZJI7gR1VdTDJ9cA9\nwDuA3cBDwLUXnjTvefSaBM+jn35/39eTNbHz6Kvq2W5xG7CV4SvhAHCkaz8C3Nwt7wfurarnq+o0\nsAjsW0tRkqTNM1bQJ9mS5HHgPPBgVX0N2FVVSwBVdR7Y2a1+JXB2pPu5rk2SNAVbx1mpql4A3pbk\ntcAXk7yZ5Z/v1vx5bWFh4aXlfr9Pv99f6yYkqWmDwYDBYLChbax5rpskfwA8C/w20K+qpSQ94KtV\ntTfJQaCq6q5u/fuBw1X16AXbcYxem84x+un39309WRMZo0/y+hfPqEnyauAXgZPAMeD2brXbgPu6\n5WPArUkuT3I1cA3w2FqKkiRtnnGGbt4AHEmyheEfhv9SVV9O8ghwNMkdwBngFoCqOpHkKHACeA54\nv4fukjQ9TlOspjh0M/3+vq8ny2mKJUnLGPSS1DiDXpIaZ9BLUuMMeklqnEEvSY0z6CWpcQa9JDXO\noJekxhn0ktQ4g16SGmfQS1LjDHpJapxBL0mNM+glqXEGvSQ1zqCXpMYZ9JLUOINekhpn0Gum9Hp7\nSLLum6Tl/HFwzRR/3Hv++/u+nix/HFyStIxBL0mNM+glqXEGvSQ1zqCXpMYZ9JLUOINekhq3atAn\n2Z3k4STfSvJkkg927TuSHE9yKskDSbaP9DmUZDHJySQ3TnIHJEkrW/WCqSQ9oFdVTyR5DfDfgQPA\n+4AfVNUnk9wJ7Kiqg0muB+4B3gHsBh4Crr3w6igvmNLFeMHU/Pf3fT1ZE7lgqqrOV9UT3fIzwEmG\nAX4AONKtdgS4uVveD9xbVc9X1WlgEdi3lqIkSZtnTWP0SfYAbwUeAXZV1RIM/xgAO7vVrgTOjnQ7\n17VJkqZg67grdsM2nwc+VFXPJLnw89maP68tLCy8tNzv9+n3+2vdhCQ1bTAYMBgMNrSNsSY1S7IV\n+K/An1bVp7u2k0C/qpa6cfyvVtXeJAeBqqq7uvXuBw5X1aMXbNMxei3jGP289/9HwA/X3XvXrjdy\n/vzpDTx/+yY5qdkfASdeDPnOMeD2bvk24L6R9luTXJ7kauAa4LG1FCVpXv2Q4R+K9d2Wls5Moeb2\njXPWzbuAPwOe5P//H/kww/A+ClwFnAFuqaq/6/ocAn4LeI7hUM/xi2zXI3ot4xG9/c2Fla3niN75\n6DVTDHr7mwsrcz56SdIyBr0kNc6g16byN1+l2eMYvTaVY+z2d4x+shyjlyQtY9BLUuMMeklqnEEv\nSY0z6CWpcQa9JDXOoJekxhn0ktQ4g16SGmfQS1LjDHpJapxBL0mNM+glqXEGvSQ1zqCXpMYZ9JLU\nOINekhpn0EtS4wx6SWqcQS9JjTPoJalxBr2kGbKNJOu+9Xp7pr0DMylVNZ0nTmpaz63JSQJs5P+r\n/e2/sf6t50oSqipr6eMRvSQ1btWgT/KZJEtJvjnStiPJ8SSnkjyQZPvIY4eSLCY5meTGSRUuSRrP\nOEf0nwV+6YK2g8BDVXUd8DBwCCDJ9cAtwF7gPcDdGX6WlyRNyapBX1V/Djx9QfMB4Ei3fAS4uVve\nD9xbVc9X1WlgEdi3OaVKktZjvWP0O6tqCaCqzgM7u/YrgbMj653r2iRJU7JZX8a2/TW3JM2xrevs\nt5RkV1UtJekBT3Xt54CrRtbb3bVd1MLCwkvL/X6ffr+/znIkqU2DwYDBYLChbYx1Hn2SPcCXquot\n3f27gL+tqruS3AnsqKqD3Zex9wDvZDhk8yBw7cVOmPc8+tnU6+1haenMBrcy3+dh23+++7eeK+s5\nj37VoE/yOaAP/BSwBBwG/gT4Y4ZH72eAW6rq77r1DwG/BTwHfKiqjr/Mdg36GeQFT/af9/6t58pE\ngn5SDPrZZNDbf977t54rXhkrSVrGoJekxhn0ktQ4g16SGmfQS1LjDHpJapxBL0mNM+glqXEGvSQ1\nzqCXpMYZ9JLUOINekhpn0Dem19tDknXfpPm2bUOv/15vz7R3YCKcvbIxzj5pf/u3Pfuls1dKkpYx\n6CWpcQa9JDXOoJekxhn0ktQ4g16SGmfQS1LjDHpJapxBL0mNM+hnjFMYSNpsToEwY5zCwP72dwqE\nlTgFgiRtSJuTonlEv8l6vT0sLZ3Z4Fbm+4jI/vZ/JfefdK6t54jeoN9kDr3Y3/6v7P6zGPQTG7pJ\nclOSbyf5H0nunNTzSJJWNpGgT7IF+LfALwFvBn4tyZsm8VzTMhgMpl3CBg2mXcAGDaZdwAYNpl3A\nBgymXcAGDaZdwCU3qSP6fcBiVZ2pqueAe4EDE3quqTDop20w7QI2aDDtAjZgMO0CNmgw7QIuua0T\n2u6VwNmR+99jGP6revrppzc0xrV9+3Yuu+yydfeXpNZMKujX5Stf+Qo33HDDhrZxxRU/xbPP/mDd\n/bdsuYIXXnh2rHU/9rGPrft5JLVo24YuXNy1642cP39688rpTOSsmyT/FFioqpu6+weBqqq7RtZp\n75QbSboEZuL0yiSXAaeAdwPfBx4Dfq2qTm76k0mSVjSRoZuq+lGS3wGOM/zC9zOGvCRNx9QumJIk\nXRpTm+smye8meSHJ60baDiVZTHIyyY3Tqm0lSf5Vkm8keTzJ/Ul6I4/NQ/2f7Op7IskXkrx25LGZ\nrj/JryT5qyQ/SvL2Cx6b6dpfNG8XEib5TJKlJN8caduR5HiSU0keSLJ9mjWuJMnuJA8n+VaSJ5N8\nsGuf+X1Isi3Jo13WPJnkcNe+9tqr6pLfgN3A/cBfA6/r2vYCjzMcTtoD/E+6TxyzdANeM7L8AeDf\nd8vXz0n9NwBbuuVPAB+fl/qB64BrgYeBt4+0z8trZ0tX2xuBVwFPAG+adl2r1PzPgLcC3xxpuwv4\n/W75TuAT065zhfp7wFu75dcw/O7wTfOyD8AV3X8vAx5heJr6mmuf1hH9p4Dfu6DtAHBvVT1fVaeB\nRcY89/5SqqpnRu7+BPBCt7yf+aj/oap6seZHGP7RhTmov6pOVdUiwwlJRs3Fa4c5vJCwqv4cePqC\n5gPAkW75CHDzJS1qDarqfFU90S0/A5xk+Jqfi32oqhfP9d7G8ECmWEftlzzok+wHzlbVkxc8dOFF\nVue6tpmT5F8n+S7w68Afds1zU/+IO4Avd8vzWP+L5qX2i11IOIt1rmZnVS3BMEiBnVOuZyxJ9jD8\ndPIIsGse9iHJliSPA+eBB6vqa6yj9omcdZPkQWDXaBPDv0QfBT4M/OIknnezrFD/R6rqS1X1UeCj\n3RjrB4CFS1/ly1ut/m6djwDPVdV/nkKJL2uc2jVzZv6MjiSvAT4PfKiqnrnIdTwzuQ/dp++3dd+l\nfTHJm1le66q1T+r0yosGeZKfZziG+o0MLx/bDXw9yT6GR2E/M7L67q7tknu5+i/ic8B/Yxj054Cr\nRh6b2fqT3A68F/gXI80zUf8a/u1HzUTtY5iZ1/gGLSXZVVVL3ckIT027oJUk2cow5P9TVd3XNc/V\nPlTV/00yAG5iHbVf0qGbqvqrqupV1c9W1dUMP7q+raqeAo4Bv5rk8iRXA9cwvNBqpiS5ZuTuzcC3\nu+VjwK1zUP9NDL8f2V9VPxx5aC7qHzE6Tj8vtX8NuCbJG5NcDtzKsPZZF5b/e9/eLd8G3Hdhhxnz\nR8CJqvr0SNvM70OS1794Rk2SVzMcCTnJemqf8jfK36E766a7f4jhWQkngRun/Y33y9T8eeCbDM+Y\nuA94w5zVvwicAb7e3e6el/oZ/mE9C/w9wyuu/3Reah+p8yaGZ34sAgenXc8Y9X4O+Bvgh8B3gfcB\nO4CHuv04DvzktOtcof53AT/q3q+Pd6/5m4DXzfo+AG/p6n2iy5yPdO1rrt0LpiSpcf44uCQ1zqCX\npMYZ9JLUOINekhpn0EtS4wx6SWqcQS9JjTPoJalx/wC9ov/Xb6Lw1QAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11b91dbe0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(train_labels, bins=20)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### How many Nans we have?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 7293.39,  7716.8 ,  8031.32, ..., 15400.2 , 15799.6 , 16779.8 ])"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.unique(train_data[:,90])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### PCA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#first normalize with respect to column\n",
    "def normalize_data(data: np.array, standardize=True):\n",
    "    means = np.mean(data, axis=0)\n",
    "    stds = np.std(data, axis=0)\n",
    "    if not standardize:\n",
    "        return data-means\n",
    "    return (data-means)/stds\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_norm_data = normalize_data(train_data)\n",
    "val_norm_data = normalize_data(val_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "For 2 components, the explained variance by them is : 0.152\n",
      "For 3 components, the explained variance by them is : 0.191\n",
      "For 4 components, the explained variance by them is : 0.225\n",
      "For 5 components, the explained variance by them is : 0.254\n",
      "For 6 components, the explained variance by them is : 0.281\n",
      "For 7 components, the explained variance by them is : 0.305\n",
      "For 8 components, the explained variance by them is : 0.327\n",
      "For 9 components, the explained variance by them is : 0.347\n",
      "For 10 components, the explained variance by them is : 0.365\n",
      "**********\n",
      "The 0.8 variance ratio is explained by 57 components \n",
      "The 0.85 variance ratio is explained by 66 components \n",
      "The 0.9 variance ratio is explained by 78 components \n",
      "The 0.95 variance ratio is explained by 92 components \n"
     ]
    }
   ],
   "source": [
    "n_components = [2,3,4,5,6,7,8,9,10]\n",
    "for n in n_components:\n",
    "    curr_pca = PCA(n_components=n)\n",
    "    curr_pca.fit_transform(train_norm_data)\n",
    "    print(\"For {} components, the explained variance by them is : {:.3f}\".format(n, np.sum(curr_pca.explained_variance_ratio_)))\n",
    "\n",
    "print(\"*\"*10)\n",
    "explained_variance = [0.8, 0.85, 0.9, 0.95]\n",
    "for target_variance in explained_variance:\n",
    "    curr_pca = PCA(n_components=target_variance)\n",
    "    curr_pca.fit_transform(train_norm_data)\n",
    "    print(\"The {} variance ratio is explained by {} components \".format(target_variance, np.sum(curr_pca.n_components_)))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Covariance matrix of features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAjgAAAJBCAYAAABLWM//AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11bea5630>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def compute_covariance(matrix):\n",
    "    return matrix.T.dot(matrix)/matrix.shape[0]\n",
    "\n",
    "cov_matrix = compute_covariance(train_norm_data)\n",
    "fig, ax = plt.subplots(figsize=(10,10))\n",
    "im = ax.matshow(cov_matrix, interpolation='none')\n",
    "fig.colorbar(im, ticks=[-1,-0.8,-0.6,-0.4,-0.2,0.0,0.2,0.4,0.6,0.8,1.0])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def get_eigenvalues(data, n_comps, standardize=True):\n",
    "    normalized_data = normalize_data(data, standardize)\n",
    "    cov_matrix = compute_covariance(normalized_data)\n",
    "    pca = PCA(n_comps)\n",
    "    pca.fit_transform(normalized_data)\n",
    "    eigenvectors = pca.components_\n",
    "    eigenvalues = []\n",
    "    for ev in eigenvectors:\n",
    "        eigenvalues.append(np.dot(ev.T, np.dot(cov_matrix, ev)))\n",
    "    print(pca.explained_variance_)\n",
    "    return eigenvalues"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[11.33334393  7.37416555  4.77442318]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[11.330312816056491, 7.372193327440167, 4.773147275910677]"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "get_eigenvalues(train_data, 3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Evaluating feature importance\n",
    "http://jotterbach.github.io/2016/03/24/Principal_Component_Analysis/"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/admin/anaconda/lib/python3.5/site-packages/ipykernel/__main__.py:4: RuntimeWarning: invalid value encountered in log\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x11c8b17f0>"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWYAAAEJCAYAAACnhI2ZAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzsvX+4HVV5NvwsA5EQS4PBkM8onpqiqRFFY4SiyDYYTWMj\nFotUFCOpyCuieUUUQWtyLAKhKEVT/CgIjVQqYlHIa4rExo1RjIZgNKaNYOw2Gt8QjR6iMXIgrPeP\nue+1n/3Mmtkze++TfU4y93Wda52ZvWZmza819/Pbee+lQoUKFSqMHjyp3wOoUKFChQqtqCbmChUq\nVBhlqCbmChUqVBhlqCbmChUqVBhlqCbmChUqVBhlqCbmChUqVBhlqCbmChUqVCgI59w859wW59yD\nzrmLR+o4IzYx768TqFChQoX9Aefck0RkuYi8RkRmisibnHMzRuJYIzIx788TqFChQoX9hJeKyEPe\n+5967x8Tkc+LyGkjcaCRYsz77QQqVKhQYT9hmoj8TC3/HOt6jkNGYqcSP4GXtvRY43x9o0jtwmIh\n4fV6XWq1v8HSjgJ9ayX2O9J9D0P7h97t927XXPkY2mvQrkmuaf2rTmqvEJH3Yf118Ws9+q5XD/v+\nHye1l2Ohgfb4c/DPHrQ7sN/vSq02DuveiJavyD+P+HiTfn+PpcNMq8fyY6nXfyu12tOwzOdqEtrN\napvJUq/vkVrtheZo3NfjaH+FMTwitdpkrPuW6UtutRF9fy+12tOx7k/N/tBuuyPp+5PD1LXF/m55\nJGm5+qxnY7/7pFabaM5DnHSPzMmmXq9LvV4Py4ODg6/03tez+o80+mb8q28UWbpCZOnSpbJ06dKW\nixLt3+b3qm+k7zf6fPzR0PebhbtKvb4P/70xt1/pMRTsW26fvy3Rd0/7TqHvIyX6/r5E333tO4W+\ne6Re3xPmhqVLl4pzrlZ4Bx2gVqu1HC9jUt4uIseo5WdgXc8xUoy57QnULvRS251chOKwX/0+YAs+\n3DNejBUbCmw0vfj+L8P+P9xGkpgX+X2BWT50iciEpSLXFT98R9iDMd+h1p09SpJj/YeI/F/8/0do\nl/8I//wU7cSk+cVjIhseE5GbRWZZxry/8J8F+y3F3yjv+zQS3UvTfc8+Af+AXfsHsDxRarVxUjt5\nsHmkpb1gr4+37xIQve/rReRPnXPPkuSp+hsReVP34yp49B5gv51AhQo9x6z2jLnCWER3E7P3fp9z\n7gIRuUcSbcNnvPf/3aPBtTl6D1DsBN4stdrDIsvxRb0gn2ElOrqlhY5fVPfXUd8ZnYxhc1a3pO+U\nQZF1YAdkymtwXeZgGay0drmILBsU2attDtQJfj8+hu3Y1zR1jS/Euk94qdVq8q8uWX5LXhrYlU5q\n40VS92EitjnbnFetJjKA42zEymG0U3y6b0Hk9r2F54q+135dpOC+a2fVRWYV7DsCz9iIPrfd9B2H\na7ov75691myESXDCOej7exG5yfT5Tuui437rIlJr6p57hriNJ47Domu993eLyHN7MpwcuL7lY/Yu\nOfD9WL4NrfbdeB7aya9O2uvvSdp3YD3VYZMmNrfZCX3aL7E888n4h+oEfXOoXYHRZMfDSTs175qc\nifbOpKk/mrQ1boN7tvvB5iYcy/Q5+GcrWqhmNtzV7PtztKedhH9g6HzoJ0lLq8Cz1ZAcL5T5AKzE\nk36CWjelB/d7LfbLxzNvn/eh70kFjrvajJcvJid+r+w/rqgqSRvt+CwsbD+WdtiFsUzOOa8G+gzw\nI/rzzK7BmBtTUaVAIx8mv8145mdi2wua1+k7/5S0J9j3/Az0uV2tX4J1R2P5fPy2AetnYfkKdR8u\nwbptWDce6/kOXYT1y9Sx+QpO/HjreQTo5Uv5Tw+Mf78q8fAf1QtjY8fY3wq0UYQR0dmPbvRiUq5w\nYGJJX+ehDFzavksplFFl9Bf9Y8x0XdlpvrSTCoxnCNuQiW5Rv5FMnNzFeZFxLB9jE9lWjPtmLF9m\nxz8zshHEjt34UB1Bpj612WUHrHpkUr9AS2PaETjOLvVyT56Pf4bQ/gotPQTAHteobai2uRbrFpMV\n8yOa7yYpIiKbKHZj+XjtpXkU2q90sJ+MZ+E+Nf6TQPGXYaOLIcn4/0pax2v7q+Y24Tpz0qA0heu2\n79FmV57+tHbP5TPU/+Bea2Ho5CU91GzyI/U/bxnfob04R0r3T6i+456FfxptxhTDR8zy79A+Ra37\nKP/pwZfj5yVe6Gf09Us1kiHZf+ycu90599/Ouc3OuRPab1WhwijAptHIHg9GfLR9l1J4vMRffzFi\njNk59y8icq/3/mbn3CEicrj3frfqkhyYjONwrD3+WaoLGMgc6FfXcKxkPtTE5DCpW7H/s6iUxTbb\nlQ4Yuw86bTLB8S9u3UZEUgaL4MIHxdm2ByWFYzDul2Es3zLX/Fo1EcxGW0QnWxb71HFohHsqWppm\nd6GdlXP8S7Gfy8392Lmr2SelNiGLgz3gTFynt6sujJegFxtvGSWiMyJjot6ZLH4aFdNUgOcbXgtj\nHY7D3c8eIWmK7pI0ZWgXZLLcmXxQeaH4zoD27lM+w+NgGN6Fe/NrrH+atGLSHLXAiQnP+hqwdko0\nm7VkhHYqx8ALxLH1bJLrwddya4mbNv3A0zE7544QkZO9928TEfHePy4iu3M3qlChQoURRf+ZcFGM\nlPHvT0TkV865myWhlPeLyGLv/d5ml4GkOYmsVOvFCOic1mw166ns+uP0JtfhQ0dGcFbGR3Jaxv8t\n+ADavMvEm3180hwD1qJZxUMZTJlYrPyCyHa4DfW6RxT52EM31wDNGjDbjFPLE9Dube3C25KLy+1Y\noDOdote9Bi1c+JbA44Xh4UxpNVftaxnOGV3lNPx2L6+l0j3acwz38M1oS7hGbTf6ZJGmzYOSxYnm\nnEOgkfIIkpeZY99rDhTx2PkMWkouH0Lrnixp8FlrIwWM0wFNYMoNLJLhFrHlEHPM8szIthfjevA8\nqE4/OXu3e+Ci+WksXzTi9q4y7nL9xUhNzIdIYmZ4l/f+fufcP4rIB0VkCTvU63+Qev0PIpKISbXa\nYVKrHRXbV4UK+xfj23epMLKI5K6odZ+7Yuww5hHRMTvnjhaRb3vvn43ll4vIxd57HTTcwYHJqnP8\nQbvCG9CSfn0y3YU6zSzfYaLF26Cdd8GA+p9MiQ+RtdrneBaQpXOTPB/bEQfZI8a947tJS1bK9mG1\nyezX4R9KROQNvCbaZ5k6zFPQ/jzSp5cw57Md5zPteaoPngV69TCimDaM2ZSMlMcLpY1d0OPm3TP6\nHv8tlikV0p+fIfHLJze3CRRZu12IyPEM/OHxtMRK74ghaYvF2M+1PXjWKDFdHN1XD3S+3ykxyBP6\nqmMeSePfvSJyrvf+QefcEkmMfzphfvaBaaSiO1ws4INO72Q3NB5OR99F+F0bl2hQoyGH4untWI4Z\nl/oGGhW1FEERlS+RdTcaKcxCiwlz933Nn454P/65qrXvFch7wAAEObW5zYY16HoWVlDEpKpCq6gY\nzUi1wIWmL9VcejIyH1S6ZP4YyyfxAzBJdeKkyevNCZ6TlDX6FoEOZOHH/vLWLjuMysr16Bm0OVdW\nYVmnmyE/mMXrbSdidtY+/7+SKOjJouy/4dYtbkNiiqEHE2WWLjGGl/V1Yh7JAJP3iMjnnHOHSsIZ\nzmnTP8G+Atcja1LOQ9akPGbQr0k5B1mTch4yJ+UYsiblPGRMyrnImpR7hYxJeaRwWYFzbjspx5Ax\nKcfQ20m5RzjIVRnFcBgOTLGVzGRqpK9VXZAt8ULHXm5OZNwft1EPYAPiKF2tTuK3Y1XSbICcrTMh\nzunF9WJMOc5936fSXcbZ4yxCC/ay7p7mT/36yDCghC+5vnUjrUZh+PbcLo6TJ4YzxJiZLWcwfLiB\n9takaXERxAS8CffoOOyXAVGTOEHrCY7PMD5qt+OZJJk/TnXlYz7AoBnzARmGfmh8zjUhEaRL6DvR\nvkj1YQAJJ1fms578AvODSFMN14gfr84EGEXuE96HYRVU0zyXHjDY/yzxsJx6wDLm0Q1OyhUqWGzr\n6ztZgcj7wHSEijEXQXLglGpBJ24iBbNuRwWwEftlkALdzVRWtTToagWF360oCfJB1WWb3Y7iOylO\nkXy61N3R8lUg1DgFrR+FNHAGmNpfY/WZVFxS/xoRJ/catRAFCs14OVEd08GzErZliLY1XurwW16X\nQ8xveCbu+FCz6+kmsRFtgWSCk+nGprNiUSoLPmNobQUQEdl3RdIyH/2k082YuI12TeuhqiJXIjjV\nLPN68fw0k+b17oHBnMxfe68yEImJrfjKWjfPxepjR5e6JWTgkDo8nl9e2gkt596Dr+WdJR7g08Zu\nSLZz7jPOuYedcz9Q665CGPZG59y/I9ikQoWxA07KFQ4wHCQh2XCD+52IfNZ7/wKse5WIrPHeP+Gc\nu1JEvPf+ksjmODC//qRqRQwv/CyDIWxXcavToMcbwtcX+WPSIc4DzX93gW5NpkGKn/tPZA+hne5s\nSH1w6cx/E9Ydi/Uh0dL5akPq3JmukvShBw/LVjUmhubOhnFmGBeKDhc3N7vKCrr7WcadY7ibgGPt\ntV4ZTKQUcUW0CEEcr0jalapW1oKM604vh6k0Omk3gSzNHV3hjlbrML6ghzYVTXZCxzzlFWqbgpLd\ndeo+MA5nOkU7G0x1pvqfhrp/lvKg4dQGqeAeblWeNrSFMngnpDWIXHNKXBOegxXUD1MKZJDWT5vb\nhANQV41rutfYfKafpbb5XA8Y7G0lJrszxy5j9t5/U0R+Y9Z9zXtP88E6iYf0VagwilHgo1FhDOIg\nYcwiIigftZKM2fx2l4h83nt/a2TT/avcDoEhtGhH/FLzmAHB5Od07p9I3SmdZI0jfxSkSbTOa1cl\nsrrjTR+ru9Z6RutOxv19Nmm2gcXoihDTGDNL1oKHkRbx8a9Wnb+aNKlqM6xa8RWzrMZNFjShg9tN\nhk/f9GFFYsaTqeLcWRQgi0mLNCUWxqRQOAlBIlrfbZ4P3nfm6mkXYKTBwJ9YKHMKuP/b4GFxjE5b\nSptEo8B+LCgVQMr00D5+Davn6kAZXpiRCtYBWCtyIiWVXFtLDxjsihIP4cID0yvDOfchEXksY1JO\nhVzWarVSJXAqVBg5dBJMUqGXGJmQ7OKVuvuNEWHMzrm3ici5IjLHe/9oxqbJgVcbvatNvBMFQ6eh\nHzt7Vfjlxn9N2rf/ECuClb51k6DHEmn6m1oE/ZmmmtY/ugOPEWYUDFkI71Q/8lv52dblTbfmj1VE\n5E7jiWL73qpIABPTkKzQ+4BW9N2qb6EEShbwpKgjYU8hP9YOwARELLD8KrTUoeuk8tTt0l2dxNBR\nOtCeLrdJecBDh9WebRTfChPlJ5IuJcXzeQjL0etGzyW+Wkx1zodbB/rweeqkGggu0GYcZyb1yDoX\nKb09TCV4jzh0+oFPjJwHPTmeieUgjcCjZqs6znTfAwb7zyUewneM7ZBs59yAJBPzcVieJyIfF5FX\neO935WwaP7D2IeU9p/P7l9DmqRosWHOMkrueY0/Dwm58SY8oIZ5+GPv9SyzzuWcI+AOqLyaJ9TCo\nzfYMNGAOCGvw2Q9gNrBlVAlQjIQq5iZ1HxZxvDTcQbURTppuYjp8nJGJvIlXSWkE418Xz2iLERYz\n4mVQCfCb+u9odf3NlA8t1UR8pDkJ6gkcO7gUE1nIwoeJdA8+Ui2T1GukFV+V7nGd+p/60ve0dgkF\nepUKYy8MwBPoakjVFFlO5L0Yj/0M24o1UJENY3Id/xy1ESK2tuM+TCNrikUWkgDt6MFE+ckSD9J7\nxq4qwzl3q4jURGSyc26bJNnjLpXEK3a1S9L6rfPen5+5k37htJ6X4K1woKDngQ1jAJyURxViUcDd\noP9GvaLoX4DJQ5Dzjs05PisHM5nR2ehrAx7uUB83Ej+K8/zYX2QqOYg0We1cMlewoDLGmino+3Us\nz8Rx1qnjkOkvK3OtwT4fwn6uxurruQ/1TS0SiisiTQOQSJP92BwJEOf3qDj0lBjK/XwL7QB2pVyi\neEn5/SM7nZtXMdoYqAKFJVNfpfryt0ZkPyKyHvclVmmEuVY4FPom6aoej5ltUteWao/T1bqbTB9K\nEtRZ3ylp0B2O1wPX1CYh0rBBQdz9SzhWpSvZB1aaCvEH8oKHytTjpCGPt4UJflN1J3PAfWgax0s4\nrReqjKtKDOYDY5cxj2k80L5LhYMUdlKu0B/0POdTxZiLgDHSWKSrkvIhHQJrW4/luYXcaroA9X10\neSPtU+z3FvzPsGdKgG9Duyl2PRkWjAdjA5gfSWk0MRLDk9FpI9joF7H6Ml3hgg760OeRpRYRyalr\npLEsalxk+koek25bsNY8hCRM2mM95R73ZrP8ufZjS0Eb50jN7LPAe0gxeEX73ZJd613xGtJ4/HSz\nnKppGAPHQMYfGfMGMIS8GospMAkW3xnrohnJ1U3vr/OwzFeJkmUs3WipdLh8XiHt7IDxL5ayNwBS\n4W68U0fQ+ofnbKOq9HJ8LxjzR0tc5I+M3QCTjJDsFzrnvu2c+55z7rvOuZfk7aNChVGHyvxwgOIg\nCTDJCMn+qoh83Ht/j3PuL0TkA977V6Y2XolPNPV6fBn0NM6YjQZaukLZr7vOuUsxdBq9DQbQUtlM\n1hUxLNDl+rdmvX5Rrb6VCcL5iYvppXuRopJeEnQtmvvs5m97wU4m0GOxQG24W7C/b2P5Oo6N1+VZ\nqjOUmEUCcAKgL96BUN8UcyKDVuxxN3I1l3LPG0DLc+5VdRuTW3oYzJZulnRnO1T1GQmXwBa3ReM1\nxERdx+O4rJoz56TmNntx/ScwdzavEwOL+D78u6TA3Og811zvmAG08OjYDFGyUFANQS8QSgLaFbUX\nXhkfKDGYq8aujtl7/034MWs8Ic2g/knSWv6gQoUxgAIJ/yuMQfSfCRdFzwNMnHMzJHHGdPg7yXv/\ns9SG54P2XkerNvVvEd0jE7JPZiCA9fXU/rPUdVHvRmd8hDSz+vSxqQhyCWxxCEnoy1QSDuB56AQ/\njAEGA2AgCKtAr1Af54Wk57bunfWe0KZr6JY3IyglsBRa/LegVQnOr8UxF2fp7ZU+dyf03FNsonSG\nhTNcfGZzm3BPyHo+0rq8DAmJ3qs2gaRyLja9wfPHSDIphkinApIYbMHr10EAkIg0PUS4Hz5PvA9Z\n96UkQhJ9ngfHD9a4eU2zb2H2qT4sQw+Y/WdAxw/Q3nC82SZc89eplfQ9/pZ0Db6bCBKTwV6n/XxP\niRf6k2M+wMROzNeKyNe99192zv21iJznvZ9rt6vX6z4akr1HXQ8+k3n5DzKBB3sjJq3jmSFNx//T\nqEUHdzMB8IPwVLXOvRX/0KUrIgK2A/fLd37iG9WP3G/EgNMOtI+EKD6+QPoj0fpR+ypKyL/Glykx\nBIQMe7i2Xrm6WHUT825z3nHTmr/dDqHqDAQpDME4OgnC2HoYPtc3N5Hzec8w3q1wGZzO+6P0T0M6\nVZ4U++DSKBrqp5ptKOZnuaGJSPNjhEn8iiXNn0ItRBIJGphzVDFn4Zi38r7iWd6J6/ePWH15ZEys\ngck+VMno230iP9J/iPyYgfuwX6p06Nr6v9BquyTd+uh+RzUm1YDTm+OOhGS/svuQ7PNLTCTXlZqY\nnXNXicgCScIxt4rIOd773WX2oTES7nILvfeLRUS89190zn0m1qn/uTEWtu9ywKEXUWU9xu0jrOmy\nk3IRbB9hsnRJd2RoRHDi0e37dIOS6Ufs/LB06dJ694MYUVXGPSLyQZXu+BL8dYSRCMneLCLne+/v\ndc6dKiJXeu9nRzbFgckY4Gu1XgURMDiAYhaNfw+aMWtPmrzscSkw5wYeyqFPJ+0kGA63QNzWRo+g\nVtnfLxcMI/twfca9Vf1GKyn91ZiTGhVYJpVxM6QKIq/wps0qR5WJDk/mWMjumEOY47bBGDnIC7Zg\nvgUyQNZhb0Dt8tAPmn1tMNMVJneGNuQx3w33S/GeJJIVVFqMyPb6kvd0MCHMwdjWdPCc7VXvQ8gH\nkvHOhCx5b1A/kjFzYwYORVwP2xU3Dvmxu35fevC1XFRiEDd1fDzn3OtF5A3e+7M73cdIhGSfKyKf\ndM6Nk+QOvyN7D/3EG9p3qXBwYuwkIatQCvvtxi4Skc93s4P+B5gUYaDUY6WqkOSB+mLqMmmceH7S\nLD6v2TVWJblj2ExfInIuWBv1e5qZiWQEgtBgSNcho3O+TX3Qub/T25zHBrVNZkBDEZZnjX40NkXy\n99KgQ33iJCptmY1M50DOyP87C/vYEBkzk1Rdzd+YWImMv0i1Dxo6VYa0HdCXN7BcqBI5GSYrb0Aq\nCAZn6sVPUdt0YKOg4ZZk17phnq3uM6tg23eHBme6X2oXO9o5hvDcUlNapOZjqaAUC9wHD2OzFnpP\n7EWAyZklBnVb6njOudXSmhvQSTKPfch7vxJ9PiQiL/bed8X8Dt6Q7AoVsrCjitc/MJFNNur1X0q9\n/suwPDjoUvmfY04MGkh3PF+aSXU7Rv8ZM8FqFT9R68gEWLniGKzfkRNgUihU1mxDPWLw3CDbhaV8\n58PNvlPgQH89vD3Oy0vKQ5AV0uW7QFpRBq7QwyJPWni9sbjz+Yol8CGeg22svj64F+rw5wx9fcoL\nRLsgfr+1L43ctLlO5LXQLwulDbJImywpB9QXd2JYC5U0dBUPc48o2TFF8DFw3Vx4T7PPCni2bAXj\nm27HQndCleZ1H1KE0uPkxC/jH3rn2FD2/Ql6lfD5/UBWR2lKUbyfRtLT3laHo6XeOyQNozsspJ19\n6gM5rheM+bQSD8edpY5XIt1xIXQcku2ce4Zzbo1zbrNzbpNz7j3m9/c5555wzj01ax8VKoxOFPhw\nVhiDGNGQ7E9J8jVa7Zx7wDl3XbsN8tAxY3bOTRWRqd77jc65p0iiIDzNe7/FOfcMEblREvo1y3v/\n6/QeTkgOPISacNdj9cUlxhOSvauEPmQguf6lInKG+iDejr4Mf2VIcLAoK5/bjXDvon/mmUwqTm8G\nqqC0D2ibOoDL1VjoITCR0hAs/dsR4hpcWLXvM3Wb1mpOuyvY7/A/NH+iHSQkG6J3DJiurnbSNgR7\nEVp9nhksdxX2S06hhcOpZNGNpKGaL6YaDEn0eUHAsrYgBJk67ZjtYq1hv/N4z7RmjzpqMj88A7dA\n78q4qJYQfTJ+bru19Xgnx64j79EMtBdG+hiwSANTal6K5cuNhCkiMh4sfjtY/DTo+EOqWOr8dXAQ\nrkPR4BQRaT4DrKrSQJsj7dDbhjm+jntd6+/r7mr+3xMd82tKTC5fHZtJjLz3O7z3G/H/70Tkv6Vp\nabtGRN6ftW2FCqMbea6CFcYuDpIkRmEniS9zXRKXhzkiUvPeX+ic+x/JZMyIpR2GjyQ9C2LpBy2m\n4WO23SRvEclIoRmDLulj9GJ78aWOVnamsbWRNOvBKoI+l8xB10XLqB9XqnQSGS19SiNRWcE3NaNU\nz1p1nZgsyp4jiw5ox4FOfLZDAnYyMvo1k5lFIhuXYJsQimt8qlvKXVEfSUb7B7MNWbCuLk5GTJsB\neMRKsOwFE1Xf30kroHu/Hqkoz8u5JvR+GcDyZEQ0XgcfdJ1bP+XfC88Q1sxzL1a/QcVCqXA1Vs+D\npHEjno1NapNrKU3a62TPrwj47Ct/7QauXVatTur+36nWUVpiClum0E3p5FvQAwZ7SokH+d4xH5L9\nFEkm5b+XJPrl6yIy13v/W0zML4kpw0NI9g8HRUSkdsELpFabJK0uU3x44I613VjLpyGYZN13m+uC\nWxMe8PV4wOElJxN0HmMdqqxwNe4JYyF0LbhU5jMcZyOOY/MLiEg6KU6BsvCpKiovMx1iIiIn76fE\n+yxQz9oX0NqJObh2natWwuWME2Pw0zfqj7AsIjsg8mcFFrQEBbEP1XLMA0KxPpIroy1ovFQfMKq+\naMNibmUanJ/V7Br/KGvgw75NGf/oembJRcguiLp3W1We4fzJqKmmEFGh1nwWoCoZxodmfCS46gJs\nvxzbkgwwv/QR/Bhpwx6Jykdbx5LK6yHS/NBm6eVPiawzhkE+czRaf1qkvkOk/rCInD5FREQGB3f2\nICT7hBKT3Xf6OjF3G2ByiCTfvVu893c6554vCU/4vksK/j1DRDY4517qvd+ptw0hl7cNYoX2Atgf\nyJiUK+xfFJGQ9jfaTsoVRhq1qSK11U0vn6VLH653v9f+qyiKott8zJ8VkV9576MWCzDmF3vvfxP5\nOTkwxT4GPDykPlTHgmFsAsM4rohrGsEvOcVWstZG0tynWAvdd463TJDQYbc21zGPw+CEWEADGQ63\nte5nz1X/ZxkKGTjxFLOcg+tNKPt0RQn3QeylkZSGTmoAcidMG7bNMWnxeFJruwfHm1giPJzGIRpE\nv6R+uwDjC+obU7cxMH+tClBujy3j7TJDXBZszg2S00nKmBzEMqqo7kXL53WL6svr3s598FT1P56n\nI2C0/jp3b++vCvTZCjVcJpvXQUEY0wqMn5LpLNyPIahdhtUmT6BNSVN8l/CerFfv6OxeGP9eWGKy\n+/7YNP45514miZPlHFQreQC+fBpeeqIbqlChQoVucZAZ/zpEcmDWW4MNIRhMRJr+4AwGoS6wUHis\nQQj9hiGmk7SaIhLY9HroUGfTkkPW9Z92g6ZzPStBRPXQFnSj4kNSIulPwHvM8ifTXUK4O9yqHoJb\nlU74QzdCEv1/Q0tS/ImItEOiR2GDxlFbTbnFrZGdDZsObota7w3WWIcbYagewgRLZOaR68ZngdVu\niiTY4TYNLM+i8VHZC3bAlMKAm7wK8CnEpI6iwHXbAolAF5NlUicU5wkJrRroO8BKOCroJYC6fSbJ\novrv+aoP2XtWaDl94X6Y/mkNXug59t7xmdfqxl4Y455b4ob8aOzqmCtUqFBh7KD/TLgo+s+YA6yF\nX4F6aDKR2sfwTyNpbrqh2XcRFXn8+tJVjTrZq9Dm3SSGwUaqqZDxsTo29dMDaCfy+Ir5MLqTueg7\nqQ3HgA86plxdZB88D+qCc/IxX4P9vzdvv3CXGr4jacezwgi8KfY+2uxKCztJcEjUxBShlDC0vrQT\ntkjdPkuUF8IQAAAgAElEQVSO41yjHgTATvM8tfOMaAHtBX+KViXk33hz0gaJCJ4bd8JzIxYRzLFQ\n78rniR4juvYPUwcci3YypLVNuB/00NQZ0P8K7XK01Nd3lGQoD7wPHAT1xREJsjP0gMFm+fTF0Bib\njNk592QR+YYk9QgOEZEveu8HnXNHSjIbPkuSmfON3jNdVIUKFSr0C2PHE6tbr4zDvfe/R+7lb0mi\n1HyDiOzy3l/lnLtYRI703n8wsnn7Axcq3yPSmuiF+sl2X+rL1f9kb1fFOhqYAJMQqEK2F9HjdgRT\nMXgfHI8DE32x3UAK+UenwGu3Dm1M18hzJguCf+sy3J8iYfSUNFCLQC6KbWN1zNRP3tl+/wEco605\nKM3Q9wvssW19RZHC7H29IlazWZmdE0CBYg0p33AL5US/GxIJq2VfD7GNTJpR+qUqUwP6PMjai+zn\nbmw3rwsGfiP2QbdpmkZan5EeMNijSgzyV2M7wEQkmaAlYc/vFJFbROQU7/3DyKdR997PSG20B/5Y\nNOgdV2AcocYc893SKKHdd6gSgeGCWesorq5U13sBXZVomBhAS8NgkS8sJ2K+1OdndWzmfqBdiioN\nXV6MZclCnuE844xBKGdvch7vU25H46hq4bkZlQ4lOO1WnlIHMEUcr1OBEObg1sZ9DagfTc5s5nqI\n5qkm+Pb24EPojTpBJEIG+NGgPiri1sjJjUnq+BEdgNtnA/dhQKk/OtF70g2P6cEmmHwX+1TVFlbu\nsrmUbX6NZGBo+WwYI+yR2OY3kftC1RFDyaiRDDnC1Qdm+6Ot4+d1nxi736G4cQ8myjLVlYf6OjF3\n7C4nIuKce5Jz7nuS3MHV3vv1InK09/5hkSSfhohM6X6YI4AFb23f50DDuInt++x3TGvfZX+jrYR2\nIGKg3wOI4LD2XUrhIHOXc84dIYn7/3tEZK33/qnqt13e+8l2m3r9Zb5e/5WIvElERGq/GpTasSKy\nuMR4cnNNMMSUTLOBNk/cJ+u1uWW1iAum5OFQT2McHfaZG/nLakysOLGYVSJMgMAy9XEm+UmJ+pQK\nGIxSRG3BfLoQyzdd0/yJblSrilxvHHMDGN8sTPBXwNBTKAcy1EQNHDjXDsO8Gh81bSPSB8FGWVVu\ndqlra93jFlK1gfWapmRWeMkDDYNknOYe0e3vj9S6KEvsDDtQ8Xxq7J0+E8dmFkeSRxof71d9OR/O\noZEdzPmaTyWt1hwWzX8ewTDGOz4y3pGpkn1YicH2gqF3jp64y3nvdzvn6iIyT0Qeds4drVQZO2Pb\n1GpHSa12lIgsTVZcO9iLoVSo0D06mpQr9BJjsEp2T9FNPuajROQx7/0jzrkJkvhiXSlJ1pJfe++X\n5Rr/9kLZOIGfcLC7zSpsdib1eTZkFt+TvcgpWya3wd7Ih3AClYIRg1EKZOJgzuciE90NOWOg4auB\nZRpVqCucFtsWSt6dcGi5EqvJVmKTR7tsdVptthLt2Rl9Y/mYh821422Zwjy6WjCikpr3k2H09tpq\nnTyTGEGHve+zSTuOQTxF6uMZA+K1asxBGhtAS9VOJ4nxrRFYJEgmWyBZ2PvA66+TDTJLXUoHTyhl\n/x48CwzemZX1fkRQN/euE5fNW7AP/cycj3XX4Vp6SFEhK14nBukoumewvkRiFteLEPDO0Q1j/v9E\nZIVz7kmSCIG3ee9XOefWicgXnHOLJEkA8Ma8nVSoUKHCfsFj7bsEjB+xURRC/wJMbsXXK1TH4DdC\neVisAUNgIABVd2e+2/RVrm+h9hvDd5lUCHrK1UuS1lYWLgt6UpwPi3swR3NMjWZfeoIsKHNMMiUk\nHtoKSzszWx+rutLTbR69VcigioSd06uBTDbGSiMpNFuA8PGtKtCHziSBpGQE7WgmF1jcCa191iCt\na0uubbDqBlg183nTukFHm1jdQzLXvOrP9E6h6xifQYbVF3EPo3Q2AZLEEJ4RXbVlPSI+Vt3cuu38\nnP2zCgz7MPc17RNRCYzHQ9/ZNLqSdcekBlv9PBIExnD932OZ8SW0vZwBaWqXqkbCcVLXPg7P+Oaf\nSgtmttSQ7J7B7i3BmCeMXcZcocKBif4GfVUYKTzRvstoQTc65mjkH357tyTKw8dF5CsdB5iELzSo\n8lro7lK1096g/iero6IfTDD4QHNbnbyeumUbskx95dFqXSRkPAbtDUC2xTgJ63myVfVlAvMJTHrO\nMVCxeCLaFelj1g1zOvks0yESYs7qL4GN0iMllpYzSxdL7w8dUAEd/HawoOBz2ysJjR4Q9Cfms0KX\nAnq+RFj+OuO33ECrnFZkfcY4Q9ATjdX6mjLAlaJdJ+lE6XWD57ehysb/K9pL0I6jWMLnlx5IOX7l\nrFADkhpsFboodFYx6ZDwP/I77Q88dd4Gpi7QnpoULk/kMw4pc/WapOVptYbKd/+13F2CMR/RX8Y8\nEpF/h0uSmGK+9/5x59xR3vvIk/IUHFhHWA0kzel4me9gYH8HmdU24boycGW3us6hCgmj6xjnT7Fe\nV3PgC46H5w7kPaB0zwmtpbwVM4/xhbET4ilolcCyAQ9lMOpxDPzAdFDF43Y1JuZGoBjPwINYdBcn\nH84JzJIWJiW+1drYB3HXBvQEUMXzLLWOcUf8Ylmruc6DnZG/OfN4eXiGWY7k995tJu8QmzCgOjWS\nJuTlwBdxAYzSK5HJcMWq5iYLoUIYxg3IDaLpAMGgHMnZItI0EGtV2DhT+YSReG+PjY3BRbyPeD92\n4Pnlt2jGc9Q2+HiG5wdjG4beI1wDrcLClPEQHsJjezBRDpWYmCeNYVWG956apSdjX16S6L8rvfeP\no0/BypYDScNJeaSQmpTzYCblIkhNygXASXmkYCflPNhJuRDMpFwIdlLOQ4Gk+r3E7iLn0UiaMgFi\ndlLuNVKTcgHYSTkXdlIuAjMpF4KZlHuFfe27jBZ0y5ifJMlbOV1E/sl7fwkiAe+UxKd5r4i833t/\nf3rrhcmBV8N4M5ei4UfSXVOguE0xNlbRhBMvDWDMb6wrjAyg5Zc6o2hqC5gdDTLbDhg1NmL1POoR\nCvhMts2ToMGcHGD3p3+6+dMdZnvWtqPIG9ierpQCyn8fDGshMIMsVUdd4UXc+Y2k7SKoIGCdelFD\nfm1KEhRHMGkHpvWvzW3CRGJDsm1I/HXSGTKqzjAgg/zh39Qmn4CL2F5YvlJqG3zgd6tCvakakjx3\nSiM5ofhZQTX6w8L0L5fZ45A48D5r1R75WhH3RILvpGHoG/F8FcpBzjFpFRCf2c3dM9gdJRjz1LHN\nmJ8QkRcx8s85NxP7PNJ7f6JzbrYkZT+fnbef/mCg3wOoINJZ0YMRhy18W6E/eG77LmVwsDDmlh05\n93eSOM2cKiLLvPf3Yv2PReQEWym7Xn+3r9d/LGS7tdo2qdUOlRbDBdmJZWjWGV/XVku5CjHZzU1m\nWbuSfT9+vCVYXqpPlO5FYD/rIZbOBqvcAio1I5KoZgX2tzAnbHgyfmOl7mgWNmlVG7D6C92wWF8t\n5gZmwbp6H2ZlFxoZtUGV+m2yopy8zsSHsd/A1EjduG2ZPL1k8e9orhr++6Qdz0AG5kcuIvUAlFjm\nYPlp6reJrCFI7mKzzXHSUAEgDbDDASOhrIHRek7kfoTEU3QNA0NeC1adMnTrY+NZ24ekQCSak3Pu\nO4knK3rTzTAYm0Wa59yutqA07RhZ+Z2D/l2rPzKSJJG9731U6t8QqX+76S43OPiD7kOyt5VgzMeM\nUcYcifybK0l82m8ledTvdc49R0QOtZOyiEitdqzUasdKUyT9v50OpUKF3iJMyhX6hdorRGqvOT4s\nL136/XrXOz1I3OWOk8RnS0f+fcw5d6gk9PR4EXlURN5H9tyCufh6rebxY0EM1BdCTxgYbV617FCI\nDu0itDkVUgI7xP5WgPmQ2WoDD/PpfMJeNxpG6MamMr3sQMG1qQyMsXpRxQSDDhw68jos+rOxOiS9\nGWhushLMiWposiBLELRLFO1PTDMaxHewpC2q74yMPsEoSiatDXlkddSnmnsV3PRUEMFDCKJhMqGU\np8WZ6n+OoUjFdICeOvR+nIJ/1iBsfE5eOk5rs6BdoojRrAh4bobxa8nolWi30Q2S3j7d1AvMA3S+\nt4BX8bDjdEZAcrtG66asOkQvo9vUeZzZ0ZzTPYPdWoIxTx+jjNl7v0lEUtnavfePicjZ3QyqQoX+\nYuwku6lQAgcDY+4er00O7MEIHXScW5TPJ5na2fh43VJmrDEvjCzQjxn6PTDkoSOTxUmxaxR0wMbd\niJWidYXkzOAWSgfK8h705/Th5rcT5xFCziNjSlWTIGuHXne10iiRgWfmDo8lF8qTOizImOE5k6op\nSAt8xJuSHhvBMEi7wB2qk2HKTBRl02juUcRnYpZrGKueKG8JywCzdKk71P7bVdtm6HSL7p/SE9x6\nhhCazXDx36quKZ9zW+HlQrMsEp4t6pa5jxCaneHvLCLN+02/704ry2cgjKlQ5fruGeyWEox5xhhl\nzBUqHLho9HsAFUYCZZIY9RldM2bly/wz7/3rnHNXicgCSfTLW0XkHO/97vSWp+LAsM6fjg+U9skN\n7JE6LTKbAl4BKdA/+qPpn/JCTUXipZ9KwbhfzYUbBfXr29T+szwpQiQjvUJ0uDW/r+3KLOlxkCEV\nDDFvAdjwZfA2YLi1HlKKibNKOVjReuiT6c0iIiovKtqcdJzMZ2E9IFIpN6erjcjSLWOO+G43oLcn\nY6Xb+HWsDE4vEB0aT107rylMK/RbZgbUY2P32OrvAU3cHK9VI2ls1GNI+/q85jZ7EBNNg+a2xINj\nCLuKSoMpWOnmVPVbG+8a2hL+S63j8IJOn9fdsPYWaacHDHZjCcZ8fGfHc869T0T+QUSO8t7/ul3/\nzP30YGJ+rySy+RGYmF8lImu89084564UEe+9vyS9JVQZuCnnukRMvcFrtTVEqa0Q76bTHRoi+fXI\nTXBe3jnAaLMKKpL52qX60daue2AR62FViQQU6/lmFsmhQGMQJxY6+1MdEVEB8ANDL6SzWTGFBlEl\n/rMuHCeeSch5PAx1QTRUuMCE2Q42M1oLrAG1A6zF/ulmtlq9X69CW+T9pKjPDGic6Bm8cUSRYrhU\nLXDC4f3XbmIljIdZLpQ2T/JD6pyPLRPZZ4GAn/UILGKmPh3rMROz7DBm3vEMjKG6jh+amKqEfbJy\n1bSg+4l5Q4mJeVb5idk59wwRuVGSF35WNxNztzX/niEi8zEYERHx3n8NgSciSUJKm5hglODR9l0q\nHJxY31f1YoWRwhMl/jrDNSLy/m6HKdK9jpkD+eOM3xeJyOfjP9G3KzHsNAuA6NBQiEzTERwSxFcY\nC86LGAtShkL0eSZfNob/HCIpXWKRFANBXKSvK5zlUxUotAucraiMc/dIiBQMfSJyOqQDG2bNYILd\ncL07IjI2qmJC2kqK6GAxO5RhNSsI5bdk3ZoVoU8Qi+mexRDqyH2ga17IVIbv83ww8+D6qISpbQjP\nP4aMmYyzRAKnk1lNBS6QRfJuXxeZiKmPZCDG4THDXTsUGTfE+JUgCgtoCMNzNaTyqNDWR/e4zVCR\nnA3WyoCoD+n9g6mS6X8Tq1+ONpWkSQPP3DPN6ljiq8zE8qdl/SCyBqHrc6BXiZFUx2RIkarkZTGC\nkX/OuddJos7d5Fz3H/ZuAkxeKyIPe+83OudqYkQN59yHJAlAubW7IY4UGv0eQIXRijFkJDqg4Z7T\nvk8Z5DDh+gMi9e81lwf/3NVspKFzbrW05gB2kiRu+7AkRpS55reO0U2AyeUi8hZJnD4nSKKNu8N7\n/1bn3NtE5FwRmeO9j+oM6vW6T6rgJqyiVnuD1Gp/IrJ5SbPTzIy6YUzSswxjj4U0X4h1DAQ5A8tv\nQr/Tta6Zur8OHPTpOE+n+VDZJHJdmU7xz7E8k/o4pS9ehD4oSCwTMyprFwGT3LwIy4+o32iwG/9O\n/EN2XSSJlK1sYdwNRaSpj2YkC/Xq3D8ZkA6osFVO8tzzYMT1eF6y1Ie6TmFWik0a0X6a3r3UM7aZ\nh23uVr8HIzJjvDMMY61GraRlEBODgyayIpvWDVMwxXN6N7KvzeNzhPsw/N3mJuGcaWD7g0ShA1lo\nkwuGVUq3nRjdcdwPq2mAQ6AXZEQKqdfrUv/qK8Py4JXSfUj22hI65pOL65idc88Xka9JkpLCSSIe\nbheRl3rvo8Wo26GbAJNLBaZ259wpkkT4vdU5N08S9cYrsiZlEV0Fl+rpP+l0KBUq9BYRx50K+xe1\nWk1qyr669IouJ2WREZOEvPc/FJU83Dn3PyLyYu/9bzrd50j4MX9KEo3Tauha1nnvbS5GCQETHozK\ngVHF9FcB8C5YZnIZtCRtgQdECJkG6+LX+WTc7Z0PNDexSZKYWIas8hidEN54Q4TwUjBEMuUb1QeX\n5k86ZwSyzn3NbPa9yYyFFSeo45z3iqRd+Y1mHxLLY3hdoLOz6SAnSBqUNsI1hN54q9r/9H/CP7yN\nYMpBT7wM65th6Ge5RB99q7eJBenREXv06A42kDR7QGFjuv87wJRPN+doK4WP13XjsN/gnYIx0N2M\nxH9Oc4tmFWgeB54jd0eeU3p9tKsM3eL1AzfCrECfQfUcUaX/Y7RMZM9zvgoeO/YZEhFZDo70UfTd\nafp8Sf0fvD7gwrcD7p10fTtLuTiiTt9vn58s/tG7sJ5mE4ZkX5YeUrj+WcnKUilRu8T+yy7npUtV\nRk8mZuTCuBf/H9ume4UKFSrsf+ynidn7FBspjT6GZJPSwvLOr/7TVZcjyHbg03s2zM23gNKsh8V6\n9rLmNo2Lk3aA1yYn0bhFqv5dAZyIbdblbQM9G9MzjoNnQqDQqso3fTq3g/1kVTy+Wn2Q6bwfY0pF\nEQJYmGjps+pH6oeZ7MkmUKfsH9NP8xzpq0rdcpHCNhQ1yHC1pwt07rtx8lR/hpBpSpZKpzoEJfuk\nDlKFhmT9Pr5eROQMtLwMTA5/LvrcwGfyxOY2sTqMnSKEsmsRw9hNFqPP+7DMy5Rb4gqixBbotGfo\nvtjBjbDT0O+b6uFeFFVI0L2rw6oSOub5Y7jmX3c4JTnwbojMfH+mfEz1ubRli5QjfTQggdnkOCHf\nm9MXiOW3EImLWDTSNLB8/MfxDz4w96n7eVLEuCciQeTfB4eVlpedkzTOPbjh0fBTIDglpZ5QWIjf\nVmScKwNOdGa38MGyWf0oW9Ml6kJJg1nZOO5YxBzuzW58DCjCpiq8lIg4KwNeEx0zQtVXmetOVZvH\nB4DXkhWSfq+6WjUTwVwo9PqboX4bR1UVZ9MG2jzjHH4bgmvm9Vh9cTfvvYqm9Dg5uuj9FdrZZv8t\nNTdhXVwLCkvXPYZj6EtNm+W0HkyUK0tMzAv6OzF3FWAikoRkO+e+55y7C8vHO+e+jXXfdc69pPth\njkGcNLl9n/2NhX191jKwsH2X/Y33tO/SFbIm5QqtOKx9l1IY+QCTnmEkQrK/KiIf997f45z7CxH5\ngPf+lakNz8DX63Z7fJ3vgWIeXIZCsUm7jXJiH0INvkn0gQR7qUPUqv2d2i7D/E6XNapVWuqlMacB\nmF8IirBj0hV/88NhtyuH9Gne1AzcY5gsRUX9SSUTm3xu0g7fkLSMAxnm2FRR2XUITrClnfZGXMco\nugZ1BwyQlEaiDN2oNxj8MMiEGqRFZTKWKSOphwojkwRl5J4oCrrZ0ZI/8a34B0zf5hsWkTCT3ASV\nVWD6RbLysSr6Vdld7HVOZdQrk1GxAMBy/wtCw/NCZgWteqPBl0UvTY7uvVckbar+oUhTzTWAdoRD\nsu8owZhPH8OMORaSLcn3hvLfJGk6so4yVD5RFTIwPBoliwpd47ESf31Gt1WybxeRj0kyEb8PjHmG\nJJ8+h7+TvPc/S2/9FRz4temfLELmOaPvY9hwSx5cG/wAhExccPUZVpRwPI0ldFIZat1GlXXLraeW\nCbC33XA7ynMD4qQwnqyUBrAiRiKydAbMNFp/XqImnL9GexzHYnTbUZD5sS+NgAiZX6n2zwxi00tU\nDQ/MqYGW7nlgvSt/0Oy6gOOmXpvnvA4tWOsWxQvI/GdhnBuwTClBR/0ez+tP4xkeggYMzgy+aNGd\ndiF9zqGBEMu51w1jWQNpkEEp0fqAwK10dUMfBsPgkfm0Sqrwzsw5Ae/qsArtH49AuAtx/YObKt6l\nlftajiMiTanPsuhLTQDWglerH7/a/dfy1hKM+awxmo85EpJNvFNEFnvvv+yc+2tJ3tq5dvt6/QdS\nr28SkfUiogNOKlToM8KkXKFfqG8SqW/4cVgeHEyHSJfGwVAlOyMk+0si8pfe+yNVv0e895EkRxcm\nB159TbI4l0mZlI65DtbDyh+BTdoxK91jqvQ8dYwFUlVSZ0f3M2tZFpFmOk7qSOFOshPeJYXcg/A9\nfAhPSos3CH5bj99m49JtBTtK1cETkeMw7k3tjh0rB083NqOL1SG6022tQlv/LgKbfpO4AOuXc/2A\n+rHR2tdW3YgC93sfpJHg4YJz3f1gsyut/gMMc6d3g3X/KwJ6lWidsH3GeJ3glbMRodLHx86Hkh5o\n+02wASyK9I1WQimIVNVqetqo+x8SgbFPI2myAkGKQEtTrJ6TVfFlGfq+Ra3rhVfGihKMeeEY1TF7\n7y/13h8DZ+q/kSQH89ki8guEaItz7lQReTBvPxUqVKiwX7CvxF+f0RM/ZpUr43XOuZeJyLWSpEL5\ng4ic773/XmQzHJi+qfhi71TpNabYBD5gIOOg49qHXaxRH7c5xq+YrGsAq6mPy3WoL4Brsd/F3I/V\nbb9GdbbWZnqRbEGrnVVRr80mR7JVxGNVrFMJlMgIkZhdB6UwwCCQiDL1/AjrP/sB9RuY5Dgccx8t\n+vQYiAWYUJ/ORPLUr+IZuVOnwOS4GfTC694w+zxF/T+ANuLLbkEPIIblMwFUiuVp32pKHwhYSdV6\njPgbp+o0lqh4zdQBUQbeBoH95iXJ4rVjLICp46iR8u83Pvznq2fvuoLjba2s3T2DvbEEY377GNUx\na5iQ7G+JyMHpu1zhwMD2vr6TFUYKo4AJF0UfI//ImC3TfLPqAk+EVCIZgrrlh9U6MjGwr12wysOb\nYgV8hhfmnje/VxGLePAlRRRWCLO2zGNAbdRImkwd3RvU/230nXsMMxERmcnQdcN2d5kJJtejBNdr\nO65XNBQcOuaHFiftcqy+dhD/REKys6p6p5iiSGCUexClxjxTUW8DU+YqsEdb9Xlqc5PVeE5s8nx6\nZYxT6zIlKnqKXBf5jRFxnAEaGfuIIIRTM1yfz96dsd6AqZIdrsG0Zpca7ifTl4bQcqYxYHrhIoE+\nkAiGVLX1zCrrgPUG0WhXa3NIPb+TesBgl5dgzBcctCHZkNHP+kKyeCvHocOuGPoDsTgYlDApbsek\nGJ1E6J9DY10kXJgBE3TtCoajvNwPGNPtODbn4zAGfGiuUdnrzkFLNyGbb6EFtq6eUUeEF1jHD/Ml\nLqGGWIH9LOTgbiq+bQBFd4roSm2w+7ykZaj9f6BdGDt3BEZsga/YjKzro++hrQ5CVQxVAfhAxnJ1\nBzLAvvdKcRjDpzZqLUBQ0xDMKjR5Yz64FKTg8jLvnHbHo338L3k8mw+GhEIZeW+FJTs2MYqI3I79\n63Bx3jNqz85kzu7Yx6hdHUi8h2erHNF85NaiXWLGdpk6ZxoAB3owUV5bYmJePEaNfyIizrmGc+77\nDL82v73POfeEc+6pWdv3FZv6et0rBLyjfZcKByfe0r5LKRwsIdnOuZ9IUg32N2Z9kWqxODBZEL+4\nyjBC9zjagjLdg7QYlmXYyWPBFtboUQA22KWl7DzWBYMhxUeyPF1Vos0xY6HAzP38djrk05XPhoJr\nV0JebxppSmThK4KGuXc29DtAhw9/Bi3HbdUGOkqB7JCqIyavooShw4bbgB9pXTSIL+dDaIM6xTJE\nXWsYyZ2YQzvkijaqsTnq2VgD+rjr5qTNUzel3EV5zpR2jIG4DDQz5znPoruceUb2qFI4DAenMZpC\n7oBxtYshhPhjH/mqt+6Z1NUlGPNFY5gxS3KxYvvoWbXYChX2O0YBY6owAhhD7nLdemV4SSqV7BOR\nf/be31C+WmxOJeHx0KMeA2YZ8toyEQ7DlDdKCgxkuBbLuXpd66JkWOtedR6pZCwDSUN7yNNoXIkk\ntwmudTT2MbJJ64apE8/IETzrrPS6t1sGBWaWMjYqVrwdrGcaaGI7Q0weyLZ0SDL16aFuHNkuz52u\ndVqCYYXorDSW2hhrpIEtYJyZ+mkF5CQ+A7bM2/OkRhLwVK5uSlU/lxRebpa3402nTW5N5HiTn9K6\nfBGOd7XqG5gyWboOrBIpx5R5P6Az1/eOKvgQ4NPACpzzRC1dYdx0Q+VtmfRTtDlDOI5Ga3Sy0kLu\ne9cBxtAHt9uJ+WXe+//rnHuaiNzjnNsiBavF1ut1SYqxJqhCsitUqEDU63Wpr24uD17eg5Ds4e7G\ntD/RM68M59wSSYSAC6RQtdjLcWDqFQvoOFNBF4RK3J1yVWIAgEmsvk59L1L6z4HWfbSEJ2ddL7CX\n6YkV/Ds/af5yQrjGpA866EFEhu9q/p/ppkU9OqWDiAcG3Yv+Bst35yR5D0noqQsk2waDPUOd8+2G\nke8AO6U+Pbhg6XLzKAJH9z/qxunNNoPuWiooJTB8m5A/BuMqFgB3y2EUIYhdz61mLGT3E3VFoKI6\nd5U0eAieOsGFjDpy3t+88ykCK9nRg+mTpt+A+r9RbNdr1f3mo3UBLxBsExvgxjhLZ1jgs2zuA4tP\ncDKMhdVTjctsbuORtjaIny2uo93rfJeU0DEPjtEAE+fc4SLyJO/975xzE0Xk1SIy6L3vabXYChX2\nP3psCK0wOjAKdMdF0U0Soz+RJGmRl2SC/5z3/krT5yci8pK4VwaVj0xADp1Ui1cF2S5ZSVbSHM0Y\nyCJs0nAwrMZdrYdvga0xFwOZJb/qZAr0eCDdKBBSS8Q8ONpiQP0/1NqSrdBRkVb2mGfEfAaAYDkW\nzOd8U3IAACAASURBVJFKr4pjD+GeBX/dQbVREe8XtW8Rkakv5T9ocW1D5WvN1HCuqXJjnYD61gG1\nzur4MySvQuyUQVPUmc9q/rQeLJQJs/YaFqkT3b0XrfUFD9IPJQ19nXCPbkRSJMbk8jGN+pW3g043\nQL1/VqkvMOrJqur6LqunH0Cb5VElIr1gzJeWYMyXH6wBJltxkYJqgEYhnQGNLk8ZYmtmRZMYrMFN\nRPYgv+9EFljj/pPAiR0umXyndnuNQt0+E8wRIrV0GaqMIqVQF2yEjHO81wEmdLfLcvI/qvnvID4o\nS0y+X04INLLsVM/lFNaaw3FChjJ7XbSlhx8LfCD3IXhkHFQn637a7MoPxiD2awMOUpU6RML9HL4j\naamyYM3F3PJNNJohR8nV2MdFz1J9GjnbZ4ETDVz4otkDLUyBW+tCFoUlEFTlRZh+qm4iMYBWn3Mb\nV82UG6BIUK/sxMTPNBonw6i8HtGW0UyNJDN452+Cikxr+pqqw+4nyotLTMzLxqgqY8xjzw/a9znQ\nMLirfZ/9jUz/5n6i0e8B9AFloh/3EzLtOR1iDKky+jcx24tumY+INAMwZkgUgSlrsdmWjOLnl2wS\njHSipBEYbMJGp3qyV/0JJ0vhJAdmGAI/bO5iEfkm2pNN2HM0KxhZBNQ325FRbdrp2CQnlwbzErBS\nMdnvLoz1dN35KGnBhHNbl6fEWDxUDJNeIHFEjIwbwZRD8nlMALNJSJTxzzJlYmLOdRpvrscf2X76\nPMGUhxCmvAvtRZH9r8f4jsUyC6HMZM3IWGkySgGQQo49zfweqUO4HM/9BVjehPYwHF8TWl7eKe+W\nVuS8xouONit4Paxkpq8TfzOumyFEW5HJ23DtpmD5s/jtmTlMOaQVwG/bs1i95EhnHWAMuct1G5L9\nx865251z/+2c2+ycO8E5d6Rz7h7n3I+cc191zkWS5FeoMIqxvq9SbAViqMf3YbjEX5/RbUj2v4jI\nvd77m51zh0jCQy8VkV3e+6uccxeLyJHe+w9GNm89cMyhPvTEbzRY0KhVqIJDRrVkXXDzS2jPhN52\nDhIQsf6aTrnbsMfE/odRQSPq7kbjJXXBZkzb1FiOYVi1Ca4Izv7UDSsXwY1ImsNK2taAFypeaL10\nmwQ+LbXsuJ2poxhCzCOub1Z3ahECNuY3161Dnu2g3jD1IDepWnPHGclkFfY3n9vCQHXZPc1taA9O\nZfcjFdTsEuyX2QOhQk3X9dMBUiZRlg28QWBLS2R5KrkQ9K1rcE/nFHjGg156DlZkGeIUgotjiW3I\nqtcoldgcSFoNvCxk9VYafEg9TzRZvArtg8aGQVystumFzvf8Ejrm68ZoSLZz7ggROdl7f7OIiPf+\nce/9I5JY6mheXSEir+96lBUq7Ff8tH2XCmMPYygkuxt3uRdK4ov2X5Iog+8Xkf8tIttNzb9fe+9j\nGeZwYOj91kLf18L2WLEZ3hlMjDOQV3XBMjU6+dNcXIQZZIUEKwSWYhluTpUHIujNyNB2qB+Nmx8T\n3qTCeLUHBI65CQn+jssJLCFCXmmz31AFJaZjBiO+5h+S9r3UG/M+FKmEXQQDSbMeE+Rs+okpdhqC\nUejpwreJEkVOGtPAZMtU8AbCM0g3zxwXL1vfMARI6UAWKq9pxzDPnJbs/gnte809ozRFweyZ6rfx\nVFI3MFz0tW5y89VxVtlnzeZM1+HjcAulVw9NMMcaI86CPc3/V1p3OXqTYON4IFn3DPbcEoz5hrHr\nlXGIiLxYRN7lvb/fOXeNiHxQrIoivSwiOiQ7CQisPVWk9qIuRlOhQoUDBvXNIvWlS8Py4OBgVSW7\n0IbOHS0i30YxVnHOvVySiXm6iNS89w8756aKyNe9938W2YU5MDwrLvj75qrlWWOz6Q4Vgs/rG7GC\n7PGfTUetv8yp9pyFrMoMDPfVXiepdI1gIMMPmPV5gGJyCKxYq+Omkp3wO9tB2scQmMPAGC2NWPZP\nXfJVUhiZlZ11YQSEUQeGXqAaN2ETNrG+YczjIih5eZwikX5ZVW20XpnXHcUfwrXMqYgTdOM26b0p\nkCCi9MLmnJisn4LeeCXtrII+mLp3hl7fj9/JvnV+cpZWZNKt1L2LvDv0SqIXCz0gJmXoj0WkKd1S\ncjSJ+Htd829hCca8ovzxnHPvlkREf1xEvpJhWyu2ry6Nf/eKyLne+weRK+Nw/PRr7/2yfOMfw8h2\npH9KGcuyoPML0zXpA6YPckz4z6qBm3MOqoX5ret1MArH9BD8n4+1JYwiCBVXcLxQnJMPosoOtgpv\nw3wzeQc3Joq4C5vbUIwOwSE28isWcMIX3lS72IHjZ5WUj4IfSD2xZfnDRqIq6Z74CyzPzzq2vs/U\n/2ZFZ8LFcrXyU09lzON1p0hu81ZLWj2Q8qlVLpRbENVG42s0mCIDjHzkm8MYJ/0oMgqQr0MqapXq\nL1VGi6o8ugTSjZAkgJO8fv14rtSSjSuSs6QdlKVzDUhFMGhm5TuJovuJ+ewSE/Mt5SZm51xNEt3r\nfO/94865o7z3GdFi7dGtH/N7RORzzrlDReQnkhSNGSciX3DOLZLkLXpjzvYRHNa+SwrWXzQHJe5N\nEx2MaW2J+8pJuRRyw1czMLV9l65QIkjBTsqF0IFRrpM0pptL3Lst32jfpxs81r5LGm3qRsbQyWtX\nBmu+277PSGNk3eDeKSJXeu8fFxHpZlIW6WvNv0NwYIp3OeJxqgw82cq9Zlmvo4hp8z1rXyUjyqbU\nEANoG81NyK4ptZJdTDLl2lsy0rUzMilGuB1udyF4pkhGsTY5PoKx69zIj1bFkwd7HIxl76eSNpoz\nl2H13AbiagjmiW1jgiBYl+6MiHqIc/WxJdQ5dMNalvfs85kiTaU6h/vlTKZLY+GeUJ1GexfLP86L\nHc8YuFNQKQq2wYXuCCxPMkE7MVgjL9UeC8xYdM4SJjCcR/UKzxUS2BWq7yWQ/q5AQIktursIfd+l\n1pEfTENulJWYtPkK/bnq64LxsnvGfEYJVnZ7acb8PUmo/zwR2Ssi7/fe35+/VTYO3pDsrJwUFSrY\ntKwV+oMwKfcIOca/+i9F6mpKGHTp/M/OudXS6uzuJLGVfViSufRI7/2JzrnZkhgatPtNKfSRMTNB\n699iOVLFOqgocC2G4MQewjPJMi6VNBimHQudtSBDYwh4EZc6C2YQ40dSGcxCEiOOG8xzPVylZmuF\nIowpjD6z+koyoIZa9zW0F3dwL60ePASyvK7ZZx8y8tHoFBh4HmPL0h/CQrUHgR8t4dYDaButm6Qq\nsWhQZAGzvAPsi7rTH6qutHTQnYw64VTQiDRZLz39J9hzj42FrpnWmJjjdtkOmrhRFZ7KL7LILEeM\n4sENMuMZ0dVCmEt8JoJPNkPVFsupHKRMGPmGEZDzS/xOyU8bF7MSNDE0m0LKdJ1D4N+7Z8yvL8GY\nv1yaMa8SkWXe+3ux/GMROcF731GCmp6HZGP9u7Fuk3Puynb7qVBhVGFP93NAhVGIkQ0w+bKIzBER\ncc49R0QO7XRSFul9SPbhkvg2t7dOrsPXK/X116441EuC0U7Fl3sHtom5plkwFPTYjFBnEWnqTsm8\nz4/0aQN+YFlA+Gz128p211hXWqYempQP+r3NYCIzmbM44kkQ9OcUxYu4s1m3rDah1CI5111LJxn5\nmO/GtvPIhgoYqkJQhz4enokhPPs2yU1wLYuJw434cVpC47kd7sdmSDdkjcEtc05zm8CUKXkV8TZo\no2PWYckMYQ4GzSypUKticJ0uRbIk3t5Z5nppxvxFtGcbthuYrn6urBcUkBXAVAraWL2j+6/l/BKM\neVVpxnyoJKLK8SLyqIi8j+y5E3RTwYQh2W8TSUKyRWS3c66n1skKFfY7jolN5hXGPEYwu5z3/jFp\npWNdYSRCsr8lxayT5sDU0SrHyrVgyHQZgv/jJlTfPi46dlu5hIAebt/NSaud9EOY6vuT9jiEHKNp\ntabnBAuIZLA7oFQKQzCd+8AIcxO/AyuNji54McAnetsDzb7HkHkzzNaezwvV/2QuVtqgXYDbKk+R\nThhTYOI25D6PxUO3fCM8FkLFcEo919kNpOnHTEmlgA64UBi31fVSfKJ0k+MPTD36t3k8tONUHxJh\nOobMxPOakox09Wx4UtCWwOQIMX2xHUvQ6Rupapcik1n393r0OS/yOyuyc1a4BB41++DGwoo72oVv\noAcBJnNKMOY1B15IdiHrZLpK9sNSq9ncsRUqVDgYUf+uSF3xiMFre1AlewzlYx6JkOwnSTHrpDkw\ndWw6/NcEUZyLj9gNOVFYwUOADPxzpq9NaiQSvDCuwf7fSy+JnFDgVJgqvAM2giVF/XMJ6oLphDtN\n/UYGS9ZjovfINqKBE/CE2AsvCgR9fQd13k44SXVlIGShKhFglhuhZ7XnFvS5H1MrY54yZUFvGcQo\nbft086dUaPfLzHIswVUWyJx14ibrWQE/8ly7hvWvNz7nmWWeRNr7M4v8DJLiM0nWg9RnpDgdIDPT\n6PKL2GUsqH+GmaPpsy8S/OxX4ll+GlbzESdJ1X799IahQ9YGE/oNP+2hI5ubTPI9YLAnl2DMa8dw\nzb+MkOytIjLNe78E1snV3vu00m4YF2k8q4DigWwxwGBs07Fuqx0rXK+uVjl3LzIPwvQcIxODFGiF\npUtUKrevMs7twuRE6Ze5MmjUog9KXY3VBlPYe/4l9X+op2bruLV/cdP+t9b2oHI4y4lo8eHKVT1k\n1ZSjioEiesQgeTb2e4vJ6bxaXYO5Rj0Q3AvNNvvUNiE/8jn4x7qIUZWlAkDiWcuU6K4ktlkImPgQ\nlsN94YcdA9igQv2tQa0QcF/vvCJpT+siUjGmnrB1MbPcAHW+5FSNwueaZUVquP/vmK50lyNLzc0r\njffrTrxb8WvQ/UT5shIT87fGripDJB6S/XsRuck5t0mSb+xbc7bvH4b7et0rEHPHte+zv8FJucKB\nhYMhu1zXuAlfr0UmGdBZasK8FWMjo2EgRcjoBuPTumua2zwPbchuBUa1C0a/jtx3lNvOdry0HDYN\neaEiR2T/F+K3T5ggDu42OiYrFluoEvLDkBhsljpWzGA64wFdKw7uUymDVCyBFINFGBZuAnB4z27V\nAQGM7DgPLQ2FebmuM/L+3g52qkOyg/qEN5zjLZIpzoCeWPqUn276jH8O/snJs00EVzoTgBNUZTnP\nII2M/F7pAJxUoBKNfsZlbZF6h24yBrwb8fy+3Y5hQP1PXQmfQaoDKSHpSjaN1t0sx7Ev4P5Bodc+\n2OxD4+WCs/APBeqYNBgyDHbPpGaXYMzrxzZjrlDhwIOdlCv0Ca9t36UMDgbjXw+QHNga0XRWNlu7\nLqS35HreOO0qDWUXWVAqjSW30fkO+X0yOlI61m9R687w8T7HMYkeKgrHkiVtR6zroVhN8nWyqsW3\nC6ZosmhWFB7A7wvQrj+ruU1gsDinrHDuPASWx210xj6w2+Ww/lxA4yivFylQgUogmfclhoVoIZ7s\nUca/iUYvvQX7nVFkv9Q7Uzr4ZLoLdac0ZlEaCQFLZWrlETQGataNa8vnno/lGWSRjWbX8PzzWaOB\n2Abz6BzXd7XuJ3Wf88Dr8p6cPqaSdhb2RQjoOFqjacPJ3Uf3DPa4Eox50xhlzDDs3SbJBOskcYn7\nO0mu8gJJ9MtbReQc7/3urP1UqDDqsL2v72SFkcLBxpidc0+SRFl5giRKpTXe+yeQJ8N77y9JbbQD\nXy/LnHTkZfgtK61lrAI2v+6GKRBkuFpczdI7hwQ/mp1SucyS3bHQ6DaglMBdaXWr1Z9vgv48K/GL\nRmCNRrfJBDbaXS7XnU+k1bvfFiyAK+IWVBxhOumsBDki0mS/vIdgmppJjfs4/rEJrWL6dht0gmfk\nPlj2YwE5C3Gsi7BM9bStCCLSDNbhy3waXPGHIPVMws27Q9WyC54b0JVfDeknVUVF2Qfug32ABHke\n2k0515JSFCWvlDeI8r7Zg/FONBIYH9/zXiCtK2Jg6txYojEw5jtRtSV4VGR58sRgU9sCy9SzcXEP\nGOyMEox5yxhlzAavEpGt3vuficjP1Pp10nyDKlQYG1hZMeYDEgchY/6MiGzw3l9n1t8lIp/33t8a\n2cwcmLpfzc6ovzM+qSHsuYieD+xuA4Ywy7JwkTQTp36VTPxXkd9sgpqMgJAo8D28Df47KsNmOtk8\n2NcGsK+Yr+wSXI/BrHtZooRPCBvXQo61llNSodMqfXm1jtkck2yVThq5Iea4zxcizSu9WaL2B+PJ\nkbJDKGT5MYfw8xnNVfPBAN+JZZtYnjXuZj1HrWzjsRGura4gbVhi2C+PpyUXjhPP48Vgw6mE/xFp\nJ1V3sgis/pjH1+wa9+rDuFc0P5zUQQVy4mSMde1L1crvdP+1nF6CMW8dwwEmIiGr0i9E5Hne+1+q\n9R8SkRd776OMuV6f4Ov1fcLPWK32NanVatJK4plrFw/XKoiNmbXhRJqGEKoYrFuWNpTQX81kk6PL\nEm1aOsosqyBmgI5AKxN9RpgorltwPOZQuC7nw8IgF8aORHNyDCTNKkRqzUcU1yACM5eUeB5CYAYy\n9zVUoA9Fc/suhG2UwbPF/UqBlWuOx/KkV6sfOTmYmpGxfCSpwqy8R5jABqGLKXPujEpcrgJaL2iz\nvU4nOhE+jDugqqLazhrn6mob3DJZiN+o9mOJxwk5kyGfaQbEMuPhbrX/I7gjRvY9H639oOs8NO+Q\nKOZgv2sKXFNG9JKgLPAqZcNlIiIyOLjvlV2HZA+UmJgbY1+V8ReSsGU9Kb9Nkm/nnKyNarVxUquN\nk+ZDVOvBUMpgpOvfdTIpjzQG+j2ACDIm5V7h6r6+X3FwUh5NCJPy6ECtVpNabZ5wilq69PF61zvt\nqH5if9ALxvxvInK3934FlueJyMdF5BVtEkXjwDkBB8G1ClarnaiHF61kQdggBYBMd6PumhU0QAaN\nMfmfNH9yGcwmqNJpyFiRPUSqYl6P5TPVb8y5y4eI2cBYRZnVN9wrwiayFwVB/w3LiyyrzqtyzFy+\nHzFtkcovGQEOufsnQN22KS1XKv8Fnw2w02HV14bRE0VE9tS9K1KV3UgyUfe8U9FCtZZSS5jfNRi2\nT3qDPCeyOO9ZL+KqZsYdzh3qlB2QQnPdF3kfuC+dQyZLTYZ7thNTQPSdnd7aN2ZIb2Zr7P4LO7UE\nY97RX8bcbQWTwyWZSu5Qqz8lyey02jn3gHMulnex/2jRDVaoUGHUoQf1V1vwRIm/PqP/ASYBdKfS\nRdrIehGssQf13Ao5x1sMoN2R/mkIgROT6LjPRLFlwnupvGskzeb7mj/l5b4VkRb3Kflx0qwDSw8V\nXozLmM71kWKH2B91vswNPaS2ISOjES6XaVIaaKDNUj/oSC2TmS8rT3WLuxykAA8JoAjBYSg8ZTMG\nAIUc22of52Hd9ZQ2aHjrRJ0SQoUL9M1wBxPJsVmAXW9b01x1TEb1GrrAnTiIFVoC/Fyb40RAIyvz\nIgcJLFZ5PkNCtYmV7lP3mQFb3C/HT9vrJrQnX6t2+J7uZ+nJJRjzrrGvY65QoUKF0Y9RwISLon8T\nMyscvAnLRyDl4mqV2Wsu/wFD0F5GIk2PhbPVhzCV6CVh4qtcYtKeH5MQJlHXZXV1sQCWLHAf0Oe2\nsGTq0ODmRN05gwk2vl/1hV6bnhWB6TfQQrLQzNay0ZeBKX/LOPs3mpuk3NXGnyTZQBDHZTjOh7P6\nqcephr6M80ll9kIgkI6Gfi+kgcBVbEjwm1VnMMGsdJJ/hfYiRXyutzrlZ0kmuB0dgz7M60O3yDym\nTOkPet3V0OOG51k5KgUGi/1ugijD0hIpvbseFNyGglRlnjONYbQT7A+RZ5zSVKg+QlaM4B3NmNfD\njfN5uF5ku4xm4OXSz1t41A4x4wdO5j/aoSsvLLwgxlB2uZEIyb5XRP5/SZ7+x0Tk/IzSUhUqjE5c\n1FcptsJIYQxNzCMRkn2jiHzce3+Pc+4vROQD3vtXpja6FZSWodG12Dgy9FdMsk7ic7KO0KB1OMMC\nHvxctXsQmEeo68axmAoUIvnpPTMxgBbjHP5U0o5nYnal97ZJ9YMukywJTHG9mjxSyYrsdcvzOiDL\n4lgw1j2rml0ydfp5dfUsssagkyXxN+uhE6uzCClgL3TxqYASuEM+pCQw3nLrIRCkLOXpcgX03FTx\nrsY2zKPBlzzKaJnAiozPeq0cpf4nu83wsGhJUWD8lFNVdEhPn9LcJuij23jo3KiO8+fcna3kE9Ex\np8ZAYAybIS3k2VmoY+bp8XluDSjq/ms5voSOefjA0DGHkGzn3BPSfAUmSVP+qVBhbICTcoUDCwch\nYw4h2c65GZJQHoe/k5BDoxXX4+sVq6KbArwMtkN3Os0k128B9JBD8HlNRb9FWHCK4YDF7EGZ7Ila\nF9lIGlq56Q1AQhjYtmYtdpxgNnOgzNNVkagnplV7AOttOs5NdzW3Oc4kLbIolBLTlq7SztVgcW1T\ndg6o/8G2dn+6dWg2tDkWwbjTkJVcv/UM0MtEOcfI/0F7dYlnLuq7K5Jvf2jn3610p+fB0/R6MyYm\n1X+JWsdnLEhIGMMenGQRbyXmAeF9oOSlcjEFW449Dq/FkPpw8f261ejkQ3Runq81dfENtBnPb4Lu\nGawrwZh7UWOwC/Q8JNs5d62IfN17/2Xn3F+LyHne+7l2u3r9Kb5ef1yS+q0itdq9UqtNkdY4fwZp\nGGNHSuWgYQ0gcDNbg4cppoLIq5kmIq3ubBCzg8ohq6agjiyMuOhpaFcifiPGMUmDVRMsSppbbm6u\n4hD4YrKiyDlYHy3cStgPFSciPX4r/g6Y9RCtp6rz2GFfzDwXRHyoNuBDlcoHgmtw37uaq+iBT8NR\nSgUENctmlcM58/5Oj6yz4xxAy3NmnpBYbhRMZMOYMMez2C63ndnsuh7nzEnQ5j3RNTDpVWBdDgNy\n3PLycohkIusDM0n9D7XZudD5sFCy9UGOjTlkbzQh5g+I1H8gUt8kIscuERGRwcHBrkOy95WYmMf1\neWIeiZDshd77xSIi3vsvgk2nUKsdIrXaISKyFGvOjHWrcMBjZvsu+x0dlKeq0FPUXiBS+2BzHl26\ndGm9232W0WT0uxLlSIRkb5bEE+Ne59ypInKl9352ZFMcmBQxltcYYu4cqKnppsWqEtEcxRTFaQTE\nfufjA7gqYnALoGhFpm5z/krT+EPb4pYMo1MeS6b4yDw+uU7/NmR9AG0jPaZpDECgWt8yXRUAshPG\nPX42LzEMLWrUgrpjNyo6W4lllyIZqRzXMQOeNPNjizTVQjW+FmSARlISSUr+iohMpV+ZnUwhOTVU\nVgCyNtal46PXQHtys2u2usYaMbUqhgFK15k+ecD2HvfMmfzId6pzDrmOcS2XY6phwMZy6zYn0pQG\nMvKGh2rysfONqf0sWA/wzfGf+WxqAYyXhZqLoPYAE9/9SNK2Pl9dM9g/lGDMh41lxqxCsnWKqXNF\n5JPOuXGS3IKM9FMVKoxSFCp7VWGsYQzZ/voYkr0PX69xYHlrEW5t6/yJSJPBktEWqI5A4xwVLGVe\nNroOvT1iVEslwIEofgF0hYG16GoPOoS1DegKGPTCZCKkRXCBu0190JlfmB/5n2K91etpEuCY2pG6\neFCaYbiX6QCWTP0k9Pfh+rxQ/UY9Z+s9+olL9vVs5jR6k/qRkcSWGVKvrtNNhqAaawg29ogWVQn0\nwcGmkCc9UYdMxg1+sR3SwrS8wA8q/WFR88jtHCNsIZwav4X7T4NzQ3U20pjN77wNx3um2iR1TLo4\ncqyQMFueDWyTa8vh8ExwUyrAC/r2K5QV9pKs/fH54YPQInF0zWCHSjDmSWOZMVeocGDiue27VBhz\nGEuMuX8TM5nsRjDlSZk9JejHVr8taeeC6VBvfLrq+nZ8FKm33WV1XLGwVaPHfTLX66T3QGDKZj9/\naT+wBSo3hGCVac11KQ8KMsHDWlcfLWkwVWhgykZ/HyUMDPAACxsf6RKYMiUW6tzN2OY+2tyGARnc\n/9bEve/ZKQlNXeMZ9prZQArFmAdYsM9cl8CUScmV1wRrH9LhZTOkg5T98UdN3TdJ22wwy6eBMQe9\nvjqfIehGJ92U0ce6JEo6SZW9/9p1kNlnGa7Pmn8c5DFF6vcZL5/Yedh1KRasxs/rw/eMiY+eafaR\nrvqpYKUcQnuDXCXdYiRTZTjnXig9jHjuNu3ne51zP3TO/cA59znn3Hj12/ucc084556at48KFUYd\nNvVViq0wQthX4q8DXCUiS7z3LxKRJSLyD92MtWMds3Pu6SLyTRGZ4b0fds7dJiJf8d5/1jn3DElC\ns58rIrO8979O7WAVPr/BIlukZt4A2gbaPI8OgsyW4ikDAj6n+kCPm2IPOUn8UyAbIltRAQl3uNYh\nDKAlWdyockPXH0zaEKJuQmmZ/Ol5zU3ienkF6tsfUOtmt7vuA+r/BlrqJ6kDRLh7NIDF6J8zvT1O\nVf/j3uwF8w5MnXp2fc+AlM6foPulDloAI9uO/ZOBRq8fRSxQwj2PtOwipBJokUJo+2AYOCUKeDes\nQCj+whx22hFyajradJ9k4CSp5HQ6QwGdkbYYqSd4yUTuQyasp5NE7CjSOrZ4QFHXX8vtJXTM00rq\nmJ1z/yEiN3nvb3fOvUlEXuu9f0vZMYb9dTkxf1uSimy/FZEvici13vuvOedul6R0xV2SNTHTXY6u\nY5wo1qjrMcfWMHtKa7seL8BsnRkNL/cF2A/nB+bpXa32zwfDugyFl8XmChCRQfy2pIjYmIFSLyM/\nDjx3qhFi+RaAYLShP57KNcKPxOnm2KkJTgddwIDHQJgXYTUnzkiAzg4Y+aamnq/YJJKREyUYgzAx\nNy5u/pQZZFGkIo7Z9iGs17Jdyt0PCGI9XeN0RBvdEfmRa6Cl8VcbhAmj3uBYjsXxV6nnlSTGGgzt\n2P5RrftTtL827XsjASxRF0lJGwFb6gSehTEh0pZVdKi6KGJ0z3XZCxV1up6Yf1ZiYn5m+Ym5XDSY\nOQAAIABJREFUWMRzQXSsY/be/8I593ER2SaJV+k9mJRPE5Gfee83OTeKRcLcaLgDFHeMxvsxq32X\n/Y2sSflARtak3FfYcmTdoVvjn3NutbRad5wkBPNDkrgNL1YRzzeJSvRa+lhdMOZJktC3M0TkERG5\nXRLW/C4Rmeu9/61z7n9E5CWx2n/1et0nVXATEadWuxlVsjUo65MdWlHNiJsi0hRXLSsdSJozf9rs\neltGzT/LKh9SExqZTJAFyeq0yCxSyHDRNhRcIpU/YrkHYABbAxGdYeexitGFEXP3A2u/CW5ZobIF\ndDQ7H2xukpXfguIqDZXadjfZ3Ku9uFcTzHFEJF2n0SJHzUX3P1qDClXEycj5sFE9GwgH/wMIyWHe\nSFw0kOmJ395fSnRktn/e7BoM5jZk3e53hRrTwnPxDw2n7zDLhL4ReJ8oIf0RVh/3L9yptEdObuis\nPpRGnydS35z8ycyknvPg4JquQ7IfymHM35HWJ2W5SKnjOeeGvPeT1PIj3vuOK9x245XxKhH5CdUU\nzrkvSZKdYUBEvu8SuvwMEdngnHup936n3jipglsTkbVc08VQOsBtVc2/ChVGK2ozRWpLPxaWly79\nz3q3+8xjzC+R1nxRnyr/EdjunDtFRTw/2HaLHHTDmF8qSUDvbBF5VERuFpH13vt/Un3+R0Re7L3/\nTWQXcR1zLCCA2GPYFgswf+KlqlOeIVCaFvfjclhLCGhg2rcTmn2DYacR33+stl2oP4egjm0QIGLi\nY4xVtWAgfXzqTmmYSmWRg/Hpuk81V51v3LQC44+EoQcY/bA1+m1XTG3a3+Ef3KQiNf+okbN9otuS\n4dHV0AQbXYxtlkWuI5Pn8LYymEar3ux7kZmHO8LiU8Yty95znnEi+hzwWD8yy9RTx5hsRih8cHm7\nVNrCGm7PVNfptrcm7d2fTVoK7zbNgNZL81FbgHu3E+HoKSlLV2r/SNd6uP8qoWN+Xnkd80mSxK4z\n4vl87/33yo2wiW50zN91zn1RRL4nyVT5PUnLR156ka6vQoX9if4VKK4wghjJABPv/X3SSrq7Qv+r\nZJMxBY2GHg+ZBRjH2puT1qYJXK82YZ5ZWrPnm75TLSsWaTJjMBAPKcSGpoqIHDEH/7AySlYVj5yQ\n7BDyHbn21MF+DctnsU8OkyUjp3NOKikSmNSazzZXBeY3gLaB1jJoBesxkEKEPZ6LbW6wHiJ5zBwI\n0hTUTl5Jh+9Gu9yMJRUMEQGfBXphRCuD08OC9gveb1MZZ1A9G+9FyyjxcQXuHUEWz5xMwTai3Qk5\nBtovrO2CKk4VsBQYeVZColi1b9hu7oR0GELkI3pjpgbgrMfnNS8H+O34je6KF5+Df24yHQfU/zaP\naHlsLMGYj69CsitUGG14bfsuFcYcxlCR7K6zyy0Wkbdj8Qbv/Sedc0dKYrp+liQ07I3e+0dSGweP\nBOiHa9C/tdb5at3mZHxR9xn2+x21za/x/1+Y4wWmHKs9R4s7WJ4juwB7jCZxMb63ZJPU8+o6fNY7\nQlvaLahnO4srTABIzOeT5zrOGoGNbnOOcvK3EkU4H+gidfTbcajincWUyfJ/r9YNgL0xcXqLvlAk\nVYtRRFKsLlxDeL5owrNcp9tUcJbdad0qbg7zQVE6SyXZj1XApmRkxr1EXfNNeMxDGDd/0MmdpNVt\n0fqTpxA7T47PMmac33y1/1V8tidLHLEpAIz7BDDmAeyv8bp01zMzxk9PDr5vm1S4PmMKgg81mTJ8\n5zeBSkfT+naOsZQroxvj30xJ3MlnS/L0/4ckWQjeISK7vPdXOecuFpEjvfcfTO1gB94gxv03coxx\nKRE6L0qwXeY57TdLsZqTBo0qiGzaox7w4FKVkW95c0Tyoai2ALLtuXAz+99YTzc5bQCjpMkAgJSx\ncqDZdy3cyWzaiJT7l5oYlv8gaS9gHxtVp4ujWvdEXPedyKQ3BS/7Tcobkja/7TTIWmMsx6ICc4LY\nm1UyTGeOhRnDGqTalr3KQfhwvlqtzIr2tIZhjayCs3zm/lSti5VaUpignom9OKdgEMSHMjd/RFbQ\nDlVsvGfqY23vzbU4HjMATvm86ttJYQuquvABuxsT/zyc32Ic70Nqk6ZBsGvVwn0lVBkn9VmV0U2u\njD8Tke947x/13u8TkW9Ikk7oddK82ytE5PWl9tq9Kqkg7AObg0J+rgY/ad+lK3BSLgNOyoUQCe9t\nh79r3yWFLWXut7UtjxSKhOB3gzaTssbeTohTJ0E7L2zfJaCbSbkEOqn1mIMRzpXRU3SjyvihiFwG\n1cWjklTfvF9EjvbePywi4r3f4ZybEt2az37jpNb1Ib+uwrPsilS8igKYcmbwxgDaN0jTGEP9A1kj\nDC5hQlYGmL34yk/AMj8kMxGi3cDkp4uOknnfkPGgjXty8/9XPtq6TWDKDDVuJM3hakKjSxXHMtFm\n6xpImgsUSx2PvsMfN4PBhHyr2v9Z5jpMoXoAEsZb0HdRkRfp+a2LM7ShihVXwEr3wdg7zkhOIioD\nICYJMmWG4lujoEizGgiNWTRcRcVxvhoMhSfTJFMeQNtobrIT9+5pNEDClSyXBFiWjf2GCVkZkYtG\nJDZUUpQBBmnxmbBqmmTCvMI1DauXeEidi43UNgVVb2ITM6WO1VimuiLUc1SGWz6Gi4wk8PdY/yV2\ndK15RbrEQaFj9t5vcc4tk+RW/E4Sd7nYx2aU+h7lWMgrHOSobOKjAj2clEVGBxMuip65yznnPiZJ\niMBiEal57x92zk2VpGL2n9n+9b9yvr5dROYnVoJa7a4kEnB1RLSlo35IIMQx48t+tdIb0GXJuowx\nF+/5pN+6Hp51vofIfOt5SXtWiWsUrfoAI+IqVOomeTmpk2sf0aGndPAmI10ssY+txhwkDJs4SiSw\nunVghDZ5DrFV3btgnKQBMiMwZ73ahlkIjkBbKJQ8q8YiPSu0sa5IDT4gZGWjBEfmSnsEmei3JA38\ntg9VO/JqOlrDMA2pDLZJVQwXya6zx6ARfc7WyMqAEjJ/XDddqb3tcxmpGh8kCF5jcz9uVPunm6jN\nfKdQr9elXm/q/AcHH+s6JHt1CR3z3LHsLuece5r3/pfOuWNE5K8kMeX9iYi8TUSWSeJAG1VW1qYl\nf7KUb2Gtm6FUqNA77OvrO1lBmLKhWbVh6dLherf7PChUGcC/IxE+M/bvhnrjC865RZJUn3tjdEvG\nGwRWB+OAzjNs02Iumda6fBGY8tU670VGchv6sN8Oo9kv1G+L0VLXdTiWc6uqGFc06352tXq5L8K6\nF2Ld0yUbdfSh6j0EP5ClkCkrtpRyY+M1pQ2WrEaVKiZTpovaTMu2NcCCTjyhdfWwax3j+9Vvd9C7\ng14ZT5EW8Fprt8IN9LQASw1SjnFrExH5NNrrjUvdWuj4qTrXbpNPx/YL6OI21DoW6nBzq5YTZMqn\nqHUMyoH3xR/AmFNmk2bA0QNHJu2Lmf+Gni43w46i7Xghudb1GWO61LQiaR32DtMC+rbzutPD7Rq0\nDDCO6robZpnPGt7ZWDBVuM4ZHiR79zT/nyBdYyypMrqamL33r4is+7UkCY4qVKhQYdTgsfZdRg36\nH5Kdh5UmaONsEyIdC/ywOjsyUAYTRBPiDKBtoDV18Fp8cY3+Ni/0NMC4Ig2B1QWnCbVtYN5W12vS\nTrYUFLDHtiwJ2974hWaXwGBiIbl6HyJyNagTmX9WSk3t6shzCteHhRlpdI0ltM8JB28Les5w20jN\nxdyE7Fn7o091RnKsleqcWYNvnt2/8RW/U20Twp0pUZzW2jcXA2hxP+7D/W3REVMXTvdK7lczfWm1\nD1CapORIPfdl6PNhtX9KG8dgeU+Ba5t6ZzIqsOh6h1O61/neUULHfPpY1jFXqHBg4tT2XSqMOYwl\nHXNXjNmEZN/ovb8W698tSQzr45LUAUxH/q3A14suMfzy6vI+TAye5WieYpcishmaJOpM6cnBYgjh\no6kVyKSu7aIGpTjragktNwnT7T5a0ihm7PduWrCxHK3AwsAFsutIlB1xC/ZHKSREU9LPWAVB2MIB\nHD8TvDJSVyegykrzmSpgoJWbP5coYtc8eJGY5EjUU5Pl3a2uLW1Jc5ApiN48F2VFHEbA41KFGtO3\n8hy525R3idLVD33X9CmQ8IiwCZuiqUIp+VBC+TFaEzU7pK6THS/PhwJki6cInzE+c5RU8MItxLYr\nItcploJXo1UC65rB3laCMZ85VhkzQrL/VpJUd4+LyH8451ZKItQsEJHjvPePO+eOiu7ABo3EHvBv\nMyAgYxA0tLXkHoAhkDd1idlvS6Y4/DYH6+hqN4PuZzQ2qgmDWd/+1eyf95FeYS15PviyQVzlBMMx\n/ovqeh7WMW0EPzDz6DoE0f9SdR6X81j2g2ImZO1tECbkAbQNs60q60PtALUR0zGxfQLHY7awmCoj\niNIwyqaypj1XgksXK4tMMPeMxqez8ly64CrGSYMGLC2x81pehN+YKDA2IQ+byYih39zHnpz3dhoD\nS2B8TQU7KbUI3UPP4ApOyJjohpT6aRItmnDd4zzzeuzjy0Z9JJKjYmM2ROQAuUP9xKhzPtO8Z993\nrcfTxwyqGE7UeH5W6DB3AzshUz2nU77MOlp6hYPF+BdCskVEnHPfkGQGeomIXOm9f1xExHufV1um\nf4gmJqqw/xFLZtRnDPeVLFUgejgpixwkqgxUhf2yJLnSHpWES94vIi+XpDr2PBHZKyLv997fn9rB\nNnzuyUSy2JKGDaSgMeIS1aedq5M21oSwaYh7DKmdYlhwngREEXATlgNT1q5lTF5Dw0tW3bXIfrOC\nLOqRySOoEqxhDazIf7rZN/OcTL5hDQYhkK1SsmCZ+xb1jVExZFYiz7m2NgBBT5g2hzKfn/+F5SA6\nx3I0kI9QnM+pD9gOsaAaPpekaEFqy1FTBNVSLMAnA22r3YjIMvS52PRJPV+Rmn+ETRClz9m6fk6A\n/mYXXN0mU8QsoreHfmjXw81VzYRNXX8tby6hyjhnrKoyckKyD5Uko9yJzrnZIvIFaab/Hj1YUDHm\nChVGNSa/v32fEhhLqoyRCMl+nYgs897fi/U/FpETbKXs+iTn649J0K3VDpX/1963R8tRVXn/tgjI\nW4yGjDwS0QDCABmeGTQfLQoigygMDIIfBiLIJ4MyAkOIOJNkRBJdCIslZj4UggwSHpkJCKzw/MZG\nHkbCIxAgYAxzCRN5DURAYCYI+/vj7F/1qV2n+nbfvje3+6Z+a/U6XVW7Tp2qc+rUPvuJ2vYApsXt\n8co4255qShsyhrG5mOeqyzizU6MPog94Q5npUaks2qY0PN1CF15Q8vzmRvW3FNzHoyyEJJVlsYTI\n01BxNzl/PJYxr+eeD02kGEe3mfkfOWcymN9K0FIGy0BQnvtNcXt+n+/LZLZywmWSaWZSydXGh2zb\nunfWXzdIptG7meORctuf236KUo+K63dtKOVox0X/+6x0/U2l2UUR6fstUBZ1B22Z/zl4hW4zZApV\n8xWbH8m92ezP/9DtOAU5xMpFjrFsdWvv9Vx7r6eouWTXs1NmzpzZsUv2JW1wzCf3KscMlLpkK4Ja\n5S4R2QHA+n5SBoDaBuGH423HrztpSYUKg4dsUq4wbAgu2bVse8aMGfVO61wnZMxApvCjS/a3VLUu\nIusjREiZgCB7PoPccw6LGCjfrs+vcux1XQh6TldT9+ZMiT5ucwdwP9SMc61TcBCILBR8Jg6a7NH1\nV5y7bw6eCzb554VRSEQ6sGccvmnn+4xL8eZnQFEWzvt53Y5/1crVCXM22vSf5uuNQju+Ydcm9+v1\nAZRBLmyc0lglsB7eu1kqwIcmBUplvUtd/wBRthH2h/WRdxNPyU59+M8ULjYaOllk9+OSNCyKxt7E\nsr73DhSfiI7d62j4XPicYvdq6gxMbl8weUzpB/xztu0VtuK7xXZvHJ0ypaSfyelPfaZBm2UhN7dq\ntZCj0qxNhiyno+sHOpaM3i3a+UjHHOzFbXDMp/Yyx1zikv02gOM6qbdChWHFxcP6TlYYIvSSjHn4\nPP8Yy3tb5xCQQoELcshxyWZP2Wda4RSH6ZF01gCK4TNjGOdxiAWdWWXSmq03TdASTu5GbuKYiGNe\n4+onpzPO5MZM9xNzuHJSvlq6BlN+SPvj2P70cBTriaGRHJGc8iV2/sk8x4YPVzZx4nFiodVTyC3o\nnXqATJfgHQ9o1BJb7HgZNsExQplwLH4f4znlEjn+qYpClmzKc5n6i7LfWP0wMeZuY/gAi59tgSbV\nNq4U7ZmRUy5wzhGuMM54Mi2A7Lk/a+cwxVhs359lNvc+AMYpx05gi93KYiceoE4hNply2LvEJiBz\nKPN5IjvDOiPK6AjLbVlRluATSPjm03Pu0FBSYeWTtgIN0y1/LO9/b39sMl1jkzlf7rKXH0AhhXwW\nkyPKRlJQyvllHb+LkWmUfykyUyXntJADJxGKALzpeKTVoqlWZi7FJJ12Dh1cZkanlObRcx+P3LLe\nzjnOvbiMBvdyROuVYzTxoj7p61YeGdEssTKbjFx8ZIrGXo/OqbnreIXk81Gbyu6Zjj3nubEDoOGo\n4sQr2YcmtjJolq8PaMoUlGbnaQbG66BiOGGOZyKe1y0Z3GaFucF8xRZEKiMqUCfRkYRjOpXQ1mN/\nK03S2SQ+MwYh598P2hBlnNXtOf9E5DIReUFEHo32bSkit4vIUyJym0gQrIrIZ0TkARF5REQWi8in\nhrLxHWGQ84kV0UZQ9rWGLsza0mq6pLWJgSR0bQv9TcrDgF8M6zy0VjDScv5dDuBHAP4l2nc2gDuj\nTNjTbN9LAA61XH+7IGgrSvLMO6QUAet5IvvK0zyIadVTnE7m8OG40hkkFGAOr2WcwAbONXeT7yUa\nOs7KvlCQMyvjxoCIAzBO+Qo7Rg4wjteb+UPYsrilSYK61dNLjtt9NTU3M677EsdBA8W+YfwJLuO3\nt+2UbfiVfjnsTMjiDCasf6oF5X7JsnEzdnTSDZ1wjiR7jrU/fcU2EVm/mNiAbMqLkvhwvzd/3aTz\nkVfymcNEIRbEBdF/ctdetENOORJ/vWwOQuSUC5nBU1w2n4t11pp5oaQogxxvpAjd7Co7RhfpzBzV\nxsQRaAFOSfpW1HeMv0kd4q5Wsj+SmXA6RzdMuK2iX45ZVe8BsNrt/gISmbBV9RFVfd7+Pw7gfWal\n0X2Y04WcWoXuwJCvpiq0hEGclIEgY271N9wYqPJvdH+ZsEXkSAAPmZVGEZ5jY9q+OM4wjfuPsH0L\nKMcyzoRf2P9OLMOyrDROlpYItIbF9une27a9mdDjifrJEFC8+3WjYay9pKuxaceyJJNjChS3mRjy\ns2eaoqewkiC3FSvNyOmXRXsyjrqpdMW4rBUvF6v3n2+OXCoOZ5FjTlTro715h49Lo/98/ncbp3y+\no425ZDrwUH/kVyzLrU/HxzsnW2k8hY8LPJr9MQZ43tyC77ZddCShKSIVxkujsbGr48D5wLlCYr9P\nPaNxDldN6xtNZgaY0D+MMscVZjD3uSgvWhXK2HmK7wg5fHpI0zv/psQ45TPzZmwpZ6266weucjMa\nM4GMFbfMRkJOORPt2DjNJuQ4/lnnIXd6iWMeLKuMXA+aGGMWgAPLTvCePbV3gdouZdQVKqxFPP9C\n/zQVhhRhfngz2545U2qdev710sTcklWGiIwFcJOq7mbby1CSCVtEtkEQpk5W1UVNqm19neLleQzb\n6DkqoGhJca1tH93kct619TTbvuhYI4iySXgO0AeDaeYKnNlaMaYnM0/c3yD5Cys3ogmcD3SUMqPy\njhk0SWNs5ZQlhz+H9e7u9keguRk/t5Qfk4OLVyPfd2FKG5VYafE0Y9kjuSrfHz4LDZCQrzqkdBZ+\nHDUNnEVLFj5nszLw/Z+yRPEZS7LY1DQvmYNy0CSOpndnRceo5rFn5wMUJR02nJt4BicDjpE97xbi\nVPOaNNSgcxaDhXEYndvsdS/LdJ5zxOlYO/ntNqwyzut2qwyDIG+uciMaztSTYdoOEXk/gJsBTO1n\nUq5QoYvx1/2TVOg5vN3Gr12IyJEi8piIvCMie7hj00RkuYgsE5EmAaqjc/rjmEVkHoAagFEIYeCn\nI4T7nA9gW1gmbFX9g4icg2CdsRxhIlcAB5XEZM5fOGV3XHAsMRnhg8Y50H375qge5rLblfbAtp/c\njA+MAyCzx62bMT45sx2M9rcxrQvZ6eWHrYQvXWk0DHmaCmfpnSwutG3OGS81Tik452ScUz705qbS\nuM4f2e9l8spYyLzCFACZ7M9z25QFJro542xdFpckl1TiPtzMGcXDy0FzSRRKrBlS9urMfMNFEwPv\nZBx6k3vO2uASLaTsdMlNX23b01ww/GYorM6cXTCAhu08OWcOIJ/tJpHRh2P7GNtNY5M4cBOTF8xz\n46cQVjTSp7xjoiKyha0xsh1zsKe3wTFf0CbHLCI7ImhfLgFwpmrwTReRjwOYh6BB2QZByj9e+5l4\n+5Uxq+qxJYcKmbBV9XsAUjZmFSr0DmZ2PAdU6EIMpYxZVZ8CABHxg+cLAK6xxCF9IrIcwD7oJ/j3\n8CdjpcUD0wdNipq0geMouf0+45jJvZyIIiZYOdFxReM91wdkHMLTRvOu55RT2mHzMCOnnAVese37\nov7ZzwWN92ul2NV8leOUaTftQ2tuF29Y+y4xQR9F2Y/b9Yxb+qN+AwVknLJxo/OedvuBArNyhcnE\n9w37r/142Dz6RxHdXlZSf3O1RThi+q67jWuaFNU711YstIsmE8e+fPLpBu1OcDB5pHHKS+z9mJBj\nTKzPM67Xzsk4ZZPNT98DWV95TLW2MXjPxOieL7PSe+SdYjRzEtwjQwZknsvOrTtOp8VxTrvigp3/\n+/wONMauF8/Q5Z4ccxR4aQO7Jhc3HK9HJZg8zykTjJj7opWvRQpVNjNbOVDePQ75Cw8uhskMbmvk\nY2euQj5UWxLDNzFfap1/gm0vSE2ChmziMmP5XdyIzGXOMNpv5kky/Vdmr5/4YDGq1jv+o/eV6D8r\n5uxnL1lm+2qj7o34fPeSZyIBxkyIuqFgZudvJAUqjE4OxeOsg7Z8HOjN6rJJ8cjEIbaXYgF+jCz2\nxtE3Gl3OwcQ5OzC/Hs3c4hjVNLejeST7ky73d9p2PO9QqTvdP8uACZoS5X3MbdvHtVkmEE6qfMSc\nkPkhXpQ4J/tI2zjlhJy6jjfZ+/6sUE61C8YTH93lt7KSH4AsfrhTvAEJpx+nPE4lzV1pJefLl+3Y\nJkb7RgsSgZ9ZSaYqVoZnoh7uYN+ZKWhsnkqaskw+baAZx7wKjeEGACJFKxARuQNAnO+K4tpzVPWm\njhsYod+JWUQuA3AogBciq4wtEYRUYxFcq/5GVV8VkfciWKbugfA9v1JVZw9mgyuMMMzyH8EKFQz9\n8pXtodnEPAZ5r4LFCdM8VS01/22CVQi6OGIb29cUg+2SfRSADVR1NxHZCMATIjJPVVcWaiW4nPGc\nA5BQ/llOPhrdU/nxSqJemgxlpm8lIgegaJ51n+2fxHBpc4v109GAGS0yizTjRGKnl4KCLaWkMWT3\nzLWtj9ruXGwBZM+D3G52Pb/WTQSu8crKzDEnMlW60B6IF6dwRHAJnzJ94wqoz8Qs5JRpZhZnrCkz\ng9zR9Q8QueHzwTNmNp1svJkeAFiAKW82N6pJ0k9yu8y6/VnXlpi7Iwc7mt4WVFBaoCN22X5R/Tu5\n55Hl5vPu3QC+a6X3TNwvv4nPb4ECfD+Tax3n3r+4TTe5j+ayYrUZyLVTdDXZtTEOHlYwe/hgvmwr\nOFPrWIuiDG/BdpWIXIjwqfkYgPuTZ0UYVJdsBLZ+ExFZDyGe2P8AeK2/a1SoUKHCUGOIzeW+KCLP\nImRxullEbgEAVX0CQaD/BEIqiVP6s8gABu5g8oqqfiA6vlpVtzRRxpUINjobIWQ1uTRZqTeXS5mZ\nFfYxxCIN3hMmVz5esQ+RSE4rjh1Mc6BUPNtSmABujbkPk6unzO7ZiHQSBeme86YjSMz9OmP+Vszv\nyrJENM20Tbk3uXdelw4mCd/1jKNlfgRy/OOs7GvQ0izxX22b4m0q2lJt86sZ33fNsmQTTfPgOceP\nwnP7WkTrHHu8GV4q9Cb1Jie69jIrzQ1OVwIAl1gs7pNLOMxYbeQzrvAZU3QuTZ4TaS2GUVBwAtk4\niJMDMFY6vf/ZR3y3nonawHE+32ho0ULZP1cCE5s8J7iVKdsfWwxe3LnDx3FtmMtd2csZTCJwlbAv\nwnp5DILd890icqeq9vkTCi7ZfwnUCvlQKlQYDnhvywprG/W7gHq04J+ZUMa1i3XWJVtELgbwa1W9\nyuguA3CLqv5rotr0hXNmZp7E5G6LzQxgb8cNAA1zOM85e8eSlGzQNP2zvhQ2pylltLFs1nPthHeG\niL95/+hKwlt4IHIa2Md2mPXIEttfCHLj2xfDyyljUyyTXdvzXmMi5Q04HloJGu+tAWJ4bb/vD/Zz\n3MfeFM3jtKhN1Bl46w/KRWmNkAormsG7P8dy6c/mz9+7pN2bR/W/VrayI1L6AQdff8y4lTF83nEl\nzkhNbjdbSaTGtIN/V5i5hvLjWNdAF3WGD+VrkFlhNHHEyUz4yuKE58Z4xxzsl9rgmK8ZAS7Zx6Px\n5q+EGUaJyCYI8pYnO21khQprF6nUTxV6HSMqUH7ski0iKxFcsmcDmC8iU2Au2Ub+YwCXi8hjtn2Z\nqj6GFB50X2HaH8cSaa9t5leeK02GibwjIqFMKw4+DxTDjMZGi7TXtEBH0462x/KaddHm8YmeCyrh\njlZG3btdwjYbQMNLIpI5bu44+uutZHotWgd8vXFKwVnVWx2kLFF4TePMNvAxYsaMQgGeMzvqbxzB\n7o2/mV2s7ctyC345d92cm3XGKXuXb5MFXxS3n/Jh45TJacaGSUCCSwaKuonEyqXs/LHu+GsnoAg/\nsVNYapxynC3kC/65uEGfY/JKAhKxP6iPyOkUzIEk87i2d6jZame8JSo42miudZxzDD63mAciAAAX\nkklEQVRKMuKF4F0pTtnG3tEmX/+O1burk3s34+oHgG6YcFvF8OX8O8VGXOYNlYg/m5xQIqSySPjJ\nwxv1L4kGF8UCjAHNYGwtBUqnxxQXBKYsSynrSiPOUekUpxqy5bRXLnllWStL3Ayc/PaP9pmihc+D\nJoKncObpa5AWzBbHOZqUwpDPhzEZqOjkSxdN4jyvoBD0kfTiuLCMisZJkGII37ZUTj4+D5qzuRx9\nQGM5T0UYTREL3mpRdDYfA8U7+BTEUUDxI2TInkXs3GSGUP698OaYK6OxsZ0bG9PtGL9FOaca75Bk\nSMaMMTjnoNfN43Izs+Oat6U17f9G59DZaLx/98mgJCLeDUKsjC+2Icq4YYQo/3oPE4bpg1TBIZW5\nYJixfFjfyWHCX/VP0uMYiBnccGH4JuYd/Y7EwBh9Un6bLsGMEib/UDxnPSfDGOXchybQ9GN/ZOZe\nB+xm+zhJOC4mmdGZXIWL7r9Rwrj/citPs5JcC5nTKTsXzyGn7LkjugjPiGh9/hi2lxmiM7FCQnbK\nD9S2vMe+UKTcegsZLJxiZ1V0DmkK0fy8GCEagu/33Khxyqnob3615JfmBe41utbptqjNUu8Zp0yF\n52ZoKDyZzSOL681zyNVF/mLZtXzuPRO7JJkBcspOecwVzCGJtDOj3T2SU2Yc5VhMxxBkXPXM9M+Y\nSIgNVpRwyldG/Twtf2yzh+2P9eWxZxltzhzwo8jDXztePeyLwUIviTIGmiW7NPaoHd9ORF4XkRZi\nFw4X7hruBlToVgx5luwKrWHwJmVghCn/kHbJXgrgcITYoyn8EMHLpRQL/y6Uh5CLLGhVgEJOvCP8\nC1PMmVesx7PmKUWcS+T9shlQUv81ap/omJNZZ/JjO36fBQOKdTinOU5qpstEnXNldk0bbSsAcqPJ\nJLLGbT1o12Z85vXdkvy12xv/2V7K1ynmvtXKzVDELVYyytkqc7PmoiTyHs70VKyXcbHrFmWsZttv\nRa8B7/0Nc9rh4ocBoeLF0PVWMmMzFx3zI64XQE40SY6WnLI36RvDldPuwMvGh9CyzSsVibujqGl7\nJfQLALDq30PJ55R0Xf8DcuDm8usa+8Zfm6ehrvJ8KzmPxdf3DjfW3xcaE9+wfItNBe0doZHrjVbH\nvUYcO2IdbMdutX1+VfABFLHcogSO9/J1DiAz85wXPadCvPD20Q1JVltFK/GY7zE75nhfWexRiMgX\nEFKrvuGPVajQE3j50f5pKvQcuoETbhUDcjCJ9v8SwBlRtP5NANyOkBXu7wG8rqoX+PoM+QunLCz4\n5f6FcWbeJTVlDeAN9L18MpXpl1wvRdYXueukwop6ztlvx8HW6Z7q88WlMlrQGsOYxtLszKdE9c9p\ncm9AQ3b6oWgfr+kzpaQyvCy0fYeUtLtgtQEUrWycbDkVbtJzd95KI+mS7eTc7bhxM2ceLS6SbtyG\ngrVNIhxAQRburD9S9+zHMC03uEhbE9Pa8mPW1FBSvpuUpxsKwcG8/DsFvk/WNmYyYVOTsvJULko0\n7vn1aN+u7vwJRrOE++nwkwvg1bFGdlIbVhl3jzCrjBkALlTVN42ZLr25gkv2/kCtNsitqVChQk+i\nXn8G9fqMbHvmzJkdu2T3klXGYHPMv0Ljk7wlwurhH1U1lRbYXTjxxU1yYjESX3/v+us52VSQG17n\nb237p/nr3RxJbA7l88rsNyncNC19GdcKFDnALM/hhhGR4zh8KMzzbTuOt7N5P5xTyh68ZvuoOeA5\nqUzLzDdIgaS/RwbAOTU6x1+zvwBFqet4Tj21cvH1UGZ+QGrMMJSpBcgvrFhiXYONKT4PdjO54ZSl\niOeIff10LIlXfoXg+U2cXTJZrFOKkctmQKEo0UtmiLNT2Tt0hZXxQn9KnoRBh/h4Dk7U5e/Nrw7j\nsKJMa3i4lYf4+nx4AwCDwDHv0wbHfP8wc8ytTszjECbmXd3+XyIkHizk4RGR6WhHlJFT5NnykIO2\nz3b7pKP59Oah8JNTK4lP+UJRt0JvxJnx9WhexEWGt7+lv3+cwse9ZH6iyTzC4kdkkc8W/e9QMoPi\nTTwnTuPBSdyW8+eayIfLRr5rPr4xAA7+myUoDA89w9E+GD0nPnefxDQ1ifsPHz+UVIamPPF8BDKf\n0SSlBCp8YN2HPY7/O/ow+2PKJS+u4aSxNGqT94i71WhSk1LZfbBeKk03jmhpPua965JOKA5MNLy0\nhXmG/cj0XOu5c3KmoOyke/M07Gd+F2KR2NGuPh+rJAbTcF1p25m4zMd0GRed1Mc/HU+Ue7YxMT/Y\n7bEyzCX7PgA7iMhKETmhLPbo4OD5/kkKuLd/Eo9SLjwFb/PZClpJCdUJEvatw40VAxjLi9o4ZyCa\n+WxSbgMpN+XBhA/x2Q0YVYh/MLiYOJB5rm9QmzCizOWaZMm+oZ/zZjatmFwRk0dMSQxW+pcscBxI\ntiyzCXlu1Om+HlqIjXXKsxhceo43GprlNfv6ZxwaXY35cRhnZV+DtFCPdwn+coOWHPPbxjGTU17P\n6ngn0Za3jFOmRTk58ivdy/DFxvYaY04y0QwnyJ2sfLJxnStMlDOZnBITbPrPekp5xlUpOWUvzvlc\nRMt0DJ9ydRQs5QHcE4rlHwv1jfcrP5p6nfJPKICM2ZlWcjztxA/wLsBrpn2lmIic8pRwveWXh83x\n10T1eu6R9TLJL9/4eJLyOQMnbI9SUHxCTpliLcYTz8Rn4xrncLWTiRpcnX1WbnxfY5832bzZSq4g\nT47af7Sjfdptc7WTyo2Y4Yj8Zt8z9kca4plJnX/MRpS5XIUK6xw4KVcYXjzbP0k76AZOuFUMXxCj\n003ew8BB5HRT8XO97C4l2ywDI1fxq0+TnE9E12FCZXJ3VBwWzHhQNJvKFHi2fZRtz0+0jW69y9m2\nhGKEXBY5jTttP/Ma7mVlrGL+sJUnu+dD7oVOPM9HbTrXaA61bco0U8pLb6rXLDJZhv2tvCsUNLli\nLrjMxTmqw68s/HUXRs/pYivpuLJ1yVhJ6S6Igjw3EViJ+LbRnuf659gm7fcrPP8MgMb4pHIxq49Z\nVqLVlA9JULhXH18ayATD900MJSVgVI5S/BQHE+Rqxo91Wy1gbtT+LKemix/O8cUx3iywkjetTCv9\nO5b57tSGjPnJHpAxp1yyfyAiy0RkiYj8m4hsHh2bJiLL7Xgqh3yFCl2OLgysVKFjDGXOv8HGQF2y\nbwdwtqq+KyKzEbJkTxORnRFiM38cwbjmThEZn0w+SIufP3f710+0gIYOTPrdit6LWvRz7dKfdB/A\nI6P/NM86ztGclaiX3AM5KBPvZvLbMhsUoNFuchHUmEdevVmIZnLVc0s+8pOitlIsebKV/8fKB+3c\njY12TnTOd0rqZS9Pi/aRWz+AXK47J8U9LvhVKCk+7LOSKxXKkeO6fufqfd1t3xz9X+g411usZAyp\niSSMuWTqBcy0sWD5QHvJnyCzRyRH+WFH+gKKIKfMuMW75g/ffHwoD50c7eQswGe3Ccet7f9WxDFT\nbv6E40Ypaz4zbC+IzDuP4KtHh6sjJF8HH8+50bPgSsIY76ft2W7PuqZH48g7Y81ybSNyXLLZ3a2w\nF9rrJmgddX10nQUtM7ul6CUZ84CyZKvqnarK+1yEhoXjYQCuUdU/WZ6/5Wj4MFWo0CP4Wv8kFXoO\nvWSV0ZGDiR27EcDVqnq1iPwIIeffPDt2KYCFqrqgUOlMk/fQXTnlkk2ZIoO00CqANp/kiv8+qtfb\ndjZ1OCgBrRkYrOWK6Mu92MqLS+R8lBmem7ieN8KnDO+MiIafStbPRc1cGy4p6xVvC+s4qIKTClB0\nCvHZjVN2zN55x+Pg6BwGtfE2w97+OCV79LLfFEfu4a09Uo5E5NbvZTQhcyJJyX4zcAlkNudep5Cz\nl7Z9fsx5Z4uLonP4LOka/3nb3+dWZkBDvn2q7WOAJTPgydofW+NwDDPYvQUPy1ZBqZUT+5HOKVxR\npmTMRCq8QA4rov8u7CdlylfbdrIfOpcxb92GjHlVL7tki8g5AN5W1av7JXao/wdQ7wOgMwBULtkV\nKlRowIdsGAyX7F4SZQyYYxaR4xEsjQ9Q1f+xfWcDUFX9vm3fCmC6qv6mWCu0Xq+jxtk4xdlGtrv1\neh01CiZrjhOMIxb+Dqi/A9TWOO7iMjtOmfbicDzXBuIkO4eu2Rbspr4CqFGoQ3t8yr3JjW5ptNsD\nNYY8/JGV3i32EKMdBdTG2j7PafO5GMNcX/+XxfaWcJT1rwlqHwWeODts7/yp6CDluSvt2T5lB82I\nArHFGIPFzzfaHxotbazLrl+vo/YPRnt3/llian41Ul8G1Ga7FQplzOTqzeKl/iJQu6s15iffvyZx\ne9E6zaUQq18qqNHqZcIS+7N7jgajrQ2fBmrbIZ/R+3TqGTS3XT/M9dmZETNmkVrx05IxviCipTXG\ngYL6K0CNY2+RlbTHd2FF6/U6avdYvVynn2olXcHNgqR+P1D7iluRem9Hc52v/w6o/bjEioRct43r\n+jGHo1ajxNO5mztLpvoMQW0XeKufjjnY0W1wzC/2CMcsiB6MiByMIED4X5yUDTcCuEpELkSIPvsx\nAPeXVVqv11F7wAYMl93nR8+Dy7r5gvp1QI0maH5JdX7jFJwuqP8aqHE5R5EAzauomLlEgHuA+iNA\njc5hs6x8J3+deZeHzX8DUOOHjCKL893ktNoG1+GCGhVfnJBpTsXJbyEH4gzUZszI3z/DBJzrJqVd\nbKK5L3pOnBBp3mc51eofno7a1BnY2YKRZZM8kPsA1ut11DgSqJSdF43hnSRPaxPyuaZk+k7Jx71e\nr2OSOYIwV2dmvXZu1JZngfoDQG22bTNUA/uBYhYz/6v/PupfLtW9aMaUo/VPAjU6qNBZhBlfOOmZ\nYq/+n9NRO3GGu4srrAwau/9+yWh3nB767IjoPha4MWwTdH0/QW0/AP9sdG8Un9dpIlgE4Dc/th01\nK1PK8OVA/Q9A7Q9O3EVslHdSqdfrYaIDGpPdGjfvWAfVFwM1phn0+Xj3tHO+a7QPNJrZ8Pk2UJRl\nURDr9d1Qq82wgybDv9ZsHflec8LHdNSOIu3goRtkx61ioFmyvw1gAwB3WBS5Rap6iqo+ISLXIfBb\nbwM4JWmR0Q24Z7gbUAHAoDsRVKhQhm4wg2sVA3XJvrwJ/Sw0eM8KFSpU6Ar0kowZqjpsPwC1inb4\naYf7+hVtd1y/F2lH6m/4XLIrVKhQoUIS/TqYVKhQoUKFtYtqYq5QoUKFLkM1MVeoUKFCl6GamCtU\nqFChy7BWA+WLyE4IORQYrGAVgBtVdVk/530SIRjSY6p6ezPaChUqVOh1rDWOWUSmArgGwYPwfvsJ\ngKvNlTumvT/6fxKC395mAKZ72iFs7xYiMltEnhSRV0TkZYsxPVtE3u9oD3bnXSYij4rIPBHZaijr\nrOqt+gwVRhzWpijjqwD2VtXZqvpz+81G4IS/6mhjR9SvAThQQw7Bg5BPkDeUA/w6hHCnNVX9gKqO\nQogivNqOxTgv+v9DAM8hOJQvBnDJENdZ1Tu09fZSWwEAErCviBxhv31FRDxd4rxNRWQP/950a70j\nGmvLYBrAkwDGJvaPBfCU2/cIgC0RvPUfcscedtu3AZgKYEy0b4ztu93RPhT9vxQhHPlYAN8CcIOj\nfarJvfj2xvUucceWDGWdVb1Vn7ntgxBCVN1iY/xSALfavoMc7Zzo/ycBrATwSwRH+UO6rd516bc2\nZcx/B+D/ichyNCIkbIcQ6OhUR7sFgAcRRB0qIn+mqs+JyKYoRpkapxbNjlDV5wF8X0SmoBx7qSoj\nr1woIpPd8WdE5CwAV6jqCwBgXPXxKEZ4GC0ip1vbthARURt1yK9KhqLOqt6hrbeX2goAFwH4jIZE\nFRlE5CMAFiJkFyImRv+/C+CLqvqQiGyPwLUv7LJ61xmsNVGGqt4KYAcAMxG43NsAzACwox2Lacep\n6vaq+hErn7ND7wI43FX9jIicJXm54FYSZNrJAS4iZ8AGeHTMP4ujETj2u0RktYi8AqAO4AMI6bNi\n/BRBBr4pgJ8B+KC1YwyAJRHdUNSZqne11TtqkOsdqvZ2c71D3da6BBFcq/Ve3k+970WWASCHVUjH\nqiO2UNWHAEBVn0bxfei2ekc0et4lW0S2BHA2grUHAzq+gBCCdLaqro5op7vT56jqSzbAf6CqX4kP\nSrAi2QYhet4fo/0H+4+J0W4N4DfNaEVkH4SY1YtFZBcABwNYpqoxF5Gi3dlon0zRJs69UlWP64/O\naP/F33sJ3SQEncBS7cc6RppY0ojIvgj38aqIbIzQf3sgJOM7T1Vf7Yf2LxAiGHrabwK4XlX7jVnX\nKq2IbADgGACrVPVOEfkyQjTuJwD8RFXfjmg3BPCliPZYo13maY3+owhZEbdFiEr5FIB5qvpaoh2k\n3cZof5uiFZFpCBP7NWgwJttau67TEGSMtG8iiAwEwDgA26nqahF5D4BHVfXPu6nedQk9PzE3g4ic\noKqlkfCa0dqL+7cIL9UEAKep6i/s2EOqukdE+w0EcUxTWvswfA6BS7gDYeKqAzgQwG2q+r2oznZo\nb0zc0gEA/h0AVPWwAdLer6r72P8T7XncgCAXvEmD8jZFe5LRXl9C+ziA3VX1TyLyEwBvIIS7/rTt\nP6IJ7ZsIaUlTtK9aXSsAzAMwX1X/K3G/nvZqo30pQXcVQh9shBDWfhO7r08jvD+TE7QbI0TU3hTA\nAqOFqh4f0X4TwKEAfgXgEAAP2zmHI4TLrQ+E1uh3Rsi/6c1Sn3B0Y93tPqeqa0Tkgwix1hc4+o8j\nbe7aX72/V9W3m9TbUnvXKXQioO72H4CVA6UFsBTApvZ/HIAHECZcoKiAbInW6NZDeHFfA7C57d8I\ngZPwdbZK+xCAnyPEzd7fyufs//6O9uF2aKP/iwF8yP5vgsA1D5R2Wdx2d8wrs9qhfRhhqXwQQs6a\nlxAUSZMBbDYQWj5rhAn3BQDr2bYk+qEd2qXR8Y0B1O3/diXjqyXa6jcyfj3v+SfB3C31WwrA2462\nTAvgPWoiCQ2KiRqAz4nIBSgqIFul/ZOqvqOqbwJYobYMVdW3UAwX2w7tXgjK0nMAvKqBg3pLVe9S\n1bsc7Z5t0L5HRLYUkVEIE8NL1oY3APypA9rHROQE+/+IiOwFACKyA4rxzNuhVVV9V1VvV9WvIuQm\nmYMgAnp6gLTvMXHGZgiT4ha2f0MUZaDt0AINB68NEbhrqOrKTmilPfPRlmmbQURuGSitiGwuIrNE\n5EoROcYdm9NqvSMNa9Xzb4iwFUI+39VuvwC4rwPaF0RkgqouAQBV/aOIHApgLoBdB0i7RkQ2tsl2\nz+ziIlugONm2TKuq7yJYlsy38gWU9G07tGjPOqYd2hMBXCQi3wHwXwB+LSLPIsgYT+yANncdDTLd\nGwHcaPLpgdBehmDquR7Cx2y+iDyNYHlwjauzHdpLASwWkd8AmASAeTI/hEYO64HQXocglqppsE6i\nknCyHTtoILQisgfSELi8Uu3QIigzlyOIsqaIyJEAjtWQsm4i1lH0vIxZRC4DcLmqFpJFicg8jTKw\ntEm7DQLX+nyC9hOqem+7tCKyoeZzJJLmgwD+TFWXRvtapk3Q/BWAT6jqt8toBkIbnbMxgK1U9T86\noRWRzQF8BKaZVzMbK6mnX1oR2UFVf9viPbRD+2EAUNXfGyf5GQTRVyGfZZu0uyCYgz2mqk/204aW\naEXkKVXdsZVjbdK+g5CtMuX4MVFVNxog7RJtmK1CRM5BkKMfBuAOjXQ56xJ6fmKuUKFCAyJyO4A7\nkbaPPlBVPzNA2scAHK6qyxPXfFZVtx0g7TIAu9hKjvuOR0j2vKmqekXiOoGelzFXqFAhh9ju2ttH\nH9UB7QyUzxff6ID2JmR53QNU9WcAzgCwpqSOEY+KY65QYR2BdGA+2u20Iw3VxFyhwjoCEVmpqtuN\nRNqRhpFglVGhQgWDiDxadggJ89Feol2XUE3MFSqMLAyV+Wg30K4zqCbmChVGFm5GsGbwwY0gIvUe\np11nUMmYK1SoUKHLUJnLVahQoUKXoZqYK1SoUKHLUE3MFSpUqNBlqCbmChUqVOgy/H/y91bRYMiV\n/QAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11c74f470>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "inv_pca = PCA()\n",
    "pca_data = inv_pca.fit_transform(train_norm_data)\n",
    "\n",
    "feature_importance = np.nan_to_num(np.log(inv_pca.inverse_transform(np.eye(train_norm_data.shape[1]))))\n",
    "\n",
    "sns.heatmap(feature_importance, cmap=\"hot\", cbar=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYMAAAEACAYAAABRQBpkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n/z9FhJjfaOJs1wAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11cb57ac8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYEAAAEACAYAAABVtcpZAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nATOzAnMImJkVmEPAzKzAHAJmZgX2fzen51LE97lmAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11e63bdd8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "means = np.mean(pca_data, axis=0)/np.sum(np.mean(pca_data, axis=0))\n",
    "variances = np.var(pca_data, axis=0)/np.sum(np.var(pca_data, axis=0))\n",
    "xs = [x for x in range(train_norm_data.shape[1])]\n",
    "mean_threshold = [0.2 for x in xs]\n",
    "variance_threshold = [0.02 for x in xs]\n",
    "plt.plot(xs, means)\n",
    "plt.plot(xs, mean_threshold)\n",
    "plt.show()\n",
    "\n",
    "plt.plot(xs, variances)\n",
    "plt.plot(xs, variance_threshold)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## SVM with all features and with carefully selected features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def mse(real, pred):\n",
    "    return np.mean(np.square(real-pred))\n",
    "\n",
    "\n",
    "weights = compute_sample_weight('balanced', train_labels)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "73.37395659208056\n"
     ]
    }
   ],
   "source": [
    "def run_experiment(clf, X, y, val_X, val_Y, weights=None):\n",
    "    clf = clf\n",
    "    clf.fit(X,y)\n",
    "    preds = clf.predict(val_X)\n",
    "    curr_mse = mse(preds, val_Y)\n",
    "    return clf,curr_mse\n",
    "\n",
    "svr = SVR()\n",
    "_, svr_mse = run_experiment(svr, train_norm_data, train_labels, val_norm_data, val_labels)\n",
    "print(svr_mse)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#random baseline\n",
    "random_labels = np.random.normal(np.mean(train_labels), np.std(train_labels), len(val_labels))\n",
    "random_mse = mse(random_labels, val_labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "71.83710478844851\n"
     ]
    }
   ],
   "source": [
    "#majority baseline\n",
    "majority_labels = [np.mean(train_labels) for x in range(len(val_labels))]\n",
    "majority_mse = mse(majority_labels, val_labels)\n",
    "print(majority_mse)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3739"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_data.shape[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Choosing features above the thresholds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#choose features above the thresholds\n",
    "mean_based_features = np.where(means >= mean_threshold)[0]\n",
    "variance_based_features = np.where(variances >= variance_threshold)[0]\n",
    "chosen_features = np.array(list(set(mean_based_features).union(set(variance_based_features))))\n",
    "\n",
    "_, restricted_mse = run_experiment(svr, train_norm_data[:, chosen_features], train_labels, \\\n",
    "                                   val_norm_data[:, chosen_features], val_labels)\n",
    "print(restricted_mse)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "forest_regressor = RandomForestRegressor(100)\n",
    "_, forest_mse = run_experiment(forest_regressor, train_norm_data, train_labels, \\\n",
    "                                   val_norm_data, val_labels)\n",
    "print(forest_mse)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Generating the frontal ratios and using them as additional features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "is_frontal = np.array([x for x in range(len(dataset_columns)) if 'frontal' in dataset_columns[x]])\n",
    "is_suptent = np.array([x for x in range(len(dataset_columns)) if 'supratentorium' in dataset_columns[x]])\n",
    "#1 detect frontal\n",
    "#2 dodać objętości frontala -> wspolczynnik frontal w calym mozgu"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "structures_to_delete = ['thalamus', 'caudate', 'putamen', 'pallidum', 'volume', 'wm', 'supratentorium', 'csf']\n",
    "\n",
    "cortex_indices = []\n",
    "\n",
    "for i,column in enumerate(dataset_columns):\n",
    "    curr_deletes = []\n",
    "    for name in structures_to_delete: \n",
    "        if name in column:\n",
    "            curr_deletes.append(name)\n",
    "    if len(curr_deletes)==0 and i>=0:\n",
    "        cortex_indices.append(i)\n",
    "\n",
    "cortex_indices = np.array(cortex_indices)\n",
    "cortex_indices = cortex_indices[cortex_indices < 123]\n",
    "\n",
    "def generate_frontal_ratio(frontal_inds, reference_inds, data):\n",
    "    coefs = []\n",
    "    for observation in data:\n",
    "        frontal_volume = np.sum(observation[frontal_inds])\n",
    "        reference_volume = np.sum(observation[reference_inds])\n",
    "        coefs.append(frontal_volume/reference_volume)\n",
    "    return coefs\n",
    "\n",
    "frontal_coef = normalize_data(generate_frontal_ratio(is_frontal, is_suptent, train_data)).reshape(-1,1)\n",
    "frontal_coef2 = normalize_data(generate_frontal_ratio(is_frontal, cortex_indices, train_data)).reshape(-1,1)\n",
    "frontal_coef_val = generate_frontal_ratio(is_frontal, is_suptent, val_data)\n",
    "frontal_coef2_val = generate_frontal_ratio(is_frontal, cortex_indices, val_data)\n",
    "frontal_coef_val = normalize_data(frontal_coef_val).reshape(-1,1)\n",
    "frontal_coef2_val = normalize_data(frontal_coef2_val).reshape(-1,1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "train_norm_data_ratios = np.append(train_norm_data, np.hstack((frontal_coef, frontal_coef2)), axis=1)\n",
    "val_norm_data_ratios = np.append(val_norm_data, np.hstack((frontal_coef_val, frontal_coef2_val)), axis=1)\n",
    "_, preds_with_adds = run_experiment(svr, train_norm_data_ratios, train_labels, val_norm_data_ratios, val_labels) #only ratios\n",
    "print(preds_with_adds)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "xgb_reg = xgboost.XGBRegressor()\n",
    "_, xgb_mse = run_experiment(xgb_reg, train_norm_data, train_labels, val_norm_data, val_labels)\n",
    "print(xgb_mse)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "baseline = majority_mse\n",
    "results = [svr_mse, restricted_mse, forest_mse, preds_with_adds, xgb_mse]\n",
    "labels = ['svr', 'svr with restricted fts', 'forest regressors', 'svr with add. features', 'xgboost']\n",
    "\n",
    "plt.bar([x for x in range(len(results))], results, color='#f0eff7', tick_label=labels, align='center')\n",
    "plt.plot([x-0.5 for x in range(len(results)+1)], [baseline for x in range(len(results)+1)], color='red')\n",
    "plt.ylim(60,80)\n",
    "plt.xticks(rotation=90)\n",
    "plt.ylabel(\"MSE\")\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Experiments with augmented feature space"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "train_norm_data_e = np.nan_to_num(normalize_data(train_data_extra))\n",
    "val_norm_data_e = np.nan_to_num(normalize_data(val_data_extra))\n",
    "\n",
    "_, svr_mse_e = run_experiment(svr, train_norm_data_e, train_labels, val_norm_data_e, val_labels)\n",
    "print(svr_mse_e)\n",
    "\n",
    "\n",
    "_, forest_mse_e = run_experiment(forest_regressor, train_norm_data_e, train_labels, val_norm_data_e, val_labels)\n",
    "print(forest_mse_e)\n",
    "\n",
    "train_norm_data_e_ratios = np.append(train_norm_data_e, np.hstack((frontal_coef, frontal_coef2)), axis=1)\n",
    "val_norm_data_e_ratios = np.append(val_norm_data_e, np.hstack((frontal_coef_val, frontal_coef2_val)), axis=1)\n",
    "_, preds_with_adds_e = run_experiment(svr, train_norm_data_e_ratios, train_labels, \\\n",
    "                                    val_norm_data_e_ratios, val_labels) #only ratios\n",
    "print(preds_with_adds_e)\n",
    "\n",
    "\n",
    "_, xgb_mse_e = run_experiment(xgb_reg, train_norm_data_e, train_labels, val_norm_data_e, val_labels)\n",
    "print(xgb_mse)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "baseline = majority_mse\n",
    "results = [svr_mse_e,  forest_mse_e, preds_with_adds_e, xgb_mse_e]\n",
    "labels = ['svr', 'forest regressors', 'svr with add. features', 'xgboost']\n",
    "\n",
    "plt.bar([x for x in range(len(results))], results, color='#f0eff7', tick_label=labels, align='center')\n",
    "plt.plot([x-0.5 for x in range(len(results)+1)], [baseline for x in range(len(results)+1)], color='red')\n",
    "plt.ylim(60,80)\n",
    "plt.xticks(rotation=90)\n",
    "plt.ylabel(\"MSE\")\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "prefrontal_list = pd.read_csv('data/prefrontal_list.csv', header=None)\n",
    "prefrontal_ids = [list(dataset_columns).index(x)-1 for x in list(prefrontal_list.values[0])]\n",
    "for x in range(123,train_norm_data_e.shape[1]):\n",
    "    prefrontal_ids.append(x)\n",
    "    \n",
    "prefrontal_ids = np.array(prefrontal_ids)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "train_norm_data_extra = np.append(train_norm_data_e, np.hstack((frontal_coef, frontal_coef2)), axis=1)[:,prefrontal_ids]\n",
    "val_norm_data_extra = np.append(val_norm_data_e, np.hstack((frontal_coef_val, frontal_coef2_val)), axis=1)[:,prefrontal_ids]\n",
    "svr.fit(train_norm_data_extra, train_labels, sample_weight=weights)\n",
    "preds_with_adds = svr.predict(val_norm_data_extra) #only ratios\n",
    "print(mse(preds_with_adds, val_labels))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "x = compute_covariance(normalize_data(train_additional_features))\n",
    "fig, ax = plt.subplots(figsize=(10,10))\n",
    "im = ax.matshow(x, interpolation='none')\n",
    "fig.colorbar(im, ticks=[-1,-0.8,-0.6,-0.4,-0.2,0.0,0.2,0.4,0.6,0.8,1.0])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "tsne = TSNE(n_components=2, random_state=42)\n",
    "reduced_space = tsne.fit_transform(train_norm_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# plt.scatter(reduced_space[:,0], reduced_space[:,1])\n",
    "# threshold = 0\n",
    "#kde\n",
    "# plt.scatter(reduced_space[np.where(train_labels <= threshold)[0], 0], reduced_space[np.where(train_labels <= threshold)[0], 1], color='red', alpha=0.4)\n",
    "# plt.scatter(reduced_space[np.where(train_labels > threshold)[0], 0], reduced_space[np.where(train_labels > threshold)[0], 1], color='blue', alpha=0.4)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Convert regression problem to classification problem"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# classes_bounds = {\n",
    "#      '{}_-15'.format(min(train_labels)): -2,\n",
    "#     '-15_0': -1,\n",
    "#     '0_15': 1,\n",
    "#     '15_{}'.format(max(train_labels)+1): 2\n",
    "# }\n",
    "classes_bounds = {\n",
    "    '{}_0'.format(min(train_labels)): 0,\n",
    "    '0_{}'.format(max(train_labels)+1): 1\n",
    "}\n",
    "\n",
    "train_classes = []\n",
    "for label in train_labels:\n",
    "    for key in classes_bounds.keys():\n",
    "        keyx = key.split(\"_\")\n",
    "        if label >= float(keyx[0]) and label < float(keyx[1]):\n",
    "            train_classes.append(classes_bounds[key])\n",
    "\n",
    "val_classes = []\n",
    "for label in val_labels:\n",
    "    for key in classes_bounds.keys():\n",
    "        keyx = key.split(\"_\")\n",
    "        if label >= float(keyx[0]) and label < float(keyx[1]):\n",
    "            val_classes.append(classes_bounds[key])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.svm import SVC\n",
    "from sklearn.utils.class_weight import compute_class_weight, compute_sample_weight\n",
    "\n",
    "class_weights = compute_sample_weight('balanced', train_classes)\n",
    "\n",
    "svm = SVC()\n",
    "svm.fit(train_norm_data_extra[:,-2:-1], train_classes, sample_weight=class_weights)\n",
    "# svm.fit(train_norm_data, train_class;es)\n",
    "pre = svm.predict(val_norm_data_extra[:,-2:-1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "accuracy_score(pre, val_classes)\n",
    "# accuracy_score(svm.predict(train_norm_data), train_classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "plt.hist(val_classes, bins=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "plt.hist(pre, bins=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [Root]",
   "language": "python",
   "name": "Python [Root]"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}