dask.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Populating the interactive namespace from numpy and matplotlib\n"
     ]
    }
   ],
   "source": [
    "import dask.dataframe as dd\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from lightgbm import LGBMClassifier, cv, Dataset, train, plot_importance, LGBMRegressor\n",
    "import re\n",
    "import datetime\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "%pylab inline\n",
    "from util import *\n",
    "import pickle\n",
    "from sklearn.model_selection import ParameterGrid, StratifiedKFold, cross_val_score, KFold\n",
    "from tqdm import tqdm\n",
    "import scipy.stats\n",
    "from sklearn.metrics import roc_auc_score, mean_squared_error\n",
    "from sklearn.cluster import KMeans, MeanShift, estimate_bandwidth\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.preprocessing import normalize, StandardScaler\n",
    "from catboost import CatBoostClassifier, CatBoostRegressor\n",
    "from sklearn.base import clone\n",
    "from dateutil.relativedelta import relativedelta\n",
    "import json\n",
    "from dask.multiprocessing import get\n",
    "import re\n",
    "from collections import defaultdict\n",
    "from xgboost import XGBClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/light/Projects/kaggle/rosbank/util.py:49: FutureWarning: Sorting because non-concatenation axis is not aligned. A future version\n",
      "of pandas will change to not sort by default.\n",
      "\n",
      "To accept the future behavior, pass 'sort=True'.\n",
      "\n",
      "To retain the current behavior and silence the warning, pass sort=False\n",
      "\n",
      "  df = pd.concat([df, pd.DataFrame(date_spaces)])\n"
     ]
    }
   ],
   "source": [
    "# Data loading, base preprocessing in module util.py\n",
    "df = load('train.csv')\n",
    "Y = df.groupby('cl_id')['target_flag'].first()\n",
    "Y_reg = df.groupby('cl_id')['target_sum'].first()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Prepare table for parallel computation\n",
    "df_dask = dd.from_pandas(df, npartitions=8)\n",
    "trx_categories = df['trx_category'].unique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/light/Projects/kaggle/rosbank/util.py:49: FutureWarning: Sorting because non-concatenation axis is not aligned. A future version\n",
      "of pandas will change to not sort by default.\n",
      "\n",
      "To accept the future behavior, pass 'sort=True'.\n",
      "\n",
      "To retain the current behavior and silence the warning, pass sort=False\n",
      "\n",
      "  df = pd.concat([df, pd.DataFrame(date_spaces)])\n"
     ]
    }
   ],
   "source": [
    "# Test dataframe data loading, preprocessing and preparation for parallel computing\n",
    "test = dd.from_pandas(load('test.csv'), npartitions=8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/light/miniconda3/lib/python3.6/site-packages/ipykernel_launcher.py:150: UserWarning: `meta` is not specified, inferred from partial data. Please provide `meta` if the result is unexpected.\n",
      "  Before: .apply(func)\n",
      "  After:  .apply(func, meta={'x': 'f8', 'y': 'f8'}) for dataframe result\n",
      "  or:     .apply(func, meta=('x', 'f8'))            for series result\n"
     ]
    }
   ],
   "source": [
    "def cl_features(d):\n",
    "    \"\"\"\n",
    "    Feature preparation\n",
    "        arg: d - pd.DataFrame - Frame for each customer\n",
    "    \"\"\"\n",
    "    result = dict()\n",
    "    trx_categories = ['POS', 'DEPOSIT', 'CACH_OUT']\n",
    "    test_mode = d['trx_category'].iloc[0] == 'foo'\n",
    "    d = d.sort_values('TRDATETIME')\n",
    "    if 'target_flag' in d.columns:\n",
    "        result['target_flag'] = d.iloc[0]['target_flag']\n",
    "        result['target_sum'] = d.iloc[0]['target_sum']\n",
    "    \n",
    "    # Time features\n",
    "    min_pos = d['day'][d['trx_category']=='POS'].min()\n",
    "        \n",
    "    result['cl_id'] = d.iloc[0]['cl_id']\n",
    "    \n",
    "    result['min_month'] = d['month'].min()\n",
    "    result['max_month'] = d['month'].max()\n",
    "        \n",
    "    result['month_diff'] = result['max_month']-result['min_month']\n",
    "    result['min_day'] = d['day'].min()\n",
    "    result['max_day'] = d['day'].max()\n",
    "    result['day_diff'] = result['max_day']-result['min_day']\n",
    "        \n",
    "    min_hour = 2\n",
    "    max_hour = 23\n",
    "    result['night'] = ((d['hour']<=min_hour)| (d['hour']>=max_hour)).sum()/(((d['hour']>min_hour)& (d['hour']<max_hour)).sum()+1)\n",
    "    result['night2'] = (((d['hour']>0) & (d['hour']<=min_hour))| (d['hour']>=max_hour)).sum()/(((d['hour']>min_hour)& (d['hour']<23)).sum()+1)\n",
    "\n",
    "    result['full_night'] = (d['hour']==0).sum()\n",
    "    result['mean_day'] = d['day'].mean()\n",
    "    \n",
    "    d['day_d'] = d['day']-result['min_day']\n",
    "    result['day_std'] = d['day_d'].std()\n",
    "    result['start_day_calendar'] = d['TRDATETIME'].min().day\n",
    "    result['end_day_calendar'] = d['TRDATETIME'].max().day\n",
    "    result['3_months2'] = d['TRDATETIME'].min()+relativedelta(months=3)>d['TRDATETIME'].max()\n",
    "    \n",
    "    amounts = defaultdict(list)\n",
    "    counts = defaultdict(list)\n",
    "    # By month calendar\n",
    "    for i in range(5):\n",
    "        start = result['min_month']+i\n",
    "        end = result['min_month']+i+1\n",
    "        \n",
    "        last_n = d[['trx_category', 'amount', 'month', 'MCC', 'MCC_group', 'TRDATETIME']][((d['month']>=start)&(d['month']<end))]\n",
    "        \n",
    "        result[f'month_mcc_unique_{i}'] = len(last_n['MCC'].unique())\n",
    "        result[f'month_trx_cat_unique_{i}'] = len(last_n['trx_category'].unique())\n",
    "        \n",
    "        user_trx_categories = last_n['trx_category'].value_counts()\n",
    "        user_trx_categories_amount = last_n.groupby('trx_category')['amount'].sum()\n",
    "        \n",
    "        for trx_category in trx_categories:\n",
    "            result[f'trx_category_{i}'+trx_category] = user_trx_categories.get(trx_category, 0)\n",
    "            result[f'trx_category_amount_{i}'+trx_category] = user_trx_categories_amount.get(trx_category, 0)\n",
    "            \n",
    "            counts[trx_category].append(user_trx_categories.get(trx_category, 0))\n",
    "            amounts[trx_category].append(user_trx_categories_amount.get(trx_category, 0))\n",
    "            \n",
    "        result['month_count_'+str(i)] = last_n.shape[0]\n",
    "        result['month_amountsum_'+str(i)] = last_n['amount'].sum()\n",
    "        result['dep_wd'] = last_n['amount'][last_n['trx_category']=='CACH_OUT'].sum() - last_n['amount'][last_n['trx_category']=='DEPOSIT'].sum()\n",
    "        \n",
    "        # By all period\n",
    "        if i==0:\n",
    "            end = 10**6\n",
    "            last_n = d[['trx_category', 'amount', 'month']][((d['month']>=start)&(d['month']<end))]\n",
    "            result['month_count2_'+str(i)] = last_n.shape[0]\n",
    "            result['month_amountsum2_'+str(i)] = last_n['amount'].sum()\n",
    "\n",
    "\n",
    "            user_trx_categories = last_n['trx_category'].value_counts()\n",
    "            user_trx_categories_amount = last_n.groupby('trx_category')['amount'].mean()\n",
    "            \n",
    "            for trx_category in trx_categories:\n",
    "                result['trx_category2_{i}'+trx_category] = user_trx_categories.get(trx_category, 0)\n",
    "                result['trx_category_amount2_{i}'+trx_category] = user_trx_categories_amount.get(trx_category, 0)\n",
    "    \n",
    "    # Index of max value in array\n",
    "    for key, value in amounts.items():\n",
    "        result[f'max_{key}_amount_month'] = np.argmax(value)\n",
    "        \n",
    "    # Index of max value in array\n",
    "    for key, value in counts.items():\n",
    "        result[f'max_{key}_count_month'] = np.argmax(value)\n",
    "                \n",
    "    # Last month\n",
    "    min_pos_date = d['TRDATETIME'][d['trx_category']=='POS'].min()\n",
    "    start = min_pos_date+relativedelta(months=2) if not test_mode else datetime.datetime.now()\n",
    "    end = min_pos_date+relativedelta(months=4) if not test_mode else datetime.datetime.now()\n",
    "\n",
    "    last_n = d[['trx_category', 'amount', 'month', 'MCC', 'MCC_group']][((d['TRDATETIME']>=start)&(d['TRDATETIME']<end))]\n",
    "    result[f'month_mcc_unique3'] = len(last_n['MCC'].unique())\n",
    "    \n",
    "    user_trx_categories = last_n['trx_category'].value_counts()\n",
    "    user_trx_categories_amount = last_n.groupby('trx_category')['amount'].sum()\n",
    "    \n",
    "    for trx_category in trx_categories:\n",
    "        result[f'trx_category3'+trx_category] = user_trx_categories.get(trx_category, 0)\n",
    "        result[f'trx_category_amount3'+trx_category] = user_trx_categories_amount.get(trx_category, 0)\n",
    "\n",
    "    result['month_count3'] = last_n.shape[0]\n",
    "    result['month_amountsum3'] = last_n['amount'].sum()\n",
    "    \n",
    "    # Last N transactions\n",
    "    for r in [15, 30]:\n",
    "        last_n = d.tail(r)\n",
    "        user_trx_categories = last_n['trx_category'].value_counts()\n",
    "        user_trx_categories_amount = last_n.groupby('trx_category')['amount'].sum()\n",
    "        \n",
    "        for trx_category in trx_categories:\n",
    "            result[f'trx_category6_{r}'+trx_category] = user_trx_categories.get(trx_category, 0)\n",
    "            result[f'trx_category_amount6_{r}'+trx_category] = user_trx_categories_amount.get(trx_category, 0)\n",
    "        \n",
    "    # By transaction type - time features\n",
    "    \n",
    "    result['has_deposit'] = (d['trx_category']=='DEPOSIT').sum()>0\n",
    "    for c in trx_categories:\n",
    "        dd = d['day'][d['trx_category']==c]\n",
    "        dt = d['TRDATETIME'][d['trx_category']==c]\n",
    "        dm = d['month'][d['trx_category']==c]\n",
    "        da = d['amount'][d['trx_category']==c]\n",
    "        \n",
    "        not_empty = dd.shape[0]>0\n",
    "    \n",
    "        result[f'min_day_{c}'] = dd.min() if not_empty else -1\n",
    "        result[f'max_day_{c}'] = dd.max() if not_empty else -1\n",
    "        result[f'day_diff_{c}'] = result[f'max_day_{c}'] - result[f'min_day_{c}'] if not_empty else -1\n",
    "        \n",
    "        result[f'min_month_{c}'] = dm.min() if not_empty else -1\n",
    "        result[f'max_month_{c}'] = dm.max() if not_empty else -1\n",
    "        \n",
    "        result[f'month_diff_{c}'] = result[f'max_month_{c}'] - result[f'min_month_{c}'] if not_empty else -1\n",
    "        dd -= result[f'min_day_{c}']\n",
    "        result[f'day_std_{c}'] = dd.std() if not_empty else -1\n",
    "         \n",
    "    if not test_mode:\n",
    "        d = d.loc[d['day']>=min_pos]\n",
    "        \n",
    "    # Index of transaction with Max amount by category\n",
    "    result['3_months'] = (d['TRDATETIME'].max()-d['TRDATETIME'].min()).days>=90\n",
    "    min_day = d['day'].min()\n",
    "    pos_tr = d.loc[d['trx_category']=='POS']\n",
    "    result['pos_sum'] = pos_tr['amount'].sum()\n",
    "    max_pos_index = pos_tr['amount'].idxmax() if not test_mode else 0\n",
    "    result['max_day_idx_pos'] = pos_tr.loc[max_pos_index]['day']-min_day if not test_mode else 0\n",
    "    \n",
    "    out_tr = d.loc[d['trx_category'].isin(['CACH_OUT', 'WD_ATM_PARTNER', 'WD_ATM_ROS', 'WD_ATM_OTHER'])]\n",
    "    max_out_index = out_tr['amount'].idxmax() if out_tr.shape[0]>0 else -1\n",
    "    result['max_day_idx_out'] = out_tr.loc[max_out_index]['day']-min_pos if out_tr.shape[0]>0 else -1\n",
    "    \n",
    "    return pd.Series(result)\n",
    "\n",
    "def features(df):\n",
    "    agg = df.groupby('cl_id').apply(lambda x: cl_features(x)).fillna(0)\n",
    "    if 'target_flag' in agg.columns:\n",
    "        return agg.drop(['target_flag', 'target_sum'], 1), agg['target_flag'], agg['target_sum']\n",
    "    return agg\n",
    "\n",
    "X, _, _ = features(df_dask)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/light/miniconda3/lib/python3.6/site-packages/ipykernel_launcher.py:2: FutureWarning: 'cl_id' is both an index level and a column label.\n",
      "Defaulting to column, but this will raise an ambiguity error in a future version\n",
      "  \n"
     ]
    }
   ],
   "source": [
    "# Real code execution\n",
    "X = X.compute(get=get).sort_values('cl_id').sort_values('cl_id')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Removing outliers. No improvements on private score\n",
    "not_outliers = np.where(Y_reg<np.exp(14))[0]\n",
    "X2 = X.copy().iloc[not_outliers]\n",
    "Y2 = Y.copy().iloc[not_outliers]\n",
    "Y2_reg = Y_reg.copy().iloc[not_outliers]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8823952017427107\n",
      "767\n"
     ]
    }
   ],
   "source": [
    "# Кросс валидация\n",
    "params = {'bagging_fraction': 0.55, 'bagging_freq': 2, 'colsample_bytree': 0.8, \n",
    "          'learning_rate': 0.01, 'max_bin': 255, 'max_depth': -1, 'metric': 'auc', \n",
    "          'min_data_in_leaf': 20, 'num_leaves': 16, 'objective': 'binary', 'reg_alpha': 0.4,\n",
    "          'reg_lambda': 0.55, 'verbose': -1\n",
    "         }\n",
    "\n",
    "_ = cv(params, Dataset(X2, Y2), verbose_eval=False, \n",
    "       seed=4, num_boost_round=2000, early_stopping_rounds=100, nfold=10\n",
    "      )\n",
    "print(max(_['auc-mean']))\n",
    "print(np.array(_['auc-mean']).argmax())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f7cb9c171d0>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABEcAAAOjCAYAAABZVGqUAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzs3XmUXVWB9+/vDnFgshHDJDQKhCGQVBJQEEUtxDAjDijYtEwOLTaNIAHsFpQX5PeiMgjSkzQCMhgVZBD8pUWxkHaCgEUYRIYGJEhDiIQZMrDfP+5NWSmSENGQ0Pt51qpF1ZnuPrfOUurDPueWWmsAAAAAWjVsaQ8AAAAAYGkSRwAAAICmiSMAAABA08QRAAAAoGniCAAAANA0cQQAAABomjgCALCElVL+rZRy9NIeBwCwYKXWurTHAACwQKWUe5KskWTuoMUb1Vp//2ccszfJebXWdf680b08lVLOTjKt1nrU0h4LACwrzBwBAJZ1u9VaVxr09aLDyF9CKWX40nz9P0cpZbmlPQYAWBaJIwDAy1Ip5S2llJ+XUmaWUm7szgiZt27/UspvSimPl1L+u5Tyd93lKyb5/5O8vpTyRPfr9aWUs0spXxy0f28pZdqgn+8ppRxZSpma5MlSyvDufheVUqaXUu4upRy8iLEOHH/esUspR5RSHiqlPFBKeW8pZedSyu2llD+UUv5p0L7HlFIuLKV8u3s+N5RSxg5aP6qU0td9H24ppbxnyOv+aynlB6WUJ5N8NMneSY7onvv3u9t9tpRyV/f4t5ZS3jfoGPuVUv6rlHJiKeWR7rnuNGj9qqWUs0opv++uv2TQul1LKf3dsf28lNKz2L9gAHgJiSMAwMtOKWXtJFck+WKSVZNMTHJRKWW17iYPJdk1yWuS7J/klFLK5rXWJ5PslOT3L2ImyoeT7JJklSTPJfl+khuTrJ1kuySHlFJ2WMxjrZnk1d19P5/kjCR/m2SLJG9PcnQpZb1B2++e5Lvdc70gySWllFeUUl7RHccPk6ye5B+SnF9K2XjQvn+T5PgkKyf5ZpLzk3y5e+67dbe5q/u6f5Xk/yQ5r5Sy1qBjbJXkt0lGJPlykjNLKaW77twkKyTZrDuGU5KklDI+yTeS/F2S1yX59ySXlVJetZjvEQC8ZMQRAGBZd0l35sHMQbMS/jbJD2qtP6i1PldrvTLJlCQ7J0mt9Ypa61214+p04sHb/8xxnFZrva/W+nSSNydZrdZ6bK11Vq31v9MJHHst5rFmJzm+1jo7yaR0osOptdbHa623JLk1ydhB219fa72wu/3J6YSVt3S/VkpyQnccVyW5PJ2QM8+ltdafdd+nZxY0mFrrd2utv+9u8+0kdyTZctAm99Zaz6i1zk1yTpK1kqzRDSg7JflkrfWRWuvs7vudJJ9I8u+11l/VWufWWs9J8mx3zACwTHnZ3jMLADTjvbXWHw1Z9oYkHyyl7DZo2SuS/CRJurd9fCHJRun8x6AVktz0Z47jviGv//pSysxBy5ZLcs1iHmtGNzQkydPdfz44aP3T6USP5712rfW57i0/r5+3rtb63KBt701nRsqCxr1ApZR9knwmyRu7i1ZKJ9jM8z+DXv+p7qSRldKZyfKHWusjCzjsG5LsW0r5h0HLXjlo3ACwzBBHAICXo/uSnFtr/fjQFd3bNi5Ksk86syZmd2eczLsNZEEf1fdkOgFlnjUXsM3g/e5LcnetdcMXM/gX4a/nfVNKGZZknSTzbgf661LKsEGBZN0ktw/ad+j5zvdzKeUN6cx62S7JL2qtc0sp/fnj+7Uo9yVZtZSySq115gLWHV9rPX4xjgMAS5XbagCAl6PzkuxWStmhlLJcKeXV3QedrpPO7IRXJZmeZE53Fsn2g/Z9MMnrSil/NWhZf5Kduw8XXTPJIS/w+tcmebz7kNblu2MYXUp581/sDOe3RSnl/d1PyjkkndtTfpnkV0meSucBq6/oPpR2t3Ru1VmYB5OsP+jnFdMJJtOTzsNsk4xenEHVWh9I5wG3/1JKeW13DO/orj4jySdLKVuVjhVLKbuUUlZezHMGgJeMOAIAvOzUWu9L5yGl/5TOH/X3JTk8ybBa6+NJDk7ynSSPpPNA0ssG7Xtbkm8l+e/uc0xen85DRW9Mck86zyf59gu8/tx0Hvg6LsndSR5O8h/pPNB0Sbg0yZ7pnM9Hkry/+3yPWenEkJ26Y/iXJPt0z3Fhzkyy6bxnuNRab01yUpJfpBNOxiT52Z8wto+k8wyV29J5EO4hSVJrnZLk40lO7477ziT7/QnHBYCXTKl1QTNLAQBYFpRSjkkystb6t0t7LADwv5WZIwAAAEDTxBEAAACgaW6rAQAAAJpm5ggAAADQNHEEAAAAaNrwpT0AWJBVVlmljhw5cmkPg2XIk08+mRVXXHFpD4NliGuCoVwTDOWaYDDXA0O5Jpau66+//uFa62pLexzziCMsk9ZYY41MmTJlaQ+DZUhfX196e3uX9jBYhrgmGMo1wVCuCQZzPTCUa2LpKqXcu7THMJjbagAAAICmiSMAAABA08QRAAAAoGniCAAAANA0cQQAAABomjgCAAAANE0cAQAAAJomjgAAAABNE0cAAACApokjAAAAQNPEEQAAAKBp4ggAAADQNHEEAAAAaJo4AgAAADRNHAEAAACaJo4AAAAATRNHAAAAgKaJIwAAAEDTxBEAAACgaeIIAAAA0DRxBAAAAGiaOAIAAAA0TRwBAAAAmiaOAAAAAE0TRwAAAICmiSMAAABA08QRAAAAoGniCAAAANA0cQQAAABomjgCAAAANE0cAQAAAJomjgAAAABNE0cAAACApokjAAAAQNPEEQAAAKBp4ggAAADQNHEEAAAAaJo4AgAAADRNHAEAAACaJo4AAAAATRNHAAAAgKaJIwAAAEDTxBEAAACgaeIIAAAA0DRxBAAAAGiaOAIAAAA0TRwBAAAAmiaOAAAAAE0TRwAAAICmiSMAAABA08QRAAAAoGniCAAAANA0cQQAAABomjgCAAAANE0cAQAAAJomjgAAAABNE0cAAACApokjAAAAQNPEEQAAAKBp4ggAAADQNHEEAAAAaJo4AgAAADRNHAEAAACaJo4AAAAATRNHAAAAgKaJIwAAAEDTxBEAAACgaeIIAAAA0DRxBAAAAGiaOAIAAAA0TRwBAAAAmiaOAAAAAE0TRwAAAICmiSMAAABA08QRAAAAoGniCAAAANA0cQQAAABomjgCAAAANE0cAQAAAJomjgAAAABNE0cAAACApokjAAAAQNPEEQAAAKBp4ggAAADQNHEEAAAAaJo4AgAAADRNHAEAAACaJo4AAAAATRNHAAAAgKaJIwAAAEDTxBEAAACgaeIIAAAA0DRxBAAAAGiaOAIAAAA0TRwBAAAAmiaOAAAAAE0TRwAAAICmiSMAAABA08QRAAAAoGniCAAAANA0cQQAAABomjgCAAAANE0cAQAAAJomjgAAAABNE0cAAACApokjAAAAQNPEEQAAAKBp4ggAAADQNHEEAAAAaJo4AgAAADRNHAEAAACaJo4AAAAATRNHAAAAgKaJIwAAAEDTxBEAAACgaeIIAAAA0DRxBAAAAGiaOAIAAAA0TRwBAAAAmiaOAAAAAE0TRwAAAICmiSMAAABA08QRAAAAoGniCAAAANA0cQQAAABomjgCAAAANE0cAQAAAJomjgAAAABNE0cAAACAppVa69IeAzzPuuuPrMM+dOrSHgbLkMPGzMlJNw1f2sNgGeKaYCjXBEO5JhjM9cBQL/aauOeEXZbAaNpTSrm+1vqmpT2OecwcAQAAAJomjgAAAABNE0cAAACApokjAAAAQNPEEQAAAKBp4ggAAADQNHEEAAAAaJo4AgAAADRNHAEAAIA/wTPPPJMtt9wyY8eOzWabbZYvfOELSZK77747W221VUaOHJk999wzs2bNSpKcffbZWW211TJu3LiMGzcu//Ef/5Ek6e/vz9Zbb53NNtssPT09+fa3vz3wGh/96EczduzY9PT0ZI899sgTTzyx0PH87ne/y0orrZQTTzxxvuVz587N+PHjs+uuuw4su+qqq7L55ptn9OjR2XfffTNnzpyBdX19fRk3blw222yzvPOd7xxYPnPmzOyxxx7ZZJNNMmrUqPziF79Ikhx++OHZZJNN0tPTk/e9732ZOXNmkmTWrFnZf//9M2bMmIwdOzZ9fX1Jkqeeeiq77LJLNtlkkyTZrJRywrzXKKV8spRyUymlv5TyX6WUTbvLX1FKOae77jellH8ctM83SikPlVJuHnzepZTjSilTu8f6YSnl9Qt987rEEQAAAPgTvOpVr8pVV12VG2+8Mf39/Zk8eXJ++ctf5sgjj8yhhx6aO++8M6997Wtz5plnDuyz5557pr+/P/39/fnYxz6WJFlhhRXyzW9+M7fccksmT56cQw45ZCAwnHLKKbnxxhszderUrLvuujn99NMXOp7PfOYz2WmnnZ63/NRTT82oUaMGfn7uueey7777ZtKkSbn55pvzhje8Ieecc06STgD51Kc+lcsuuyy33HJLvvvd7w7s9+lPfzo77rhjbrvtttx4440Dx5wwYUJuvvnmTJ06NRtttFH+7//9v0mSM844I0ly00035corr8xhhx2W5557LkkyceLE3HbbbUlya5K3lVLmDfyCWuuYWuu4JF9OcnJ3+QeTvKrWOibJFkn+rpTyxu66s5PsuIC35Cu11p7usS5P8vmFvnldSySOlFJWKaV8akkc+wVe973z6hLPV0rZ74WKWSnlzFLKjd3KdmEpZaXu8mNKKfd3y9vNpZT3DNrnE6WU27pf15ZSthm0btdSyq+7x7y1lPJ3S+4MAQAAlrxSSlZaaaUkyezZszN79uyUUnLVVVdljz32SJLsu+++ueSSSxZ5nI022igbbrhhkuT1r399Vl999UyfPj1J8prXvCZJUmvN008/nVLKAo9xySWXZL311stmm2023/Jp06bliiuuGAgxSTJjxoy88pWvzEYbbZSkEzcuuuiiJMkFF1yQ97///Vl33XWTJKuvvnqS5NFHH81Pf/rTfPSjH02SvPKVr8wqq6ySJNl+++0zfPjwJMlb3vKWTJs2LUly66235l3vetfAcVZZZZVMmTIlK6ywQrbddtt5w6lJbkiyTvc8Hxs0/BW76+dtt2IpZXiS5ZPMSvJYd5+fJvnD0PdkEcdaqCU1c2SVJM+LI92TWZLem2SJxpGX4ByWpP2SvNB0okNrrWNrrT1JfpfkoEHrTumWtw8m+UYpZVgpZdckf5dkm1rrJkk+meSCUsqapZRXJPl6kt1qrWOTjE/S9xc9IwAAgKVg7ty5GTduXFZfffVMmDAhG2ywQVZZZZWBWLDOOuvk/vvvH9j+oosuGrhF5r777nve8a699trMmjUrG2ywwcCy/fffP2uuuWZuu+22/MM//MPz9nniiSfypS99aeC2nsEOOeSQfPnLX86wYX/8s3/EiBGZM2dOpkyZkiS58MILB8Zy++2355FHHklvb2+22GKLfPOb30zSuVVotdVWy/7775/x48fnYx/7WJ588snnvd43vvGNgdkrY8eOzWWXXZY5c+bk7rvvzvXXX7+gc14uyW5JfjxvQSnl70spd6Uzc+Tg7uILkzyZ5IF0/kY9sdb6vCAyVCnl+FLKfUn2ztKaOZLkhCQbdGcZXFdKuaaUclmSW0spb+7OSnh1KWXFUsotpZTRCztQKeXI7r1FN867H6mU8vHucW8spVxUSlmhlPLWJO9J8pXu627Q/ZpcSrm+O4ZNuvtvUEr5Zfe4XyylPNFdXkopX+nOjLiplLJnd3nvkHM4tpRyyKAxHl9K+fRCxr9SKeXHpZQbusfcvbv8jd2ZFmeXUm4vpZxfSnl3KeVnpZQ7SilbdrdbtZRySfc9+2Uppae7/JhSysRBr3Nz95hv7N6HdUb3vf1hKWX5UsoeSd6U5Pzu+7P8gsY7r7CVTpZcPgsobLXW3ySZk2REkiOTHF5rfbi77oYk5yT5+yQrJxmeZEZ33bO11t8u7HcNAADwcrHccsulv78/06ZNy7XXXjvvVpEF2m233XLPPfdk6tSpmTBhQvbdd9/51j/wwAP5yEc+krPOOmu+mHHWWWfl97//fUaNGjXf80jmOeaYY3LooYcOzGKZ5/LLL8/qq6+eLbbYYr7lpZRMmjQphx56aLbccsusvPLKWW655ZIkc+bMyfXXX58rrrgi//mf/5njjjsut99+e+bMmZMbbrghBx54YH79619nxRVXzAknnDDfcY8//vgMHz48e++9d5LkgAMOyDrrrJM3velNOeSQQ/LWt7514HXmvVaS9ZOcVmv973nLa63/XGvdIJ2/M4/qLt4yydx0/kP/ekkOK6Wsv9A3+4/H+lyt9a+TnJ/5/6P/Ai2pWRCfTTK61jqulNKb5Iruz3cnSTcyfDGdP77Pq7XevKCDdO892j3JVrXWp0opq3ZXfa/WekZ3my8m+Wit9Wvd415ea72wu+7HST5Za72jlLJVkn9J8q4kpyY5tdb6rVLKJwe95PuTjEsyNp0//K8rpfy0u27zeedQOvc3fS/JV0spw5Lslc4vbEGeSfK+WutjpZQRSX7ZHWeSjExnFsYBSa5L8jdJtkkn8vxTOjNh/k+SX9da31tKeVeSb3bHuCgbJvlwrfXjpZTvJPlArfW8UspBSSbWWqcsaudSyllJdk7nHrDDFrB+qyTPJZmeZLMk1w/ZZEqSfWutf+ie673d38XlSb5Va31uIa/7iSSfSJIRI1bL58fMWdBmNGqN5ZPDXBMM4ppgKNcEQ7kmGMz1wFAv9pqY93DRwd74xjfmvPPOy/Tp0/PjH/84yy23XG655ZYsv/zyz9t+5MiRufbaaweWP/nkkzn00EOz995755lnnlng8TfeeON8/etfz3rrrTff8h/+8Ic577zzcvDBB+eJJ57IsGHDct999+Xhhx/OD3/4w3zve9/LrFmz8tRTT2XChAn53Oc+lyQ57rjjkiTXXXddVllllfT19WXWrFnZeOONc9111yVJNtxww1xwwQXp6enJiBEj8vTTT6evry8bbLBBLrjggmy33XZJksmTJ+f73/9+TjrppFx99dUDY9t9992z++67J0kOOuigzJw5c+DcvvSlLyXJM7XWry7kbZ6U5F+73/9Nksm11tlJHiql/Cyd//D/3wvZd6jzk/wgyfOn1wzyUt0icu28MNJ1bDox4Jn8carMgrw7yVm11qeSZNDUmdHdKLJKkpWS/OfQHUvnWRlvTfLdQfdmvar7z63TCQ9JckGSeY/03SadP97nJnmwlHJ1kjencz/TwDnUWu8ppcwopYxPskY68WLGQs6hJPn/SinvSCcorN3dJ0nurrXe1B3vLUl+XGutpZSbkrxx0Jg+0H3dq0opryulvGYR79m84/Z3v79+0LEWS611/1LKckm+lmTPJGd1Vx1aSvnbJI8n2bM71hc61sdKKWPS+V1OTDIhndt7FrTt19O5DSfrrj+ynnTTy/kOJv7SDhszJ64JBnNNMJRrgqFcEwzmemCoF3tN3LN3b6ZPn55XvOIVWWWVVfL000/n6KOPzpFHHpkZM2Zk+vTp2WuvvTJp0qTsv//+6e3tzQMPPJC11lorSXLxxRdn9OjR6e3tzaxZs7LTTjvlU5/6VA45ZODmhNRac9ddd2XkyJGptebyyy/P2972tvT29s43lqlTpw58f8wxx2SllVbKxIkT59umr68vJ554Yi6//PIkyUMPPZTVV189zz77bI477rh8/vOfT29vb9ZYY40cdNBB2WabbTJr1qz87ne/y5e//OWMHj06p5xyStZaa61svPHG6evry9vf/vb09vZm8uTJueyyy3L11VdntdVWG3jNp556KrXWrLjiirnyyiuz6qqrZr/99kuSHHXUUVlhhRWSZL77bEopG9Za7+j+uEuSed//Lp1JDueWUlZM8pYkC4sqCzrW7kkWPq2n66X6X4ehNyS9Lp2o8Yokr17A+hdydpL31lpvLKXsl6R3AdsMSzKz+4yMv4ShY/yPdP7IXzPJNxax395JVkuyRa11dinlnnTOOUmeHbTdc4N+fi4v/LuZk/lvi3r1oO8HH3duOjN0/iS11rmllElJjsgf48gptdYTh2x6azpPDL5q0LItktwy6Fg3JbmplHJukruzkDgCAADwcvDAAw9k3333zdy5c/Pcc8/lQx/6UHbddddsuumm2WuvvXLUUUdl/PjxAw8xPe2003LZZZdl+PDhWXXVVXP22WcnSb7zne/kpz/9aWbMmDGw7Oyzz05PT0/23XffPPbYY6m1ZuzYsfnXf+1MpLjssssyZcqUHHvssS9q7F/5yldy+eWX57nnnsuBBx448ODUUaNGZccdd0xPT0+GDRuWj33sYxk9uvMEjK997WvZe++9M2vWrKy//vo566zOn4gHHXRQnn322UyYMCFJ56Gs//Zv/5aHHnooO+ywQ4YNG5a111475557bpLOQ2KPP/74eR/lu2kppT/J6bXW/0hyUCnl3UlmJ3kkybx7j/45yVndCQUlnQkUU5OklPKtdHrAiFLKtCRfqLWemeSEUsrG6fxtfW86z8ZcpFLrCz609U9WSnldkhtqrW/o3lYzsda666D1l6UzTWa9JGvVWhd4/08pZcd0Hpzy7nm31XRv1Xg4nQevPpLO9Jj7a637lVK+1n3ds7r7/zydP+i/232GRk83qFyR5Ju11m93b+U4uda6Uinl/ek8XHTnJKumc3vIVkk2WcA5vDLJTekEng27s00WdA6fTjKy1voPpZRt04kI8+ZCXV5rHd3d7uzuzxd2b9u5vNY6upRyWpLptdbjuu/lKbXW8d0ZHLvWWvcqpWyezkycDRZw3IlJVqq1HlNK+X73XH+ykLGWJBvUWu/sfv+VJKm1TiylHJPkiaFxpHQ+teboJDvWWmeUUsYluaz7vj2e5E211r7utu9O8tV5Y1uUddcfWYd96NQX2oyG+K89DOWaYCjXBEO5JhjM9cBQL3rmyAm7LIHRtKeUcn2t9U1LexzzLJH/dej+kfyzUsrNSZ5O8uC8daWUfZLMrrVe0L114+ellHfVWq9awHEmd//YnlJKmZVOCPmndP4Y/1U6z7z4VToP/kw6weWMUsrBSfZIZ9bGv5ZSjkonYkxKcmOSQ5KcV0r5XJLJSR7t7n9xOrfc3JjOg0iPqLX+T+k+yHXI2GaVUn6SzuyUBYaRrvOTfL97q8yULMZ0niGOSeeTYaYmeSp/rGcXJdmnW89+leT2xTjW2Un+rZTydJKta61PD1lfkpzTvW2npPM+HLioA9ZaLyulrJ3O77GmE0T+ttb6QCll5SRHlFL+PZ3r4MmYNQIAAMAyZoml01rr3yxk+TfTeahoulFhqxc4zgnpfPrN4GX/mj8+nGXw8p/l+R/lu+MCDnt/krd0n5mxV5KNu/vXJId3vwYfty9DPoK2+yDWt6TzQNVFjf/hdILLgowetN1+g76/Z9667nNW3jt0x27Y2H4xjnvioO8vSieqLGyszyV520LWHbOI/Rb2+3g8nVk4AAAAsMxqdV7ZFklO7946MjOdT4tZbKWUTdP55JWLBz3kBQAAAHgZWibiSPfTTM4dsvjZWusiZ5W8WLXWa9L5uN4Xu/+t6Xwm84CX+hz+XKWUi/PHZ5/Mc2St9Xmf/AMAAAD/my0TcaT7aSZ/qU+VWSpebudQa33f0h4DAAAALAuGvfAmAAAAAP97iSMAAABA08QRAAAAoGniCAAAANA0cQQAAABo2jLxaTUw1PKvWC6/PWGXpT0MliF9fX25Z+/epT0MliGuCYZyTTCUa4LBXA8M5ZpgMDNHAAAAgKaJIwAAAEDTxBEAAACgaeIIAAAA0DRxBAAAAGiaOAIAAAA0TRwBAAAAmiaOAAAAAE0TRwAAAICmiSMAAABA08QRAAAAoGniCAAAANA0cQQAAABomjgCAAAANE0cAQAAAJomjgAAAABNE0cAAACApokjAAAAQNPEEQAAAKBp4ggAAADQNHEEAAAAaJo4AgAAADRNHAEAAACaJo4AAAAATRNHAAAAgKaJIwAAAEDTxBEAAACgaeIIAAAA0DRxBAAAAGiaOAIAAAA0TRwBAAAAmiaOAAAAAE0TRwAAAICmiSMAAABA08QRAAAAoGniCAAAANA0cQQAAABomjgCAAAANE0cAQAAAJomjgAAAABNE0cAAACApokjAAAAQNPEEQAAAKBp4ggAAADQNHEEAAAAaJo4AgAAADRNHAEAAACaJo4AAAAATRNHAAAAgKaJIwAAAEDTxBEAAACgaeIIAAAA0DRxBAAAAGiaOAIAAAA0TRwBAAAAmiaOAAAAAE0TRwAAAICmiSMAAABA08QRAAAAoGniCAAAANA0cQQAAABomjgCAAAANE0cAQAAAJomjgAAAABNE0cAAACApokjAAAAQNPEEQAAAKBp4ggAAADQNHEEAAAAaJo4AgAAADRNHAEAAACaJo4AAAAATRNHAAAAgKaJIwAAAEDTxBEAAACgaeIIAAAA0DRxBAAAAGiaOAIAAAA0TRwBAAAAmiaOAAAAAE0TRwAAAICmiSMAAABA08QRAAAAoGniCAAAANA0cQQAAABomjgCAAAANE0cAQAAAJomjgAAAABNE0cAAACApokjAAAAQNPEEQAAAKBp4ggAAADQNHEEAAAAaJo4AgAAADRNHAEAAACaJo4AAAAATRNHAAAAgKaJIwAAAEDTxBEAAACgaeIIAAAA0DRxBAAAAGiaOAIAAAA0TRwBAAAAmiaOAAAAAE0TRwAAAICmiSMAAABA08QRAAAAoGniCAAAANA0cQQAAABomjgCAAAANE0cAQAAAJomjgAAAABNE0cAAACApokjAAAAQNPEEQAAAKBpw5f2AGBBnp49N2/87BVLexgsQw4bMyf7uSYYxDXBUIu6Ju45YZeXeDQAwMuJmSMAAABA08QRAAAAoGniCAAAANA0cQQAAABomjgCAAAANE0cAQAAAJomjgAAAABNE0cAAACApokjAAAAQNPEEQCgCQcccEBWX331jB49emDZMccck7XXXjvjxo3LuHHj8oMf/CBJcuWVV2Y64OHyAAAgAElEQVSLLbbImDFjssUWW+Sqq65Kkjz++OMD244bNy4jRozIIYcckiQ5+eSTs+mmm6anpyfbbbdd7r333gWOo7e3NxtvvPHAMR566KEkyb333pvtttsuPT096e3tzbRp0wb2OfLIIzN69OiMHj063/72tweW33333dlqq60ycuTI7Lnnnpk1a9YLHuuII47IZpttllGjRuXggw9OrTVJMmvWrHziE5/IRhttlE022SQXXXRRkuR3v/tdtt1224wfPz49PT0v+B4N9p73vGe+9/sPf/hDJkyYkA033DATJkzII488Mt/21113XYYPH54LL7xw4b9IAFgCxBEAoAn77bdfJk+e/Lzlhx56aPr7+9Pf35+dd945STJixIh8//vfz0033ZRzzjknH/nIR5IkK6+88sC2/f39ecMb3pD3v//9SZLx48dnypQpmTp1avbYY48cccQRCx3L+eefP3CM1VdfPUkyceLE7LPPPpk6dWo+//nP5x//8R+TJFdccUVuuOGG9Pf351e/+lVOPPHEPPbYY0k60eTQQw/NnXfemde+9rU588wzF3msn//85/nZz36WqVOn5uabb851112Xq6++Okly/PHHZ/XVV8/tt9+eW2+9Ne985zuTJF/84hfzoQ99KL/+9a8zadKkfOpTn1rkezTP9773vay00krzLTvhhBOy3Xbb5Y477sh2222XE044YWDd3Llzc+SRR2b77bdf9C8SAJYAcYQXpZRyTCll4iLWH1tKefcClveWUi5fsqMDgOd7xzvekVVXXXWxth0/fnxe//rXJ0k222yzPP3003n22Wfn2+b222/PQw89lLe//e1Jkm233TYrrLBCkuQtb3nLfLM1Fsett96ad73rXQPHuvTSSweWv+Md78jw4cOz4oorpqenJ5MnT06tNVdddVX22GOPJMm+++6bSy65ZJHHKqXkmWeeyaxZs/Lss89m9uzZWWONNZIk3/jGNwYiyrBhwzJixIiBfebFmEcffXTgfVnUe/TEE0/k5JNPzlFHHTXfOV566aXZd999nzfeJPna176WD3zgAwOxCABeSuIIS0St9fO11h8t7XEAwAs5/fTT09PTkwMOOOB5t3kkyUUXXZTNN988r3rVq+ZbPmnSpOy5554ppTxvnzPPPDM77bTTQl9z//33z7hx43LccccN3NYyduzYfO9730uSXHzxxXn88cczY8aMjB07NpMnT85TTz2Vhx9+OD/5yU9y3333ZcaMGVlllVUyfPjwJMk666yT+++/f5HH2nrrrbPttttmrbXWylprrZUddtgho0aNysyZM5MkRx99dDbffPN88IMfzIMPPpikc+vReeedl3XWWSc777xzvva1r73ge3T00UfnsMMOG4hF8zz44INZa621kiRrrrnmwGvcf//9ufjii3PggQcu9D0DgCVJHGGxlFL2KaVMLaXcWEo5dzG2P7uUskf3+x1LKbeVUm5I8v4lPlgAWEwHHnhg7rrrrvT392ettdbKYYcdNt/6W265JUceeWT+/d///Xn7Tpo0KR/+8Ieft/y8887LlClTcvjhhy/wNc8///zcdNNNueaaa3LNNdfk3HM7/7d64okn5uqrr8748eNz9dVXZ+21185yyy2X7bffPjvvvHPe+ta35sMf/nC23nrrLLfccos8r4Ud684778xvfvObTJs2Lffff3+uuuqqXHPNNZkzZ06mTZuWt771rbnhhhuy9dZbZ+LEzgTRb33rW9lvv/0ybdq0/OAHP8hHPvKRPPfccwt9j/r7+3PXXXflfe973yLHWEoZCEuHHHJIvvSlL2XYMP9qCsDSMXxpD4BlXyllsyRHJXlrrfXhUsqqSQ5ezH1fneSMJO9KcmeSby9i208k+USSjBixWj4/Zs6fO3T+F1lj+eQw1wSDuCYYalHXRF9fX5Lkf/7nf/Lkk08O/DzYmDFjcsEFFwysmz59ej7zmc/kiCOOyH333Zf77rtvYNs777wzjz/+eB5//PH5jnX99dfntNNOy1e/+tX84he/WOhY77jjjiTJ5ptvnosvvjjrrrtukuTggzv/9/r000/nggsuSH9/f5LkbW97W972trclSY477rg888wzuemmmzJ9+vT8+Mc/znLLLZdbbrklyy+//MB4FnSsSZMmZY011siUKVOSJJtssknOO++87LXXXnn1q1+dVVddNX19fVlnnXVy2mmnpa+vL6eddlq+/OUvDxx35syZufTSS/Pa1752ge/RpZdemp///OdZc801M3fu3MycOTPjxo3LV7/61bzmNa/JRRddlNe97nWZMWNGVl555fT19eW//uu/cs011yTp3Lpz6aWX5rbbbss222yz0PdwcTzxxBML/F3TJtcDQ7kmGEwcYXG8K8l3a60PJ0mt9Q8LmkK8EJskubvWekeSlFLOSzeADFVr/XqSryfJuuuPrCfd5PLkjw4bMyeuCQZzTTDUoq6Je/bu7fzznnuy4oorpre38/MDDzwwcJvHKaeckq222iq9vb2ZOXNm3vnOd+bUU08deODqYJMnT84BBxwwcJwk+fWvf51/+Zd/yY9+9KNsuOGGCxzHnDlzMnPmzIwYMSKzZ8/O6aefnh122CG9vb15+OGHs+qqq2bYsGH53Oc+lwMPPDC9vb0DgeF1r3tdpk6dmgcffDATJ07M8OHDs/3222f69OnZa6+9MmnSpOy///6LPNaDDz6YM844I9tss01qrTnuuONyyCGHZNttt83uu++epPNpOmeffXbe/OY3p7e3N6NGjcpTTz2V3t7e/OY3v0mSvPe9782jjz66wPeot7c3p5xyysD7veuuuw5Enj333DN33HFHPvCBD+SEE07IXnvtld7e3jzwwAMD+++3337ZddddB56l8ufo6+ub73dE21wPDOWaYDBzFwGAJsy7JeW3v/1t1llnnZx55pk54ogjMmbMmPT09OQnP/nJwB/1p59+eu68884ce+yxz/vI3ST5zne+87xbag4//PA88cQT+eAHP5hx48blPe95z8C6cePGJUmeffbZ7LDDDunp6cm4ceOy9tpr5+Mf/3iSzr+kb7zxxtloo43y4IMP5nOf+1ySZPbs2Xn729+eTTfdNJ/4xCdy3nnnDTxn5Etf+lJOPvnkjBw5MjNmzMhHP/rRRR5rjz32yAYbbJAxY8Zk7NixGTt2bHbbbbeBYx1zzDHp6enJueeem5NOOilJctJJJ+WMM87I2LFj8+EPfzhnn312Sikv+B4tyGc/+9lceeWV2XDDDfOjH/0on/3sZ1/EbxIA/vLKvIeAwcJ0b6u5OMnWtdYZg26reaLWeuJC9jk7yeXdr9uTbFtrvauU8q0kK9dad13Ua667/sg67EOn/iVPg5c5swQYyjXBUIucOXLCLi/xaFgW+K/CDOZ6YCjXxNJVSrm+1vqmpT2OefxbJS+o1npLKeX4JFeXUuYm+XWSexZz32e6zxK5opTyVJJrkqy8xAYLAAAAfyJxhMVSaz0nyTl/wvb7Dfp+cjrPHgEAAIBljmeOAAAAAE0zc4Q/Synln5O8bcjiU2utZy2N8QAAAMCfShzhz1Jr/fulPQYAAAD4c7itBgAAAGiaOAIAAAA0TRwBAAAAmiaOAAAAAE0TRwAAAICm+bQalknLv2K5/PaEXZb2MFiG9PX15Z69e5f2MFiGuCYYyjUBALxYZo4AAAAATRNHAAAAgKaJIwAAAEDTxBEAAACgaeIIAAAA0DRxBAAAAGiaOAIAAAA0TRwBAAAAmiaOAAAAAE0TRwAAAICmiSMAAABA08QRAAAAoGniCAAAANA0cQQAAABomjgCAAAANE0cAQAAAJomjgAAAABNE0cAAACApokjAAAAQNPEEQAAAKBp4ggAAADQNHEEAAAAaJo4AgAAADRNHAEAAACaJo4AAAAATRNHAAAAgKaJIwAAAEDTxBEAAACgaeIIAAAA0DRxBAAAAGiaOAIAAAA0TRwBAAAAmiaOAAAAAE0TRwAAAICmiSMAAABA08QRAAAAoGniCAAAANA0cQQAAABomjgCAAAANE0cAQAAAJomjgAAAABNE0cAAACApokjAAAAQNPEEQAAAKBp4ggAAADQNHEEAAAAaJo4AgAAADRNHAEAAACaJo4AAAAATRNHAAAAgKaJIwAAAEDTxBEAAACgaeIIAAAA0DRxBAAAAGiaOAIAAAA0TRwBAAAAmiaOAAAAAE0TRwAAAICmiSMAAABA08QRAAAAoGniCAAAANA0cQQAAABomjgCAAAANE0cAQAAAJomjgAAAABNE0cAAACApokjAAAAQNPEEQAAAKBp4ggAAADQNHEEAAAAaJo4AgAAADRNHAEAAACaJo4AAAAATRNHAAAAgKaJIwAAAEDTxBEAAACgaeIIAAAA0DRxBAAAAGiaOAIAAAA0TRwBAAAAmiaOAAAAAE0TRwAAAICmiSMAAABA08QRAAAAoGniCAAAANA0cQQAAABomjgCAAAANE0cAQAAAJomjgAAAABNE0cAAACApokjAAAAQNPEEQAAAKBp4ggAAADQNHEEAAAAaJo4AgAAADRNHAEAAACaJo4AAAAATRNHAAAAgKaJIwAAAEDTxBEAAACgaeIIAAAA0DRxBAAAAGiaOAIAAAA0TRwBAAAAmiaOAAAAAE0TRwAAAICmiSMAAABA08QRAAAAoGniCAAAANA0cQQAAABo2p8cR0opry2l9CyJwQAAAAC81BYrjpRS+koprymlrJrkhiRnlFJOXrJDAwAAAFjyFnfmyF/VWh9L8v4k36y1bpXk3UtuWAAAAAAvjcWNI8NLKWsl+VCSy5fgeAAAAABeUosbR45N8p9J7qq1XldKWT/JHUtuWAAAAAAvjeGLs1Gt9btJvjvo5/9O8oElNSh4evbcvPGzVyztYbAMOWzMnOy3jF0T95ywy9IeAgAA8BewuA9k3aiU8uNSys3dn3tKKUct2aEBAAAALHmLe1vNGUn+McnsJKm1Tk2y15IaFAAAAMBLZXHjyAq11muHLJvzlx4MAAAAwEttcePIw6WUDZLUJCml7JHkgSU2KgAAAICXyGI9kDXJ3yf5epJNSin3J7k7yd5LbFQAAAAAL5EXjCOllGFJ3lRrfXcpZcUkw2qtjy/5oQEAAAAseS94W02t9bkkR3S/f1IYAQAAAP43WdxnjvyolDKxlPLXpZRV530t0ZEBAAAAvAQW95kje3b/+feDltUk6/9lhwMAAADw0lqsmSO11vUW8CWMAM2bOXNm9thjj2yyySYZNWpUfvGLX+Too49OT09Pxo0bl+233z6///3vkySPPvpodtttt4wdOzabbbZZzjrrrCTJT37yk4wbN27g69WvfnUuueSS573Wvffem+222y49PT3p7e3NtGnT5lv/2GOPZZ111slBBx00sGzHHXcceL1PfvKTmTt3bpLkD3/4QyZMmJANN9wwEyZMyCOPPJIkeeSRR/K+970vPT092XLLLXPzzTfP9xpz587N+PHjs+uuuw4s23vvvbPxxhtn9OjROeCAAzJ79uwkSV9fX/7qr/5q4LyOPfbYJMkzzzyTLbfccmBcX/jCFwaOtd9++2W99dYb2Ke/v3++17/uuusyfPjwXHjhhX/CbwkAABZtseJIKWWfBX0t6cEBLOs+/elPZ8cdd8xtt92WG2+8MaNGjcrhhx+eqVOnpr+/P7vuuutAFPjnf/7nbLrpprnxxhvT19eXww47LLNmzcq2226b/v7+9Pf356qrrsoKK6yQ7bff/nmvNXHixOyzzz6ZOnVqPv//2LvzeL+mQ2/8n5UQxHA1SKul0aCDDI6hTSnJCaWUVktLNc+vItw+qJY+qKgrt9XbK7l0UOMtbalLFEVabXVQB/VcNdRJgiItqRBDRVVEHoT1++N8c+5JcjJ0iKj9fr9e55Xvd629hr3P5vXKJ2uvPWFCTjjhhEXqTzrppIwcOXKRsssuuyxTp07NXXfdlT/+8Y+5/PLLkyQTJ07MrrvumhkzZmTXXXfNxIkTkyT//u//nra2tkybNi3f/e53c9RRRy3S3+mnn553vOMdi5SNGTMm9957b6ZPn5758+fn/PPP767beeedu89twoQJSZI11lgjv/zlLzN16tR0dnbm2muvzS233NLd5tRTT+1u09bW1l3+0ksv5fjjj+/12gAAwN9iRfcceWePn52TfCHJB1fSnBqplPKFUsqxK6HfzUopd7U+b19K+Ubr8xqllF+UUjpLKQeUUnYupdzd+r7WUvqZ36q/p5RybutNRimlDCml/LKUcl8pZUYp5aRSSmnVvb6Uck0pZWqr3Y//3ucIq8qf//zn3HjjjTnkkEOSJP369cv666+f9dZbr/uYefPmpfWfQ0opmTt3bmqtefbZZzNgwICsttqiTzdeccUV2XPPPdO/f/8lxrvnnnuyyy67JElGjx6dKVOmdNfdcccdefzxx5cIDhbOZcGCBXnhhRe65zJlypQcdNBBSZKDDjqoe6VKzzHe/va3Z+bMmXn88ceTJA8//HB+9KMf5dBDD11kjPe///0ppaSUkne9611LrGhZXCkl66yzTpLkxRdfzIsvvtg9r2U544wzst9++2XgwIHLPRYAAP4SK/pYzad7/Pxzkm2TrLNyp8bfW6319lrrZ1pft2mVtdVav5dkTJJTWt/nL6WL39da25IMT7JVkg+1gpQfJJlYa31bkq2T7JjkiFabk5P8vNa6da11qyTjV8rJwSrw4IMPZqONNsrBBx+cbbbZJoceemjmzZuXJDnxxBOz6aab5uKLL+5eOXLkkUfmt7/9bd74xjdm2LBhOf3009Onz6L/G7700ktz4IEH9jre1ltvnSuvvDJJctVVV2Xu3LmZM2dOXn755RxzzDE57bTTem33vve9LwMHDsy6666bj3zkI0mSxx9/PBtvvHGS5A1veEN3ANJzjFtvvTV/+MMfusOOo48+Ov/xH/+xxJwXevHFF3PRRRdljz326C777//+72y99dbZc889c/fdd3eXv/TSS2lra8vAgQOz2267ZcSIEd11J554YoYPH57Pfvazef7555MkjzzySK666qocfvjhvY4NAAB/ixVdObK4eUne8vecSBOVUk4spdxfSvlVkre1yv65lHJba6XF90sp/Usp65ZSHiylrN46Zr2e33vpd7tW+6npsYluKaW9tYpjYJL/SvLO1kqQ/51k/yRfKqVcvLx511oXJPm/SbZI8vEkN9daf9aqey7JkfmfEGTjJA/3aDvtL7tK8Oq1YMGC/OY3v8nhhx+eO++8M2uvvXb34ylf/vKXM2vWrIwZMyZnnnlmkuSnP/1p2traMnv27HR2dubII4/MM888093fo48+munTp+d973tfr+OddtppueGGG7LNNtvkhhtuyJve9Kb07ds3Z599dt7//vdnk0026bXdT3/60zz66KN5/vnn88tf/nKJ+oWrPpJk/Pjxefrpp9PW1pYzzjgj22yzTfr27ZtrrrkmAwcOzHbbbbfU63HEEUdk5MiR2XnnnZMk2267bf7whz9k6tSp+fSnP50PfehD3cf27ds3nZ2defjhh3Prrbd2721yyimn5N57781tt92Wp556KpMmTUrSFcxMmjRpqcEMAAD8LVbobTWllB+m6+00SVegslWSy1fWpJqglLJdko8laUvX7+E3Se5IcmWt9bzWMf+W5JBa6xmllI4keyW5utXuylrri0vp/jtJjqy13lhKOXXxylrrE6WUQ5McW2vduzXWDkmuqbUud5fDUkr/JLsmmZBkt9a8e/b/+1LKOqWU9ZKcleR7pZQjk/wiyXdqrbOX0u8nk3wySTbccKNMGLZgeVOhQV6/VnLMq+yeeOihh7Lhhhtm/vz56ejoyOabb55LLrkku+66a/cxgwcPzvjx4zN69Oicdtpp+fjHP54bbrghSfK6170uF198cfceHldccUVGjBiRm2++ealjfuYzXYu/5s+fn0suuSSdnZ25+uqrM3369Hz1q1/N/Pnzs2DBgjz11FP55Cc/uUjbt771rTn77LOz+uqrZ7311sv3v//9bLDBBpkzZ07WXXfddHR0JOl6zOaggw5KrTUHHnhgHnnkkUyePDk/+9nPcuWVV+aFF17Ic889l9122y0nnnhikuTCCy/MjBkzcvLJJ3f301P//v0zd+7cTJkyJf/0T/+0SN1mm22Ws846Kwcc0PVitPvuuy9Jss022+R73/teRo4cmV/96le56aabknQ9zjRlypQcccQRgZ6effbZXu8/mss9QU/uBxbnnqCnFX2Vb8+12guS/KHWuuyHylmenZNc1VppkVLKD1rlQ1uhyPrpenTpp63y85N8Ll3hyMFJ/rm3Tksp6ydZv9Z6Y6vooiR7/p3mvHkppTNdQdmUWutPSim7LatBrfWnpZTBSfZozePOUsrQWusfezn2m0m+mSRvHrxF/cr0Fb09aYJjhi3Iq+2emDlxn3zta1/LxhtvnLe97W3p6OjIzjvvnDe96U3Zcsstk3Ttk7Hddtulvb0922yzTZ566qm0t7fn8ccfz+OPP56PfvSj2XDDDZN0rdo45ZRT0t7e3ut4Tz75ZAYMGJA+ffrkxBNPzOGHH5729vZFjr/gggty++2358wzz8yzzz6buXPnZuONN86CBQtyzjnnZNddd017e3sOOOCAzJgxI/vtt18mTpyYj33sY2lvb8/TTz+d/v37p1+/fjnvvPOy++67Z6+99spee+3VPUZHR0dOO+20XHPNNUmS888/P/fdd1+uu+66rLXW/2xZ9Nhjj+X1r399Sim59dZb069fv3zwgx/Mk08+mdVXXz3rr79+5s+fn5NOOinHH3982tvb8+ijj2bjjTdOrTVXX311Ro0a1V2+0NixY7P33ntnww03XOq1opk6OjrcEyzCPUFP7gcW556gpxX9m8b7a63H9ywopUxavIy/iwuSfKjWOrWUMjZJe5LUWm9ubYranqRvrfWupfaw8izcc6Sne5Is8nqMVhjybK31mSSptT6V5JIkl5RSrmkd//1XYL6w0p1xxhkZM2ZMXnjhhQwePDjf+c53cuihh+a+++5Lnz59MmjQoJx77rlJut4mM3bs2AwbNiy11kyaNKk7GJk5c2ZmzZqVUaNGLdL/hAkTsv322+eDH/xgOjo6csIJJ6SUkpEjR+ass85a5tzmzZuXD37wg3n++efz8ssvZ/To0TnssMOSdAUx+++/f771rW9l0KBBueyyy5Ikv/3tb3PQQQellJIhQ4bkW9/61nKvwWGHHZZBgwZlhx12SJLsu+++mTBhQq644oqcc845WW211bLWWmvl0ksvTSkljz76aA466KC89NJLefnll7P//vt3vxp4zJgx+eMf/5haa9ra2rqvHQAArEyl1rr8g0r5Ta1128XKptVah6+0mb3GlVK2TVcQMiL/81jNf6Zrr46tkvwpyY+TPFJrHdtqc0ySY5J8qdZ6zjL6npbkiFrrr0opk5LsVWsd2gpWjq217t3zc6vNBVnGYzWllM1a9UMXK18ryd1JPllr/UXr++VJftp6HGiXJLfUWp8rpayb5NYkn6i13ras6/PmwVvUPvufvqxDaJhX58qRvZZ/ECuNf+1hce4JFueeoCf3A4tzT6xapZQ7aq3br+p5LLTMne1KKYeXUqYneVspZVqPnweT2Fjzb1Br/U2S7yWZmuQnSRaGBScl+XWSm5Pcu1izi5O8Lsnk5XR/cJKzWo/ALP/9mH+D1ptt9knyL6WU+5JMT9e5nNk6ZLskt7cCm/9Ocv7yghEAAAB4JS3vn2EvSddf3E/Joq9gndt6VIK/Qa31y0m+3EvV0laF7JTkilrr08vp9450vVJ3oc+1yjuSdCz+ufV97HL6nJlk6FLqpqf1+E8vdacmWWJTWAAAAHi1WGY4Umv9c5I/JzkwSVqvgF0zyTqllHVqrQ+t/CmSJKWUM9K1oen7V/VcAAAA4LVkRV/l+4EkX03yxiRPJBmU5LdJhqy8qdFTrfXTi5eVUs5K8p7Fik+vtX7nrx2nlDIsXW+46en5WuuIv7ZPAAAAeDVb0d0N/y3Ju5P8ota6TSlldJL/tfKmxYqotX5qJfQ5Pcnib6QBAACA16xlbsjaw4u11jlJ+pRS+tRar0/yqtlVFgAAAOCvtaIrR54upayT5KYkF5dSnkgyb+VNCwAAAOCVsaIrR/ZJ8lySo5Ncm+T3ST6wsiYFAAAA8EpZoZUjtdZ5pZRBSbastV5YSumfpO/KnRoAAADAyrdCK0dKKf+c5Iok/9kqelOSq1fWpAAAAABeKSu658inkrwrya+TpNY6o5QycKXNisZba/W+uW/iXqt6GryKdHR0ZOaY9lU9DQAA4DVoRfcceb7W+sLCL6WU1ZLUlTMlAAAAgFfOioYjN5RSPp9krVLKbkkuT/LDlTctAAAAgFfGioYj45P8Mcn0JP87yY+T/MvKmhQAAADAK2WZe46UUt5ca32o1vpykvNaPwAAAACvGctbOdL9RppSyvdX8lwAAAAAXnHLC0dKj8+DV+ZEAAAAAFaF5YUjdSmfAQAAAF4TlrnnSJKtSynPpGsFyVqtz2l9r7XW9Vbq7AAAAABWsmWGI7XWvq/URAAAAABWhRV9lS8AAADAa5JwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04VXvVe8AACAASURBVAgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGi01Vb1BKA38198KZuN/9GqngavIscMW5Cxr8A9MXPiXit9DAAA4NXFyhEAAACg0YQjAAAAQKMJRwAAAIBGE44AAAAAjSYcAQAAABpNOAIAAAA0mnAEAAAAaDThCAAAANBowhEAAACg0YQjAL3YbLPNMmzYsLS1tWX77bdPkkydOjU77LBDhg0blg984AN55plnkiQXX3xx2traun/69OmTzs7OJMn3vve9DB8+PEOGDMnxxx/f61gvvvhiDjrooAwbNizveMc7csopp3TXjRs3LgMHDszQoUMXaXPSSSdl+PDhaWtry+67757Zs2cvUn/bbbdltdVWyxVXXJEk+cMf/pBtt902bW1tGTJkSM4999wkydy5cxeZ+4Ybbpijjz66u82uu+6a4cOHp729PQ8//PAy+0qSyZMnZ9iwYRk+fHj22GOPPPnkk911Z5xxRt7+9rdnyJAh+dznPpckeeGFF3LwwQdn2LBh2XrrrdPR0fEX/JYAAODvQzgCsBTXX399Ojs7c/vttydJDj300EycODHTp0/Phz/84Zx66qlJkjFjxqSzszOdnZ256KKL8pa3vCVtbW2ZM2dOjjvuuFx33XW5++6789hjj+W6665bYpzLL788zz//fKZPn5477rgj//mf/5mZM2cmScaOHZtrr712iTbHHXdcpk2bls7Ozuy99945+eSTu+teeumlHH/88dl99927yzbeeOP893//dzo7O/PrX/86EydOzOzZs7Puuut2z72zszODBg3KvvvumyQ59thj84lPfCLTpk3LhAkTcsIJJyyzrwULFuSoo47K9ddfn2nTpmX48OE588wzu6/llClTMnXq1Nx999059thjkyTnnXdekmT69On5+c9/nmOOOSYvv/zy3/R7AwCAv5RwJEkpZf1SyhGrYNwPlVK2eqXHXZZSyqBSym9KKZ2llLtLKYf1qJtZSpleSplWSvlZKeUNrfJ/KqV8t5Tyu1LK71uf/6lV16eU8o1Syl2ttreVUt6yqs4P/hb3339/Ro4cmSTZbbfd8v3vf3+JYyZPnpyPfexjSZIHHnggW265ZTbaaKMkyXvf+95e25RSMm/evCxYsCDz589Pv379st566yVJRo4cmQEDBizRZmF9ksybNy+llO7vZ5xxRvbbb78MHDiwu6xfv35ZY401kiTPP/98rwHE/fffnyeeeCI777xzkuSee+7JLrvskiQZPXp0pkyZssy+aq2ptWbevHmpteaZZ57JG9/4xiTJOeeck/Hjx3e3Wzi3nmMMHDgw66+/fncYBQAArxThSJf1kywRjpRSVlvJ434oyUoNR/6Kc3g0yQ611rYkI5KML6W8sUf96Frr8CS3J/l8q+xbSR6otW5Ra908yYNJzm/VHZDkjUmG11qHJflwkqf/urOBV04pJbvvvnu22267fPOb30ySDBkypDsguPzyyzNr1qwl2n3ve9/LgQcemCTZYostct9992XmzJlZsGBBrr766l7bfOQjH8naa6+djTfeOG9+85tz7LHH9hqILO7EE0/Mpptumosvvrh75cgjjzySq666KocffvgSx8+aNSvDhw/PpptumuOPP747uFjo0ksvzQEHHNAdtGy99da58sorkyRXXXVV5s6dmzlz5iy1r9VXXz3nnHNOhg0blje+8Y255557csghhyTpCl5uuummjBgxIqNGjcptt93WPcYPfvCDLFiwIA8++GDuuOOOXq8RAACsTMKRLhOTbN5aLXFbKeWmUsoPktxTSnlna6XEmqWUtVurKYYuraNSyvGtFRJTSykTW2X/3Op3ainl+6WU/qWUHZN8MMmprXE3b/1cW0q5ozWHt7fab15KuaXV77+VUp5tlZdSyqk9VmUc0CpvX+wcTi6lHN1jjl8upRzV2/xrrS/UWp9vfV0jS79HbkyyRSlliyTbJflSj7qTk2xfStk8ycZJHq21vtzq/+Fa65+W8buAV4Vf/epX+c1vfpOf/OQnOeuss3LjjTfm29/+ds4+++xst912mTt3bvr167dIm1//+tfp379/9/4gr3vd63LOOefkgAMOyM4775zNNtssffv2XWKsW2+9NX379s3s2bPz4IMP5itf+UoeeOCB5c7xy1/+cmbNmpUxY8Z0P75y9NFHZ9KkSenTZ8n/dDfddNNMmzYtv/vd73LhhRfm8ccfX6T+0ksv7Q52kuS0007LDTfckG222SY33HBD3vSmN3XPv7e+XnzxxZxzzjm58847M3v27AwfPrx7/5QFCxbkqaeeyi233JJTTz01+++/f2qtGTduXDbZZJNsv/32Ofroo7Pjjjv2eo0AAGBlWtkrI/5RjE8ytNbaVkppT/Kj1vcHk6QVMvxbkrWS/Fet9a7eOiml7JlknyQjaq3PlVIW/tPvlbXW81rH/FuSQ2qtZ7T6vabWekWr7rokh9VaZ5RSRiQ5O8kuSU5PcnqtdXLPx1yS7JukLcnWSTZMclsp5cZW3bYLz6GUslmSK5N8vZTSJ8nHkrxraRejlLJp6xpskeS4WuvsXg7bO8n0dK186ay1vrSwotb6UimlM8mQJJcl+VUpZeck17Wu351LGfeTST6ZJBtuuFEmDFuwtCnSQK9fKznmFbgnem4IOmPGjCTJNttsk8mTJ+eAAw7I5z/ftWBq1qxZGThw4CLHn3XWWRkxYsQiZeuuu24mTZqUJPnhD3+YNddcc4lNR7/+9a9nq622ys0335wkGTx4cC688MKMHj06SfLYY49l3rx5S92sdPDgwRk/fnxGjx6dX/3qV7npppuSJH/+858zZcqU3Hvvvdlpp50WabPBBhvk3HPPzahRo5Ikv/vd7zJ37tzMnTt3kXE+85nPJEnmz5+fSy65pHuj2d76ev3rX58//elPmTVrVmbNmpUtt9wykydPzk477ZT+/ftn8ODBueGGG5J0bcQ6ZcqUrL/++tlnn32yzz77JEmOPPLIPP300yu0Meuzzz5rA1cW4Z5gce4JenI/sDj3BD0JR3p368JgpOXkJLcl+X9JPrOMdu9N8p1a63NJUmt9qlU+tBWKrJ9knSQ/XbxhKWWdJDsmubzH3gFrtP7cIV2P4CTJJUlOa33eKcnkVjDxeCnlhiTvTPJMz3Ootc4spcwppWyT5PVJ7qy1zlnaSdRaZyUZ3nqc5upSyhW11oX/xHx9KeWlJNOS/EuSkcu4Hqm1PlxKeVu6Qp5dklxXSvlorXWJXSlrrd9M8s0kefPgLepXprs9+R/HDFuQV+KemDmmPfPmzcvLL7+cddddN/PmzcvnP//5TJgwIVtttVUGDhyYl19+OWPHjs1xxx2X9vb2JMnLL7+cMWPG5KabbsrgwYO7+3viiScycODA/OlPf8rRRx+dyy67LG9961sXGfPXv/517r333rS3d439hz/8IZMmTcrw4cO75jRzZtZee+3usZKu4GbLLbdM0rXHyHbbbZf29vY8+uij3ceMHTs2e++9dz7ykY/k4YcfzgYbbJC11lorf/rTn/L73/8+//Ef/5Fhw4YlSa699tqMGzdukTGefPLJDBgwIH369MmJJ56Yww8/vPutNb31tcEGG+SLX/xihgwZko022ijXXXdd3vOe96S9vT3jxo3L7Nmz097envvvvz99+vTJPvvsk/nz56fWmrXXXjs///nPM2DAgIwdO3aFflcdHR2LzBfcEyzOPUFP7gcW556gJ3/77N28xb5vkK5QY/Uka/ZSvzwXJPlQrXVqKWVskvZejumT5OnWXh9/D4vP8fwkY5O8Icm3V6SDWuvsUspdSXZOckWreHSttfvdnKWUe5K0lVL6LHx0prU6pS3JPa1+nk/ykyQ/KaU8nq6gZ8lXdsCrxOOPP54Pf/jDSboeB/n4xz+ePfbYI6effnrOOuusJMm+++6bgw8+uLvNjTfemE033XSRYCRJjjrqqEydOjVJMmHChO5g5Ac/+EFuv/32nHzyyfnUpz6Vgw8+OEOGDEmtNQcffHB3MHLggQemo6MjTz75ZDbZZJN88YtfzCGHHJLx48fnvvvuS58+fTJo0KBFXqfbm9/+9rc55phjUkpJrTXHHntsdzCSJJdddll+/OMfL9Kmo6MjJ5xwQkopGTlyZPe5L6uvf/3Xf83IkSOz+uqrZ9CgQbnggguSdL2SeNy4cRk6dGj69euXCy+8MKWUPPHEE3nf+96XPn365E1velMuuuiiFf9FAQDA30mpta7qOaxypZQNkvym1jqo9VjNsbXWvXvU/yDJpUnekmTjWuuRS+lnjyQTkrx34WM1tdanSilPpuvxkz8l+XGSR2qtY0spZ7TG/U6r/f9N8rVa6+Wla/nI8Fag8qMk3621fq/16MlXa63rlFL2TfK/k7w/yYB0bZI6IsnbezmHful6DGb1JFv2fAxmsXPYJMmcWuv8Usrrkvw6yX611umllJlJtu8ZjrTaXJmuR2tObn2fkGTrWut+pZRtkzzWClr6pCsomlZrPS3L8ObBW9Q++5++rENomFds5cjEvVb6GPx9+NceFueeYHHuCXpyP7A498SqVUq5o9a6/aqex0I2ZE3SesTk5tYqiVN71pVSPpHkxVrrJenauPWdpZRdltLPtUl+kOT21p4bx7aqTkpXyHBzknt7NLk0yXGllDtbm5eOSXJIKWVqkrvTtX9Jkhyd5P+UUqalax+QP7fKr0rX4y1Tk/wyyedqrY8tZW4vJLk+yWVLC0Za3pHk16053JDktFrr9GUcnySHJHlr6XqN7++TvLVVliQDk/ywdW2nJVmQ5Mzl9AcAAACvGI/VtNRaP76U8u8m+W7r80vpWpmxrH4mpitE6Vl2TpJzejn25iz5Kt89eun2kSTvrrXWUsrHkryt1b4mOa7107PfjiQdPctaqzbeneSjy5n/z5MMX0rdZksp/1OS/7WUumuTXLusMQEAAGBVEo78Y9guyZmtR22eTjLuL2lcStkqyTVJrqq1zlgJ8wMAAIB/WMKRv0IpZViSxXcNfL7WusxVJX+tWutN6Xpd71/b/p4ki+wS+UqfAwAAALxaCUf+Cq09OP5eb5VZJV4L5wAAAAB/DzZkBQAAABpNOAIAAAA0mnAEAAAAaDThCAAAANBowhEAAACg0bythleltVbvm/sm7rWqp8GrSEdHR2aOaV/V0wAAAF6DrBwBAAAAGk04AgAAADSacAQAAABoNOEIAAAA0GjCEQAAAKDRhCMAAABAowlHAAAAgEYTjgAAAACNJhwBAAAAGk04AgAAADSacAQAAABoNOEIAAAA0GjCEQAAAKDRhCMAAABAowlHAAAAgEYTjgAAAACNJhwBAAAAGk04AgAAADSacAQAAABoNOEIAAAA0GjCEQAAAKDRhCMAAABAowlHAAAAgEYTjgAAAACNJhwBAAAAGk04AgAAADSacAQAAABoNOEIAAAA0GjCEQAAAKDRhCMAAABAowlHAAAAgEYTjgAAAACNJhwBAAAAGk04AgAAADSacAQAAABoNOEIAAAA0GjCEQAAAKDRhCMAAABAowlHAAAAgEYTjgAAAACNJhwBAAAAGk04AgAAADSacAQAAABoNOEIAAAA0GjCEQAAAKDRhCMAAABAowlHAAAAgEYTjgAAAACNJhwBAAAAGk04AgAAADSacAQAAABoNOEIAAAA0GjCEQAAAKDRhCMAAABAowlHAAAAgEYTjgAAAACNJhwBAAAAGk04AgAAADSacAQAAABoNOEIAAAA0GjCEQAAAKDRhCMAAABAowlHAAAAgEYTjgAAAACNJhwBAAAAGk04AgAAADSacAQAAABoNOEIAAAA0GjCEQAAAKDRhCMAAABAowlHAAAAgEYTjgAAAACNJhwBAAAAGk04AgAAADSacAQAAABoNOEIAAAA0GjCEQAAAKDRhCMAAABAowlHAAAAgEYTjgAAAACNJhwBAAAAGk04AgAAADSacAQAAABoNOEIAAAA0GjCEQAAAKDRhCMAAABAowlHAAAAgEYTjgAAAACNJhwBAAAAGk04AgAAADSacAQAAABoNOEIAAAA0GjCEQAAAKDRhCMAAABAowlHAAAAgEYTjgAAAACNJhwBAAAAGk04AgAAADSacAQAAABoNOEIAAAA0GjCEQAAAKDRhCMAAABAowlHAAAAgEYTjgAAAACNJhwBAAAAGk04AgAAADSacAQAAABoNOEIAAAA0GjCEQAAAKDRhCMAAABAowlHAAAAgEYTjgAAAACNJhwBAAAAGk04AgAAADSacAQAAABotNVW9QSgN/NffCmbjf/Rqp7Gq87MiXut6ikAAAC85lg5AgAAADSacAQAAABoNOEIAAAA0GjCEQAAAKDRhCMAAABAowlHAAAAgEYTjgAAAACNJhwBAAAAGk04AgAAADSacAT+wcyaNSujR4/OVlttlSFDhuT000/vrjvjjDPy9re/PUOGDMnnPve5JMnMmTOz1lprpa2tLW1tbTnssMO6j7/jjjsybNiwbLHFFvnMZz6TWusS45166qndbYcOHZq+ffvmqaeeSpKMGzcuAwcOzNChQxdpM3Xq1Oywww4ZNmxYPvCBD+SZZ55JksyZMyejR4/OOuuskyOPPLL7+Llz53aP0dbWlg033DBHH310kuSzn/1s2tracuihh+atb31r1l9//STJ9ddfv0ibNddcM1dffXWSZMyYMXnb296WoUOHZty4cXnxxReTJBdffHGGDx+eYcOGZccdd8zUqVO753D66adn6NChGTJkSL7+9a93l5900kkZPnx42trasvvuu2f27NlJknvvvTc77LBD1lhjjZx22ml/0e8QAAB4dRGOwD+Y1VZbLV/5yldyzz335JZbbslZZ52Ve+65J9dff32mTJmSqVOn5u67786xxx7b3WbzzTdPZ2dnOjs7c+6553aXH3744TnvvPMyY8aMzJgxI9dee+0S4x133HHdbU855ZSMGjUqAwYMSJKMHTu21zaHHnpoJk6cmOnTp+fDH/5wTj311CTJmmuumS996UtLhAnrrrtu9xidnZ0ZNGhQ9t133yTJ1772tXR2dub888/Ppz/96e7y0aNHdx//y1/+Mv3798/uu++epCscuffeezN9+vTMnz8/559/fpLkLW95S2644YZMnz49J510Uj75yU8mSe66666cd955ufXWWzN16tRcc801+d3vftd9/tOmTUtnZ2f23nvvnHzyyUmSAQMG5Bvf+MYi1xkAAPjHJBz5K5RS1i+lHNHje3sp5ZpVOae/RimlrZTy/pU8xqBSym9KKZ2llLtLKYctvxXLsvHGG2fbbbdN0hUqvOMd78gjjzySc845J+PHj88aa6yRJBk4cOAy+3n00UfzzDPP5N3vfndKKfnEJz7RvfJiaSZPnpwDDzyw+/vIkSO7g5Ke7r///owcOTJJsttuu+X73/9+kmTttdfOTjvtlDXXXHOpY9x///154oknsvPOOy93/IWuuOKK7Lnnnunfv3+S5P3vf39KKSml5F3velcefvjhJMmOO+6Y173udUmSd7/73d3lv/3tbzNixIj0798/q622WkaNGpUrr7wySbLeeut1jzNv3ryUUpJ0Xd93vvOdWX311Zd6LgAAwD8G4chfZ/0kRyz3qFe/tiQrNRxJ8miSHWqtbUlGJBlfSnnjSh6zMWbOnJk777wzI0aMyP3335+bbropI0aMyKhRo3Lbbbd1H/fggw9mm222yahRo3LTTTclSR555JFssskm3cdssskmeeSRR5Y61nPPPZdrr702++2333LnNWTIkEyZMiVJcvnll2fWrFkrfE6XXnppDjjggO4QYqHHHnssDz74YHbZZZde2/QWmrz44ou56KKLssceeyxR961vfSt77rlnkmTo0KG56aabMmfOnDz33HP58Y9/vMicTzzxxGy66aa5+OKLu1eOAAAArx2v+XCklLJZKeXeUsoFpZT7SykXl1LeW0q5uZQyo5TyrlLKgFLK1aWUaaWUW0opw1ttv1BK+XYppaOU8kAp5TOtbicm2by1GuLUVtk6pZQrWmNdXBb/m92ic5pQSrmtlHJXKeWbC49tjfO1UsrtpZTfllLeWUq5sjXPf+vR/v+02t5VSjm6x3ne1eOYY0spX+jR76RSyq2ta7BzKaVfkpOTHNA6jwNKKaNanztLKXeWUtZdfFVMKeXMUsrY1ueZpZRTWsffXkrZtpTy01LK7xeuEKm1vlBrfb7VfI004J57pTz77LPZb7/98vWvfz3rrbdeFixYkKeeeiq33HJLTj311Oy///6ptWbjjTfOQw89lDvvvDNf/epX8/GPf7x7D5C/xA9/+MO85z3v6XWlyOK+/e1v5+yzz852222XuXPnpl+/fis8ztKCjuuvvz4f+chH0rdv30XKH3300UyfPj3ve9/7lmhzxBFHZOTIkUusQrn++uvzrW99K5MmTUqSvOMd78jxxx+f3XffPXvssUfa2toWGefLX/5yZs2alTFjxuTMM89c4XMBAAD+May2qifwCtkiyUeTjEtyW5KPJ9kpyQeTfD7JrCR31lo/VErZJcl307WqIknenmR0knWT3FdKOSfJ+CRDW6shUkppT7JNkiFJZie5Ocl7kvxqKfM5s9Z6cqvtRUn2TvLDVt0LtdbtSylHJZmSZLskTyX5fSnla0k2S3JwulZhlCS/LqXckORPy7kGq9Va39V6jOZfa63vLaVMSLJ9rfXI1lx+mORTtdabSynrJPl/y+kzSR6qtba15nZB67zXTHJXknNb/W6a5Efp+j0cV2ud3VtHpZRPJvlkkmy44UaZMGzBCgzfLB0dHUmSBQsW5IQTTsiIESMyYMCAdHR0pH///hk8eHBuuOGGJMkLL7yQKVOmdG9gutAGG2yQyZMnZ8MNN8z999/f3ed1112XUkr398WdeeaZGTVq1BL1jz32WObNm7dE+ec///kkXRvIDhw4cJH6e++9N4888sgSbX73u99l7ty5mTt37hJ1v/jFL/LZz352ifIrrrgiI0aMyM0337xI+YUXXpgZM2bk5JNPXqTN73//+0yYMKF7T5SFNt9883zlK19Jkpx33nnZaKONlhhr8ODBGT9+fEaPHt1dtnDD26VdN1aeZ5991nVnEe4JFueeoCf3A4tzT9BTU8KRB2ut05OklHJ3kutqrbWUMj1dYcOgJPslSa31l6WUDUopCzca+FFr5cPzpZQnkrx+KWPcWmt9uDVGZ6vfpYUjo0spn0vSP8mAJHfnf8KRH7T+nJ7k7lrro60+H0iyabpCnatqrfNa5Vcm2blHu6W5svXnHa259ebmJF8tpVyc5Mpa68PLWACzUM/5rlNrnZtkbinl+VLK+rXWp2uts5IMbz1Oc3Up5Ypa6+OLd1Rr/WaSbybJmwdvUb8yvSm354qbOaY9tdYcdNBBec973rPIW1XGjRuX2bNnp729Pffff3/69OmTffbZJ08++WQGDBiQvn375oEHHsgf//jHfPSjH82AAQMyadKkrLnmmhkxYkQmTZqUT3/602lvb19i3D//+c+5++67c+2112bttddedE4zZ2bttddepN0TTzyRgQMH5uWXX87YsWNz3HHHLVI/c+bMPPvss0uMde2112bcuHFLlN97772ZN29ePvWpTy3xuM348eNzyimnLNLm/PPPz3333Zfrrrsua621Vnf5Qw89lEMPPTSXX355dtxxx0X6WTjnhx56KHfccUduueWWrL/++pkxY0a23HLLJF1vA9puu+0WGaujoyPrrLNOr9eNlaujo8N1ZxHuCRbnnqAn9wOLc0/QU1P+9vl8j88v9/j+crquwYsr2PalLP2ardBxpZQ1k5ydrhUbs1qPvvTcnbLn3Baf97J+Xwuy6CMri+94ubCvpc6t1jqxlPKjdO1DcnMp5X1/Qb/LnW+tdXbr0Z+dk1yxjHNhGW6++eZcdNFFGTZsWNrauhY4/fu//3vGjRuXcePGZejQoenXr18uvPDClFJy4403ZsKECVl99dXTp0+fnHvuud2Pxpx99tkZO3Zs5s+fnz333LN7D46Fb7RZ+Nrfq666KrvvvvsSwciBBx6Yjo6OPPnkk9lkk03yxS9+MYccckgmT56cs846K0my77775uCDD+5us9lmm+WZZ57JCy+8kKuvvjo/+9nPstVWWyVJLrvssvz4xz9e4pwvvfTS7LLLLksEIzNnzsysWbMyatSoRcoPO+ywDBo0KDvssEP3HCZMmJCTTz45c+bMyRFHdG0ZtNpqq+X2229Pkuy3336ZM2dOVl999Zx1zWUr7QAAEKZJREFU1lndK27Gjx+f++67L3369MmgQYO6r81jjz2W7bffPs8880z69OmTr3/967nnnnsW2cAVAAD4x9CUcGR5bkoyJsmXWo/IPFlrfWYZqybmpusxm7/GwnDhydajKx/JXxYU3JTkglLKxHQ9VvPhJP9fkseTDCylbJDk2XQ9qrPkO1YXtch5lFI2b62wmV5KeWe6Him6I8lWpZQ1kqyVZNcsfUXMEkopmySZU2udX0p5XbpWvnxtRduzpJ122im11l7r/uu//muJsv3222+pm6huv/32ueuuu5YoXxiKLDR27NiMHTt2ieMmT57ca79HHXVUjjrqqF7rZs6c2Wt5kjzwwAO9ln/hC1/odcnjZptt1usmsgsW9P5I1vnnn9/9Wt/FLdyodnEL37SzuDe84Q3db7sBAAD+sQlHunwhybdLKdOSPJfkoGUdXGud09rQ9a4kP0nXfhorpNb6dCnlvHTtyfFYuvZAWWG11t+UUi5Icmur6Pxa651JUko5uVX+SJJ7V6C769P19pjOJKck2amUMjpdqz7uTvKTWuvzpZTLWvN9MMmdf8l8k7wjyVdKKTVdYc5pCx9xAgAAgFeD13w4UmudmWRoj+9jl1L3oV7afmGx7z37+fhih3f0qDtyOXP6lyT/0kt5e4/PHYv12bPuq0m+2kv7byT5xnL6fTKtPUdqrU8leWePQ7+3lPl+LsnneinfrMfnC9K1IevidT9PMry3fgEAAODVwGtVAQAAgEZ7za8cWZVKKVclectixcfXWn+6KuYDAAAALEk4shLVWj+8qucAAAAALJvHagAAAIBGE44AAAAAjSYcAQAAABpNOAIAAAA0mnAEAAAAaDRvq+FVaa3V++a+iXut6mkAAADQAFaOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOOAAAAAI0mHAEAAAAaTTgCAAAANJpwBAAAAGg04QgAAADQaMIRAAAAoNGEIwAAAECjCUcAAACARhOO8P+3d+8x39Z1HcDf7wDRhATSMYZOjNEcc0VoHtIcaeHhD9FNG+UUnZV5aNlmk9pKM9c01mHm0jwQWBbgqcilyMysbKioIAczUXHKSJYIeGgW8O2P3/dhP2+f+4Hn4Xm47+e5Xq/t2u97f6/T9/pdn/t7/e7P/b2uHwAAACya5AgAAACwaJIjAAAAwKJJjgAAAACLJjkCAAAALJrkCAAAALBokiMAAADAokmOAAAAAIsmOQIAAAAsmuQIAAAAsGiSIwAAAMCiSY4AAAAAiyY5AgAAACya5AgAAACwaJIjAAAAwKJJjgAAAACLJjkCAAAALJrkCAAAALBokiMAAADAokmOAAAAAIsmOQIAAAAsmuQIAAAAsGiSIwAAAMCiSY4AAAAAiyY5AgAAACya5AgAAACwaJIjAAAAwKJJjgAAAACLJjkCAAAALJrkCAAAALBokiMAAADAokmOAAAAAIsmOQIAAAAsmuQIAAAAsGiSIwAAAMCiSY4AAAAAiyY5AgAAACya5AgAAACwaJIjAAAAwKJJjgAAAACLJjkCAAAALJrkCAAAALBokiMAAADAokmOAAAAAIsmOQIAAAAsmuQIAAAAsGiSIwAAAMCiSY4AAAAAiyY5AgAAACya5AgAAACwaJIjAAAAwKJJjgAAAACLJjkCAAAALJrkCAAAALBokiMAAADAokmOAAAAAIsmOQIAAAAsmuQIAAAAsGgdY2x1G+D7tP1mks9tdTvYVu6f5L+3uhFsK2KCjcQEG4kJ1okHNhITW+vBY4wHbHUjdjh4qxsAm/jcGOMRW90Ito+2l4oJ1okJNhITbCQmWCce2EhMsM5tNQAAAMCiSY4AAAAAiyY5wnb15q1uANuOmGAjMcFGYoKNxATrxAMbiQnu4IGsAAAAwKIZOQIAAAAsmuQI20rbJ7f9XNtr2p651e1h32p7bdsr2l7W9tJZd1Tbi9t+fr4eOevb9vUzNj7T9uS17Zwxl/982zO26njYfW3PbntD2yvX6vZaDLR9+Iyxa+a6vWePkN21SUy8qu11s6+4rO1T1+b91jy/n2v7pLX6nV5P2j6k7cdm/flt73XPHR17ou2D2n647dVtr2r767NeX7FAu4gH/cSCtb1324+3vXzGxe/N+p2ey7aHzp+vmfOPW9vWbsULB5Axhsm0LaYkByX5QpIfSXKvJJcnOXGr22Xap+f82iT331D3h0nOnOUzk7xulp+a5P1JmuTRST42649K8sX5euQsH7nVx2a6yzHw+CQnJ7lyX8RAko/PZTvXfcpWH7Npj2LiVUlevpNlT5zXikOTPGReQw7a1fUkyQVJTp/lNyV50VYfs+lOY+KYJCfP8uFJ/nOee33FAqddxIN+YsHT/N09bJYPSfKx+Tu903OZ5MVJ3jTLpyc5f0/jxXTgTEaOsJ08Msk1Y4wvjjH+N8l5SU7b4jZxzzstybmzfG6Sp6/Vv32sXJLkiLbHJHlSkovHGDeOMb6R5OIkT76nG82eGWP8S5IbN1TvlRiY835ojHHJGGMkefvattimNomJzZyW5LwxxnfHGF9Kck1W15KdXk/maIAnJHnXXH89vtimxhjXjzE+NcvfTPLZJMdGX7FIu4iHzegnFmD+vn9r/njInEY2P5fr/ce7kjxxnvvdipd9fFjcwyRH2E6OTfKVtZ+/ml1f7Nj/jSQfbPvJtr8y644eY1w/y/+V5OhZ3iw+xM2BZ2/FwLGzvLGe/dNL5y0SZ++4fSK7HxM/nOSmMcatG+rZT8yh7z+R1X+F9RULtyEeEv3EorU9qO1lSW7IKvn5hWx+Lu84/3P+zVmde583F0xyBNhKjxtjnJzkKUle0vbx6zPnf/B8pdaCiQGmNyY5PslJSa5P8kdb2xy2QtvDkrw7ycvGGLesz9NXLM9O4kE/sXBjjNvGGCcleWBWIz0eusVNYj8jOcJ2cl2SB639/MBZxwFqjHHdfL0hyXuzupB9bQ5xzny9YS6+WXyImwPP3oqB62Z5Yz37mTHG1+aH3tuTvCWrviLZ/Zj4ela3WBy8oZ5tru0hWf0h/I4xxntmtb5ioXYWD/oJdhhj3JTkw0kek83P5R3nf86/X1bn3ufNBZMcYTv5RJIT5lOl75XVw5Eu3OI2sY+0vW/bw3eUk5ya5MqszvmObxA4I8nfz/KFSZ47v4Xg0UlunsOpL0pyatsj5xDaU2cd+6+9EgNz3i1tHz3vI37u2rbYj+z4A3h6RlZ9RbKKidPntw48JMkJWT1Yc6fXkzm64MNJnjnXX48vtqn5+/u2JJ8dY/zx2ix9xQJtFg/6iWVr+4C2R8zyfZL8XFbPo9nsXK73H89M8k/z3O9WvOz7I+OedPCdLwL3jDHGrW1fmtWHl4OSnD3GuGqLm8W+c3SS964+4+TgJH8zxvhA208kuaDtC5J8OcnPz+X/MatvILgmyXeSPD9Jxhg3tv39rC5aSfLqMcZdfZgjW6zt3yY5Jcn92341ySuTvDZ7LwZenOScJPfJ6hso3r+PD4m7aZOYOKXtSVndNnFtkhcmyRjjqrYXJLk6ya1JXjLGuG1uZ7PrySuSnNf2NUk+ndUfWWxvj03ynCRXzOcJJMlvR1+xVJvFwy/oJxbtmCTntj0oqwEAF4wx3tf26uz8XL4tyV+1vSarh4CfnuxxvHCA6CpBBgAAALBMbqsBAAAAFk1yBAAAAFg0yREAAABg0SRHAAAAgEWTHAEAAAAWTXIEANjvtb2t7WVr03F7sI0j2r5477fuju0/re2Z+2r7m+zz6W1PvCf3CQD7I1/lCwDs99p+a4xx2N3cxnFJ3jfGeNhurnfQGOO2u7PvfaHtwUnemtUxvWur2wMA25mRIwDAAantQW3PavuJtp9p+8JZf1jbD7X9VNsr2p42V3ltkuPnyJOz2p7S9n1r23tD2+fN8rVtX9f2U0me1fb4th9o+8m2/9r2oTtpz/PavmGWz2n7xraXtP3i3NfZbT/b9py1db7V9k/aXjXb/IBZf9Jc9zNt39v2yFn/z23/tO2lSV6R5GlJzprHdHzbX57vx+Vt3932B9fa8/q2/z7b88y1Nrxivk+Xt33trLvT4wWA/cnBW90AAIC94D5tL5vlL40xnpHkBUluHmP8ZNtDk3y07QeTfCXJM8YYt7S9f5JL2l6Y5MwkDxtjnJQkbU+5k31+fYxx8lz2Q0l+dYzx+baPSvLnSZ5wJ+sfmeQxWSUwLkzy2CS/lOQTbU8aY1yW5L5JLh1j/Ebb303yyiQvTfL2JL82xvhI21fP+pfN7d5rjPGI2a4TsjZypO1NY4y3zPJr5nv0Z3O9Y5I8LslDZ3ve1fYpSU5L8qgxxnfaHjWXffMeHC8AbFuSIwDAgeB/diQ11pya5MfWRkHcL8kJSb6a5A/aPj7J7UmOTXL0Huzz/GQ1EiXJTyV5Z9sd8w69C+v/wxhjtL0iydfGGFfM7V2V5Lgkl832nT+X/+sk72l7vyRHjDE+MuvPTfLOje3axMNmUuSIJIcluWht3t+NMW5PcnXbHe/Hzyb5yzHGd5JkjHHj3TheANi2JEcAgANVsxpdcdH3VK5ujXlAkoePMf6v7bVJ7r2T9W/N996CvHGZb8/XH0hy006SM3fmu/P19rXyjp83+4x2Vx4W9+1dzDsnydPHGJfP9+GUnbQnWb13m9nT4wWAbcszRwCAA9VFSV7U9pAkafujbe+b1QiSG2Zi5GeSPHgu/80kh6+t/+UkJ7Y9tO0RSZ64s52MMW5J8qW2z5r7adsf30vH8ANJdox8+cUk/zbGuDnJN9r+9Kx/TpKP7GzlfP8xHZ7k+vmePPsu7P/iJM9fezbJUfv4eAFgS0iOAAAHqrcmuTrJp9pemeQvshqR8Y4kj5i3szw3yX8kyRjj61k9l+TKtmeNMb6S5IIkV87XT+9iX89O8oK2lye5KqvndOwN307yyNn+JyR59aw/I6sHrX4myUlr9Rudl+Q323667fFJfifJx5J8NPO4d2WM8YGsnj9y6Xymy8vnrH11vACwJXyVLwDANtW98BXFAMCdM3IEAAAAWDQjRwAAAIBFM3IEAAAAWDTJEQAAAGDRJEcAAACARZMcAQAAABZNcgQAAABYNMkRAAAAYNH+H3S0nmYRbpdbAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 1152x1152 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = train(params, Dataset(X2, Y2), num_boost_round=770)\n",
    "plot_importance(model, figsize=(16,16), importance_type='gain', max_num_features=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/light/miniconda3/lib/python3.6/site-packages/ipykernel_launcher.py:150: UserWarning: `meta` is not specified, inferred from partial data. Please provide `meta` if the result is unexpected.\n",
      "  Before: .apply(func)\n",
      "  After:  .apply(func, meta={'x': 'f8', 'y': 'f8'}) for dataframe result\n",
      "  or:     .apply(func, meta=('x', 'f8'))            for series result\n",
      "/home/light/miniconda3/lib/python3.6/site-packages/ipykernel_launcher.py:2: FutureWarning: 'cl_id' is both an index level and a column label.\n",
      "Defaulting to column, but this will raise an ambiguity error in a future version\n",
      "  \n"
     ]
    }
   ],
   "source": [
    "# Preparting features for test dataset\n",
    "X_test = features(test).compute(get=get).sort_values('cl_id')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## A"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def ensemble_a(x, y):\n",
    "    params = {'bagging_fraction': 0.55, 'bagging_freq': 2, 'colsample_bytree': 0.8, \n",
    "          'learning_rate': 0.01, 'max_bin': 255, 'max_depth': -1, 'metric': 'auc', \n",
    "          'min_data_in_leaf': 20, 'num_leaves': 16, 'objective': 'binary', 'reg_alpha': 0.4,\n",
    "          'reg_lambda': 0.55, 'verbose': -1}\n",
    "\n",
    "    model = train(params, Dataset(x, y), num_boost_round=770)\n",
    "    k_model = CatBoostClassifier(eval_metric='AUC', iterations=750, logging_level='Silent', max_depth=7,\n",
    "                                 learning_rate=0.02, random_seed=np.random.choice(10000), l2_leaf_reg=32\n",
    "                                )\n",
    "    k_model.fit(x, y)\n",
    "\n",
    "    xg_model = XGBClassifier(n_estimators=770, learning_rate=0.01, subsample=0.55, colsample_bytree=0.8,\n",
    "                      max_depth=7, min_child_weight=1, reg_alpha=0.4, reg_lambda=0.55\n",
    "                     )\n",
    "\n",
    "    xg_model.fit(x, y)\n",
    "\n",
    "    a_pred = 0.4*model.predict(X_test)+0.4*xg_model.predict_proba(X_test)[:,1] + 0.2*k_model.predict_proba(X_test)[:,1]\n",
    "    \n",
    "    return a_pred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "a_pred = ensemble_a(X2, Y2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Pseudo labeling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "valid_idx = np.where((a_pred>=0.95)|(a_pred<0.05))[0]\n",
    "X3 = pd.concat([X2, X_test.iloc[valid_idx]])\n",
    "Y3 = np.array(Y2.tolist()+ (a_pred[valid_idx]>=0.9).tolist())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(6085, 127)"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X3.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "a_pred = ensemble_a(X3, Y3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "pd.DataFrame({'_ID_': X_test.index, '_VAL_': a_pred}).to_csv('submission_a.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.94384193, 0.8212521 , 0.70903157, 0.76372934, 0.97636015])"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a_pred[:5]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## B"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3.8265194013908603\n",
      "654\n"
     ]
    }
   ],
   "source": [
    "params = {'bagging_fraction': 0.55, 'bagging_freq': 2, 'colsample_bytree': 0.5, \n",
    "          'learning_rate': 0.01, 'max_bin': 255, 'max_depth': -1, 'metric': 'rmse', \n",
    "          'min_data_in_leaf': 20, 'num_leaves': 16, 'objective': 'rmse', \n",
    "          'reg_alpha': 0.4,\n",
    "          'reg_lambda': 0.55, 'verbose': -1\n",
    "         }\n",
    "\n",
    "_ = cv(params, Dataset(X2, np.log(1+Y2_reg)), verbose_eval=False,\n",
    "       seed=200, num_boost_round=2000, early_stopping_rounds=100, stratified=False, nfold=10\n",
    "      )\n",
    "print(min(_['rmse-mean']))\n",
    "print(np.array(_['rmse-mean']).argmin())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f7cf4f27b38>"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x1152 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "params = {'bagging_fraction': 0.55, 'bagging_freq': 2, 'colsample_bytree': 0.8, \n",
    "          'learning_rate': 0.01, 'max_bin': 255, 'max_depth': -1, 'metric': 'rmse', \n",
    "          'min_data_in_leaf': 20, 'num_leaves': 16, 'objective': 'rmse', \n",
    "          'reg_alpha': 0.4,\n",
    "          'reg_lambda': 0.55, 'verbose': -1\n",
    "         }\n",
    "model = train(params, Dataset(X2, np.log(1+Y2_reg)), num_boost_round=770)\n",
    "plot_importance(model, figsize=(16,16), importance_type='gain', max_num_features=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "from xgboost import XGBRegressor\n",
    "\n",
    "params = {'bagging_fraction': 0.55, 'bagging_freq': 2, 'colsample_bytree': 0.5, \n",
    "          'learning_rate': 0.01, 'max_bin': 255, 'max_depth': -1, 'metric': 'rmse', \n",
    "          'min_data_in_leaf': 20, 'num_leaves': 16, 'objective': 'rmse', 'reg_alpha': 0.4,\n",
    "          'reg_lambda': 0.55, 'verbose': -1}\n",
    "\n",
    "model = train(params, Dataset(X2, np.log(1+Y2_reg)), num_boost_round=770)\n",
    "model_pred = np.exp(model.predict(X_test))-1\n",
    "\n",
    "xgb = XGBRegressor(n_estimators=770, learning_rate=0.01, subsample=0.55, colsample_bytree=0.5,\n",
    "                  max_depth=5, min_child_weight=1, reg_alpha=0.4, reg_lambda=0.75)\n",
    "xgb.fit(X2, np.log(1+Y2_reg))\n",
    "xgb_pred = np.exp(xgb.predict(X_test))-1\n",
    "\n",
    "k = CatBoostRegressor(eval_metric='AUC', iterations=800, \n",
    "                      logging_level='Silent', max_depth=8, learning_rate=0.02,\n",
    "                      l2_leaf_reg=12\n",
    "                      )\n",
    "\n",
    "k.fit(X2, np.log(1+Y2_reg))\n",
    "k_pred = np.exp(k.predict(X_test))-1\n",
    "\n",
    "b_pred = 0.4*model_pred+0.4*xgb_pred+0.2*k_pred\n",
    "\n",
    "b_pred=np.clip(b_pred, 0, b_pred.max())\n",
    "pd.DataFrame({'_ID_': X_test.index, '_VAL_': b_pred}).to_csv('submission_b.csv', index=False)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}