BRAIN-SMALL.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The autoreload extension is already loaded. To reload it, use:\n",
      "  %reload_ext autoreload\n"
     ]
    }
   ],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "from sklearn.model_selection import train_test_split\n",
    "import scipy.sparse\n",
    "import scipy.io\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "from IPython.display import clear_output\n",
    "\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.decomposition import FactorAnalysis\n",
    "\n",
    "from benchmarking import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_path = \"/home/ubuntu/single-cell-scVI/data/neuron_9k/\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(9128, 27998)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gene_names = pd.read_csv(data_path + \"filtered_gene_bc_matrices/mm10/genes.tsv\", index_col=0, header=None,sep=\"\\t\")\n",
    "barcodes = pd.read_csv(data_path + \"filtered_gene_bc_matrices/mm10/barcodes.tsv\", index_col=0, header=None,sep=\"\\t\")\n",
    "expression = pd.DataFrame(scipy.io.mmread(data_path + \"filtered_gene_bc_matrices/mm10/matrix.mtx\").T.A, \\\n",
    "                          columns=gene_names.index, index=barcodes.index)\n",
    "expression.shape "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "selected = np.std(expression, axis=0).argsort()[-3000:][::-1]\n",
    "expression = expression[gene_names.index[selected]]\n",
    "gene_names = expression.columns.astype(str)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>ENSMUSG00000052305</th>\n",
       "      <th>ENSMUSG00000069919</th>\n",
       "      <th>ENSMUSG00000069917</th>\n",
       "      <th>ENSMUSG00000073940</th>\n",
       "      <th>ENSMUSG00000072235</th>\n",
       "      <th>ENSMUSG00000092341</th>\n",
       "      <th>ENSMUSG00000019874</th>\n",
       "      <th>ENSMUSG00000049775</th>\n",
       "      <th>ENSMUSG00000064357</th>\n",
       "      <th>ENSMUSG00000029580</th>\n",
       "      <th>...</th>\n",
       "      <th>ENSMUSG00000025137</th>\n",
       "      <th>ENSMUSG00000055204</th>\n",
       "      <th>ENSMUSG00000041112</th>\n",
       "      <th>ENSMUSG00000020801</th>\n",
       "      <th>ENSMUSG00000070372</th>\n",
       "      <th>ENSMUSG00000027879</th>\n",
       "      <th>ENSMUSG00000031917</th>\n",
       "      <th>ENSMUSG00000027782</th>\n",
       "      <th>ENSMUSG00000036707</th>\n",
       "      <th>ENSMUSG00000018669</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>AAACCTGAGACGCTTT-1</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>107</td>\n",
       "      <td>2</td>\n",
       "      <td>24</td>\n",
       "      <td>52</td>\n",
       "      <td>6</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>AAACCTGAGATCTGCT-1</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>131</td>\n",
       "      <td>176</td>\n",
       "      <td>0</td>\n",
       "      <td>149</td>\n",
       "      <td>61</td>\n",
       "      <td>51</td>\n",
       "      <td>...</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>AAACCTGAGGTCATCT-1</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>49</td>\n",
       "      <td>113</td>\n",
       "      <td>1</td>\n",
       "      <td>27</td>\n",
       "      <td>19</td>\n",
       "      <td>7</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>AAACCTGCAAGAGTCG-1</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>187</td>\n",
       "      <td>52</td>\n",
       "      <td>3</td>\n",
       "      <td>148</td>\n",
       "      <td>87</td>\n",
       "      <td>27</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>AAACCTGCACACATGT-1</th>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>32</td>\n",
       "      <td>111</td>\n",
       "      <td>133</td>\n",
       "      <td>226</td>\n",
       "      <td>151</td>\n",
       "      <td>46</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 3000 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "0                   ENSMUSG00000052305  ENSMUSG00000069919  \\\n",
       "0                                                            \n",
       "AAACCTGAGACGCTTT-1                   0                   1   \n",
       "AAACCTGAGATCTGCT-1                   0                   0   \n",
       "AAACCTGAGGTCATCT-1                   1                   0   \n",
       "AAACCTGCAAGAGTCG-1                   0                   0   \n",
       "AAACCTGCACACATGT-1                   2                   0   \n",
       "\n",
       "0                   ENSMUSG00000069917  ENSMUSG00000073940  \\\n",
       "0                                                            \n",
       "AAACCTGAGACGCTTT-1                   0                   0   \n",
       "AAACCTGAGATCTGCT-1                   0                   0   \n",
       "AAACCTGAGGTCATCT-1                   0                   0   \n",
       "AAACCTGCAAGAGTCG-1                   0                   1   \n",
       "AAACCTGCACACATGT-1                   1                   0   \n",
       "\n",
       "0                   ENSMUSG00000072235  ENSMUSG00000092341  \\\n",
       "0                                                            \n",
       "AAACCTGAGACGCTTT-1                   1                 107   \n",
       "AAACCTGAGATCTGCT-1                 131                 176   \n",
       "AAACCTGAGGTCATCT-1                  49                 113   \n",
       "AAACCTGCAAGAGTCG-1                 187                  52   \n",
       "AAACCTGCACACATGT-1                  32                 111   \n",
       "\n",
       "0                   ENSMUSG00000019874  ENSMUSG00000049775  \\\n",
       "0                                                            \n",
       "AAACCTGAGACGCTTT-1                   2                  24   \n",
       "AAACCTGAGATCTGCT-1                   0                 149   \n",
       "AAACCTGAGGTCATCT-1                   1                  27   \n",
       "AAACCTGCAAGAGTCG-1                   3                 148   \n",
       "AAACCTGCACACATGT-1                 133                 226   \n",
       "\n",
       "0                   ENSMUSG00000064357  ENSMUSG00000029580  \\\n",
       "0                                                            \n",
       "AAACCTGAGACGCTTT-1                  52                   6   \n",
       "AAACCTGAGATCTGCT-1                  61                  51   \n",
       "AAACCTGAGGTCATCT-1                  19                   7   \n",
       "AAACCTGCAAGAGTCG-1                  87                  27   \n",
       "AAACCTGCACACATGT-1                 151                  46   \n",
       "\n",
       "0                          ...          ENSMUSG00000025137  \\\n",
       "0                          ...                               \n",
       "AAACCTGAGACGCTTT-1         ...                           0   \n",
       "AAACCTGAGATCTGCT-1         ...                           3   \n",
       "AAACCTGAGGTCATCT-1         ...                           0   \n",
       "AAACCTGCAAGAGTCG-1         ...                           0   \n",
       "AAACCTGCACACATGT-1         ...                           0   \n",
       "\n",
       "0                   ENSMUSG00000055204  ENSMUSG00000041112  \\\n",
       "0                                                            \n",
       "AAACCTGAGACGCTTT-1                   1                   0   \n",
       "AAACCTGAGATCTGCT-1                   1                   0   \n",
       "AAACCTGAGGTCATCT-1                   0                   0   \n",
       "AAACCTGCAAGAGTCG-1                   0                   0   \n",
       "AAACCTGCACACATGT-1                   4                   0   \n",
       "\n",
       "0                   ENSMUSG00000020801  ENSMUSG00000070372  \\\n",
       "0                                                            \n",
       "AAACCTGAGACGCTTT-1                   1                   1   \n",
       "AAACCTGAGATCTGCT-1                   0                   2   \n",
       "AAACCTGAGGTCATCT-1                   0                   0   \n",
       "AAACCTGCAAGAGTCG-1                   1                   0   \n",
       "AAACCTGCACACATGT-1                   1                   3   \n",
       "\n",
       "0                   ENSMUSG00000027879  ENSMUSG00000031917  \\\n",
       "0                                                            \n",
       "AAACCTGAGACGCTTT-1                   0                   0   \n",
       "AAACCTGAGATCTGCT-1                   0                   0   \n",
       "AAACCTGAGGTCATCT-1                   0                   0   \n",
       "AAACCTGCAAGAGTCG-1                   0                   0   \n",
       "AAACCTGCACACATGT-1                   2                   1   \n",
       "\n",
       "0                   ENSMUSG00000027782  ENSMUSG00000036707  ENSMUSG00000018669  \n",
       "0                                                                               \n",
       "AAACCTGAGACGCTTT-1                   0                   0                   0  \n",
       "AAACCTGAGATCTGCT-1                   1                   0                   0  \n",
       "AAACCTGAGGTCATCT-1                   0                   0                   0  \n",
       "AAACCTGCAAGAGTCG-1                   0                   0                   0  \n",
       "AAACCTGCACACATGT-1                   1                   0                   0  \n",
       "\n",
       "[5 rows x 3000 columns]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "expression.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## now clusters and qc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "raw_qc = pd.read_csv(data_path + \"molecule_qc.txt\", index_col=0, sep=\"\\t\")\n",
    "cell_qc = raw_qc.columns\n",
    "raw_qc = raw_qc.loc[expression.index].as_matrix()\n",
    "#clusters = pd.read_csv(data_path + \"analysis/clustering/graphclust/clusters.csv\", index_col=0)\n",
    "clusters = pd.read_csv(data_path + \"analysis/clustering/kmeans_10_clusters/clusters.csv\", index_col=0)\n",
    "clusters = clusters.loc[expression.index].as_matrix()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 1,  2,  3,  4,  5,  6,  7,  8,  9, 10])"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.unique(clusters)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "expression_train, expression_test, \\\n",
    "c_train, c_test, \\\n",
    "qc_train, qc_test = \\\n",
    "                train_test_split(expression.as_matrix(), clusters, raw_qc, random_state=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {
    "collapsed": true,
    "deletable": false,
    "editable": false,
    "run_control": {
     "frozen": true
    }
   },
   "outputs": [],
   "source": [
    "scipy.sparse.save_npz(data_path + \"data_train.npz\", scipy.sparse.coo_matrix(expression_train))\n",
    "scipy.sparse.save_npz(data_path + \"data_test.npz\", scipy.sparse.coo_matrix(expression_test))\n",
    "\n",
    "\n",
    "np.savetxt(data_path + \"label_train\", c_train)\n",
    "np.savetxt(data_path + \"label_test\", c_test)\n",
    "\n",
    "np.save(data_path + \"qc_train\", qc_train)\n",
    "np.save(data_path + \"qc_test\", qc_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# train scVI"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from model import vae\n",
    "from model import scVI_final as scVI\n",
    "from model.vae import format_time\n",
    "import tensorflow as tf\n",
    "import scipy.stats as stats\n",
    "import time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "batch_size = 128\n",
    "learning_rate = 0.0004\n",
    "epsilon = 0.01"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "log_library_size = np.log(np.sum(expression_train, axis=1))\n",
    "mean, var = np.mean(log_library_size), np.var(log_library_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def next_batch(data, batch_size):\n",
    "    index = np.random.choice(np.arange(data.shape[0]), size=batch_size)\n",
    "    return data[index]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def train_model(num_epochs, expression_train, expression_test, step, kl=1):\n",
    "    iterep = int(expression_train.shape[0]/float(batch_size))-1\n",
    "    training_history = {\"t_loss\":[], \"v_loss\":[], \"time\":[], \"epoch\":[]}\n",
    "    training_history[\"n_hidden\"] = model.n_hidden\n",
    "    training_history[\"model\"] = model.__class__.__name__\n",
    "    training_history[\"n_input\"] = model.n_input\n",
    "    training_history[\"dropout_nn\"] = model.dropout_rate\n",
    "    training_history[\"learning_rate\"] = learning_rate\n",
    "    training_history[\"dispersion\"] = model.dispersion\n",
    "    training_history[\"epsilon\"] = epsilon\n",
    "    training_history[\"n_layers\"] = model.n_layers\n",
    "    begin = time.time()\n",
    "    \n",
    "    \n",
    "    for t in range(iterep * num_epochs):\n",
    "    \n",
    "        # arange data in batches\n",
    "        x_train = next_batch(expression_train, batch_size)\n",
    "        x_test = next_batch(expression_test, batch_size)\n",
    "\n",
    "        #prepare data dictionaries\n",
    "        dic_train = {expression: x_train, training_phase:True, kl_scalar:kl}\n",
    "        dic_test = {expression: x_test, training_phase:False, kl_scalar:kl} \n",
    "\n",
    "        # run an optimization set\n",
    "        _, l_tr = sess.run([step, model.loss], feed_dict=dic_train)\n",
    "        end_epoch, epoch = t % iterep == 0, t / iterep\n",
    "\n",
    "        if end_epoch:\n",
    "            now = time.time()\n",
    "            print epoch\n",
    "\n",
    "            l_t = sess.run((model.loss), feed_dict=dic_test)\n",
    "\n",
    "            training_history[\"t_loss\"].append(l_tr)\n",
    "            training_history[\"v_loss\"].append(l_t)\n",
    "            training_history[\"time\"].append(format_time(int(now-begin)))\n",
    "            training_history[\"epoch\"].append(epoch)\n",
    "            \n",
    "            print 'Train / Test performance:', l_tr, l_t\n",
    "            print \"Elapsed Time:\", training_history[\"time\"][-1]\n",
    "            if np.isnan(l_tr):\n",
    "                break\n",
    "    return training_history"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def sample_posterior(X):\n",
    "    # shape and simulation\n",
    "    results = {}\n",
    "\n",
    "    dic_x = {expression: X, training_phase:False, kl_scalar:1.} \n",
    "    z_m, l_m = sess.run((model.z, model.library), feed_dict=dic_x)\n",
    "    dic_z = {model.z: z_m, model.library:l_m, training_phase:False, kl_scalar:1.}\n",
    "    rate, dropout = sess.run((model.px_rate, model.px_dropout), feed_dict=dic_z)\n",
    "    \n",
    "    dispersion = np.tile(sess.run((tf.exp(model.px_r))), (rate.shape[0], 1))\n",
    "    \n",
    "    results[\"library\"] = l_m\n",
    "    results[\"mean\"] = rate\n",
    "    results[\"latent\"] = z_m\n",
    "    results[\"dispersion\"] = dispersion\n",
    "    results[\"dropout\"] = dropout\n",
    "    return results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def process_proba(logit_dropout, mean=None, dispersion=None, logits=False, agg=None):\n",
    "    \n",
    "    p = 1 / (1 + np.exp(- logit_dropout))\n",
    "    if mean is not None:\n",
    "        p = (1-p) / scipy.special.gamma(dispersion) * np.power(dispersion / (dispersion + mean), dispersion) \n",
    "        \n",
    "    p[p > 1] = np.max(p[p < 1])\n",
    "    p[p == 0] = np.min(p[p > 0])\n",
    "\n",
    "    return p"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def normalize_plot(x, y, z, s=\"\", logit=True, normalize=True, plot=True):\n",
    "    \"\"\"\n",
    "    plot to verify that x and y is correlated with z\n",
    "    \"\"\"\n",
    "    \n",
    "    if logit:\n",
    "        # take logits before averaging\n",
    "        x = np.log(x / (1-x))\n",
    "        y = np.log(y / (1-y))\n",
    "        \n",
    "    # test shapes\n",
    "    if z.shape[0] == x.shape[0]:\n",
    "        #average across genes\n",
    "        x = np.mean(x, axis=1)\n",
    "        y = np.mean(y, axis=1)\n",
    "    \n",
    "    elif z.shape[0] == x.shape[1]:\n",
    "        # we got a gene problem, ave ac. cells\n",
    "        x = np.mean(x, axis=0)\n",
    "        y = np.mean(y, axis=0)\n",
    "    \n",
    "    else: \n",
    "        print \"Wrong shapes !\"\n",
    "    \n",
    "    if normalize:\n",
    "        # normalization workflow\n",
    "        x = x - np.mean(x)\n",
    "        x /= np.std(x)\n",
    "        y = y - np.mean(y)\n",
    "        y /= np.std(y)\n",
    "        z = z - np.mean(z)\n",
    "        z /= np.std(z)\n",
    "    \n",
    "    if plot:\n",
    "        # scattering workflow\n",
    "        f, axarr = plt.subplots(nrows=1, ncols=2, sharey=True, figsize=(10, 5))\n",
    "        axarr[0].scatter(x, z, alpha=0.5)\n",
    "        axarr[1].scatter(y, z, alpha=0.5)\n",
    "    \n",
    "    if normalize:\n",
    "        if plot:\n",
    "            axarr[0].axis((-3,3,-3,3))\n",
    "            axarr[1].axis((-3,3,-3,3))   \n",
    "        # regression workflow\n",
    "        a_x, intercept, r_value_x, p_value_x, std_err = stats.linregress(x,z)\n",
    "        \n",
    "        if plot:\n",
    "            l = np.linspace(-3, 3)\n",
    "            axarr[0].plot(l, a_x * l, color='black')\n",
    "            supp = \"a={0:.2f} \\np={1:.1e}\".format(a_x, p_value_x)\n",
    "            if a_x >0.1:\n",
    "                axarr[0].text(1, -2.2, supp, fontsize=12)\n",
    "            else:\n",
    "                axarr[0].text(1, 2.2, supp, fontsize=12) \n",
    "            \n",
    "            \n",
    "        a_y, intercept, r_value_y, p_value_y, std_err = stats.linregress(y,z)\n",
    "        \n",
    "        if plot:\n",
    "            supp = \"a={0:.2f} \\np={1:.1e}\".format(a_y, p_value_y)\n",
    "            l = np.linspace(-3, 3)\n",
    "            axarr[1].plot(l, a_y * l, color='black')\n",
    "            if a_y >0.1:\n",
    "                axarr[1].text(1, -2.2, supp, fontsize=12)\n",
    "            else:\n",
    "                axarr[1].text(1, 2.2, supp, fontsize=12) \n",
    "      \n",
    "    if plot:\n",
    "        axarr[0].set_xlabel(\"Bernouilli zero prob\")\n",
    "        axarr[1].set_xlabel(\"NB zero prob\")\n",
    "        axarr[0].set_ylabel(s)\n",
    "\n",
    "    return a_x, r_value_x, a_y, r_value_y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "list_plots = [\"emp_gene_detection\", \"emp_lib_size\", \\\n",
    "              cell_qc[2], cell_qc[4], cell_qc[5]]\n",
    "trials = 5\n",
    "list_cell_types = np.unique(clusters)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running scVI on 3000 genes\n",
      "scVI will run without batch correction\n",
      "Will work on mode numeric for incorporating library size\n",
      "Will work on mode gene for modeling inverse dispersion param\n",
      "Will apply zero inflation\n",
      "1 hidden layers at 128 each for a final 5 latent space\n",
      "0\n",
      "Train / Test performance: 47658.3 42731.8\n",
      "Elapsed Time: 0:00:00\n",
      "1\n",
      "Train / Test performance: 22224.6 22608.8\n",
      "Elapsed Time: 0:00:02\n",
      "2\n",
      "Train / Test performance: 10041.6 17999.2\n",
      "Elapsed Time: 0:00:05\n",
      "3\n",
      "Train / Test performance: 5349.95 13259.5\n",
      "Elapsed Time: 0:00:07\n",
      "4\n",
      "Train / Test performance: 4783.15 12040.1\n",
      "Elapsed Time: 0:00:10\n",
      "5\n",
      "Train / Test performance: 4874.37 7149.76\n",
      "Elapsed Time: 0:00:12\n",
      "6\n",
      "Train / Test performance: 4819.45 5763.26\n",
      "Elapsed Time: 0:00:15\n",
      "7\n",
      "Train / Test performance: 4677.26 5526.7\n",
      "Elapsed Time: 0:00:17\n",
      "8\n",
      "Train / Test performance: 4453.1 4905.2\n",
      "Elapsed Time: 0:00:19\n",
      "9\n",
      "Train / Test performance: 4593.36 5452.77\n",
      "Elapsed Time: 0:00:22\n",
      "10\n",
      "Train / Test performance: 4590.08 5050.8\n",
      "Elapsed Time: 0:00:25\n",
      "11\n",
      "Train / Test performance: 4379.72 5081.59\n",
      "Elapsed Time: 0:00:29\n",
      "12\n",
      "Train / Test performance: 4290.5 4595.81\n",
      "Elapsed Time: 0:00:31\n",
      "13\n",
      "Train / Test performance: 4432.26 4766.34\n",
      "Elapsed Time: 0:00:35\n",
      "14\n",
      "Train / Test performance: 4355.15 4531.39\n",
      "Elapsed Time: 0:00:38\n",
      "15\n",
      "Train / Test performance: 4289.28 4637.52\n",
      "Elapsed Time: 0:00:41\n",
      "16\n",
      "Train / Test performance: 4326.91 4457.66\n",
      "Elapsed Time: 0:00:43\n",
      "17\n",
      "Train / Test performance: 4212.07 4913.34\n",
      "Elapsed Time: 0:00:46\n",
      "18\n",
      "Train / Test performance: 4207.3 4851.56\n",
      "Elapsed Time: 0:00:49\n",
      "19\n",
      "Train / Test performance: 4310.54 4976.66\n",
      "Elapsed Time: 0:00:52\n",
      "20\n",
      "Train / Test performance: 4375.47 4668.43\n",
      "Elapsed Time: 0:00:54\n",
      "21\n",
      "Train / Test performance: 4157.88 4223.96\n",
      "Elapsed Time: 0:00:57\n",
      "22\n",
      "Train / Test performance: 4162.52 4397.13\n",
      "Elapsed Time: 0:01:00\n",
      "23\n",
      "Train / Test performance: 3971.27 4709.67\n",
      "Elapsed Time: 0:01:02\n",
      "24\n",
      "Train / Test performance: 4125.5 4520.08\n",
      "Elapsed Time: 0:01:05\n",
      "25\n",
      "Train / Test performance: 4235.04 4610.91\n",
      "Elapsed Time: 0:01:08\n",
      "26\n",
      "Train / Test performance: 4074.61 4536.87\n",
      "Elapsed Time: 0:01:11\n",
      "27\n",
      "Train / Test performance: 4195.74 4442.57\n",
      "Elapsed Time: 0:01:14\n",
      "28\n",
      "Train / Test performance: 4086.45 4240.29\n",
      "Elapsed Time: 0:01:16\n",
      "29\n",
      "Train / Test performance: 4035.25 4347.66\n",
      "Elapsed Time: 0:01:19\n",
      "30\n",
      "Train / Test performance: 4115.18 4487.82\n",
      "Elapsed Time: 0:01:23\n",
      "31\n",
      "Train / Test performance: 4327.61 4414.39\n",
      "Elapsed Time: 0:01:25\n",
      "32\n",
      "Train / Test performance: 4004.63 4609.01\n",
      "Elapsed Time: 0:01:28\n",
      "33\n",
      "Train / Test performance: 4052.09 5121.7\n",
      "Elapsed Time: 0:01:30\n",
      "34\n",
      "Train / Test performance: 3990.97 4200.83\n",
      "Elapsed Time: 0:01:33\n",
      "35\n",
      "Train / Test performance: 4132.84 4532.01\n",
      "Elapsed Time: 0:01:35\n",
      "36\n",
      "Train / Test performance: 4049.15 4327.2\n",
      "Elapsed Time: 0:01:37\n",
      "37\n",
      "Train / Test performance: 4103.27 4610.06\n",
      "Elapsed Time: 0:01:40\n",
      "38\n",
      "Train / Test performance: 4161.51 4178.24\n",
      "Elapsed Time: 0:01:42\n",
      "39\n",
      "Train / Test performance: 3832.36 3982.34\n",
      "Elapsed Time: 0:01:45\n",
      "40\n",
      "Train / Test performance: 4122.27 4387.9\n",
      "Elapsed Time: 0:01:48\n",
      "41\n",
      "Train / Test performance: 4032.0 4209.68\n",
      "Elapsed Time: 0:01:50\n",
      "42\n",
      "Train / Test performance: 3995.26 4577.88\n",
      "Elapsed Time: 0:01:52\n",
      "43\n",
      "Train / Test performance: 3952.98 4404.6\n",
      "Elapsed Time: 0:01:55\n",
      "44\n",
      "Train / Test performance: 4273.9 4464.58\n",
      "Elapsed Time: 0:01:57\n",
      "45\n",
      "Train / Test performance: 4123.69 4213.12\n",
      "Elapsed Time: 0:01:59\n",
      "46\n",
      "Train / Test performance: 3917.44 4528.68\n",
      "Elapsed Time: 0:02:02\n",
      "47\n",
      "Train / Test performance: 4072.64 4560.98\n",
      "Elapsed Time: 0:02:04\n",
      "48\n",
      "Train / Test performance: 3866.33 4119.55\n",
      "Elapsed Time: 0:02:06\n",
      "49\n",
      "Train / Test performance: 3910.5 4360.73\n",
      "Elapsed Time: 0:02:08\n",
      "50\n",
      "Train / Test performance: 4004.38 4611.31\n",
      "Elapsed Time: 0:02:11\n",
      "51\n",
      "Train / Test performance: 4149.78 4115.52\n",
      "Elapsed Time: 0:02:13\n",
      "52\n",
      "Train / Test performance: 3955.5 4502.08\n",
      "Elapsed Time: 0:02:15\n",
      "53\n",
      "Train / Test performance: 4132.22 5739.29\n",
      "Elapsed Time: 0:02:18\n",
      "54\n",
      "Train / Test performance: 3872.21 4244.97\n",
      "Elapsed Time: 0:02:20\n",
      "55\n",
      "Train / Test performance: 3906.21 4431.55\n",
      "Elapsed Time: 0:02:22\n",
      "56\n",
      "Train / Test performance: 3914.0 3931.05\n",
      "Elapsed Time: 0:02:24\n",
      "57\n",
      "Train / Test performance: 3876.98 4207.35\n",
      "Elapsed Time: 0:02:27\n",
      "58\n",
      "Train / Test performance: 3992.75 4514.63\n",
      "Elapsed Time: 0:02:29\n",
      "59\n",
      "Train / Test performance: 4055.51 4331.14\n",
      "Elapsed Time: 0:02:31\n",
      "60\n",
      "Train / Test performance: 4193.2 4510.15\n",
      "Elapsed Time: 0:02:33\n",
      "61\n",
      "Train / Test performance: 3924.22 4142.77\n",
      "Elapsed Time: 0:02:35\n",
      "62\n",
      "Train / Test performance: 3902.32 4174.23\n",
      "Elapsed Time: 0:02:39\n",
      "63\n",
      "Train / Test performance: 3862.54 4201.08\n",
      "Elapsed Time: 0:02:42\n",
      "64\n",
      "Train / Test performance: 4008.15 4289.16\n",
      "Elapsed Time: 0:02:44\n",
      "65\n",
      "Train / Test performance: 3940.31 4096.87\n",
      "Elapsed Time: 0:02:47\n",
      "66\n",
      "Train / Test performance: 4022.93 4047.92\n",
      "Elapsed Time: 0:02:50\n",
      "67\n",
      "Train / Test performance: 3962.72 4180.1\n",
      "Elapsed Time: 0:02:53\n",
      "68\n",
      "Train / Test performance: 3834.26 4118.0\n",
      "Elapsed Time: 0:02:55\n",
      "69\n",
      "Train / Test performance: 3787.23 4580.54\n",
      "Elapsed Time: 0:02:57\n",
      "70\n",
      "Train / Test performance: 3831.14 4553.69\n",
      "Elapsed Time: 0:03:00\n",
      "71\n",
      "Train / Test performance: 3790.77 4193.52\n",
      "Elapsed Time: 0:03:03\n",
      "72\n",
      "Train / Test performance: 4007.08 4139.97\n",
      "Elapsed Time: 0:03:05\n",
      "73\n",
      "Train / Test performance: 3998.67 4122.7\n",
      "Elapsed Time: 0:03:08\n",
      "74\n",
      "Train / Test performance: 3854.1 3988.43\n",
      "Elapsed Time: 0:03:10\n",
      "75\n",
      "Train / Test performance: 3912.73 4045.77\n",
      "Elapsed Time: 0:03:12\n",
      "76\n",
      "Train / Test performance: 3687.56 3893.39\n",
      "Elapsed Time: 0:03:15\n",
      "77\n",
      "Train / Test performance: 3776.66 4047.99\n",
      "Elapsed Time: 0:03:17\n",
      "78\n",
      "Train / Test performance: 3895.04 3944.95\n",
      "Elapsed Time: 0:03:19\n",
      "79\n",
      "Train / Test performance: 3909.95 3885.75\n",
      "Elapsed Time: 0:03:22\n",
      "80\n",
      "Train / Test performance: 3789.73 4004.01\n",
      "Elapsed Time: 0:03:24\n",
      "81\n",
      "Train / Test performance: 3877.53 4107.66\n",
      "Elapsed Time: 0:03:27\n",
      "82\n",
      "Train / Test performance: 3890.47 3985.24\n",
      "Elapsed Time: 0:03:29\n",
      "83\n",
      "Train / Test performance: 3856.03 4092.67\n",
      "Elapsed Time: 0:03:31\n",
      "84\n",
      "Train / Test performance: 3842.95 4014.3\n",
      "Elapsed Time: 0:03:34\n",
      "85\n",
      "Train / Test performance: 3825.82 4181.23\n",
      "Elapsed Time: 0:03:36\n",
      "86\n",
      "Train / Test performance: 3824.75 3808.64\n",
      "Elapsed Time: 0:03:38\n",
      "87\n",
      "Train / Test performance: 3894.9 6036.12\n",
      "Elapsed Time: 0:03:40\n",
      "88\n",
      "Train / Test performance: 3954.48 4008.57\n",
      "Elapsed Time: 0:03:43\n",
      "89\n",
      "Train / Test performance: 3913.97 4039.73\n",
      "Elapsed Time: 0:03:45\n",
      "90\n",
      "Train / Test performance: 3984.71 4126.32\n",
      "Elapsed Time: 0:03:47\n",
      "91\n",
      "Train / Test performance: 3764.78 4001.96\n",
      "Elapsed Time: 0:03:49\n",
      "92\n",
      "Train / Test performance: 3739.44 3708.85\n",
      "Elapsed Time: 0:03:52\n",
      "93\n",
      "Train / Test performance: 3803.05 3776.65\n",
      "Elapsed Time: 0:03:54\n",
      "94\n",
      "Train / Test performance: 3749.78 4032.27\n",
      "Elapsed Time: 0:03:56\n",
      "95\n",
      "Train / Test performance: 3793.94 3919.43\n",
      "Elapsed Time: 0:03:59\n",
      "96\n",
      "Train / Test performance: 3838.85 4444.4\n",
      "Elapsed Time: 0:04:01\n",
      "97\n",
      "Train / Test performance: 3883.05 4887.02\n",
      "Elapsed Time: 0:04:03\n",
      "98\n",
      "Train / Test performance: 3831.97 4181.27\n",
      "Elapsed Time: 0:04:05\n",
      "99\n",
      "Train / Test performance: 3808.53 4062.88\n",
      "Elapsed Time: 0:04:08\n",
      "Running scVI on 3000 genes\n",
      "scVI will run without batch correction\n",
      "Will work on mode numeric for incorporating library size\n",
      "Will work on mode gene for modeling inverse dispersion param\n",
      "Will apply zero inflation\n",
      "1 hidden layers at 128 each for a final 5 latent space\n",
      "0\n",
      "Train / Test performance: 53515.4 43839.2\n",
      "Elapsed Time: 0:00:00\n",
      "1\n",
      "Train / Test performance: 23714.8 21353.1\n",
      "Elapsed Time: 0:00:02\n",
      "2\n",
      "Train / Test performance: 10016.3 12863.3\n",
      "Elapsed Time: 0:00:04\n",
      "3\n",
      "Train / Test performance: 5266.58 7107.47\n",
      "Elapsed Time: 0:00:06\n",
      "4\n",
      "Train / Test performance: 5164.5 8220.02\n",
      "Elapsed Time: 0:00:09\n",
      "5\n",
      "Train / Test performance: 4924.85 7995.01\n",
      "Elapsed Time: 0:00:11\n",
      "6\n",
      "Train / Test performance: 4478.96 6540.66\n",
      "Elapsed Time: 0:00:13\n",
      "7\n",
      "Train / Test performance: 4740.57 4889.65\n",
      "Elapsed Time: 0:00:16\n",
      "8\n",
      "Train / Test performance: 4541.41 4819.76\n",
      "Elapsed Time: 0:00:18\n",
      "9\n",
      "Train / Test performance: 4422.01 5267.5\n",
      "Elapsed Time: 0:00:20\n",
      "10\n",
      "Train / Test performance: 4366.19 5065.85\n",
      "Elapsed Time: 0:00:22\n",
      "11\n",
      "Train / Test performance: 4697.75 5135.8\n",
      "Elapsed Time: 0:00:25\n",
      "12\n",
      "Train / Test performance: 4512.17 4666.67\n",
      "Elapsed Time: 0:00:27\n",
      "13\n",
      "Train / Test performance: 4425.62 4731.82\n",
      "Elapsed Time: 0:00:29\n",
      "14\n",
      "Train / Test performance: 4472.65 5146.4\n",
      "Elapsed Time: 0:00:31\n",
      "15\n",
      "Train / Test performance: 4247.46 4474.73\n",
      "Elapsed Time: 0:00:34\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "16\n",
      "Train / Test performance: 4208.64 4596.3\n",
      "Elapsed Time: 0:00:36\n",
      "17\n",
      "Train / Test performance: 4188.95 4495.0\n",
      "Elapsed Time: 0:00:38\n",
      "18\n",
      "Train / Test performance: 4229.58 4811.77\n",
      "Elapsed Time: 0:00:40\n",
      "19\n",
      "Train / Test performance: 4276.41 5268.07\n",
      "Elapsed Time: 0:00:43\n",
      "20\n",
      "Train / Test performance: 4271.07 4557.11\n",
      "Elapsed Time: 0:00:45\n",
      "21\n",
      "Train / Test performance: 4294.77 6426.33\n",
      "Elapsed Time: 0:00:47\n",
      "22\n",
      "Train / Test performance: 4362.69 4397.47\n",
      "Elapsed Time: 0:00:50\n",
      "23\n",
      "Train / Test performance: 4102.57 4449.22\n",
      "Elapsed Time: 0:00:52\n",
      "24\n",
      "Train / Test performance: 4146.11 4440.79\n",
      "Elapsed Time: 0:00:55\n",
      "25\n",
      "Train / Test performance: 4156.87 4742.9\n",
      "Elapsed Time: 0:00:57\n",
      "26\n",
      "Train / Test performance: 4081.43 4618.17\n",
      "Elapsed Time: 0:00:59\n",
      "27\n",
      "Train / Test performance: 4127.15 4434.51\n",
      "Elapsed Time: 0:01:02\n",
      "28\n",
      "Train / Test performance: 4317.05 4640.01\n",
      "Elapsed Time: 0:01:04\n",
      "29\n",
      "Train / Test performance: 4075.08 4183.5\n",
      "Elapsed Time: 0:01:06\n",
      "30\n",
      "Train / Test performance: 4076.21 4701.59\n",
      "Elapsed Time: 0:01:09\n",
      "31\n",
      "Train / Test performance: 4130.65 4491.48\n",
      "Elapsed Time: 0:01:12\n",
      "32\n",
      "Train / Test performance: 4135.63 4242.92\n",
      "Elapsed Time: 0:01:15\n",
      "33\n",
      "Train / Test performance: 4039.62 4149.82\n",
      "Elapsed Time: 0:01:18\n",
      "34\n",
      "Train / Test performance: 4234.75 4853.41\n",
      "Elapsed Time: 0:01:21\n",
      "35\n",
      "Train / Test performance: 4346.56 5238.06\n",
      "Elapsed Time: 0:01:25\n",
      "36\n",
      "Train / Test performance: 3930.41 4433.51\n",
      "Elapsed Time: 0:01:28\n",
      "37\n",
      "Train / Test performance: 4167.25 4235.0\n",
      "Elapsed Time: 0:01:32\n",
      "38\n",
      "Train / Test performance: 4105.17 4601.65\n",
      "Elapsed Time: 0:01:36\n",
      "39\n",
      "Train / Test performance: 4000.01 4313.14\n",
      "Elapsed Time: 0:01:39\n",
      "40\n",
      "Train / Test performance: 4197.22 4525.65\n",
      "Elapsed Time: 0:01:43\n",
      "41\n",
      "Train / Test performance: 4034.22 4108.23\n",
      "Elapsed Time: 0:01:46\n",
      "42\n",
      "Train / Test performance: 4110.81 5240.99\n",
      "Elapsed Time: 0:01:49\n",
      "43\n",
      "Train / Test performance: 3935.78 4969.74\n",
      "Elapsed Time: 0:01:51\n",
      "44\n",
      "Train / Test performance: 4010.0 4357.05\n",
      "Elapsed Time: 0:01:54\n",
      "45\n",
      "Train / Test performance: 3964.16 4213.77\n",
      "Elapsed Time: 0:01:58\n",
      "46\n",
      "Train / Test performance: 4106.6 4628.89\n",
      "Elapsed Time: 0:02:01\n",
      "47\n",
      "Train / Test performance: 4048.2 4306.36\n",
      "Elapsed Time: 0:02:05\n",
      "48\n",
      "Train / Test performance: 3984.11 4864.2\n",
      "Elapsed Time: 0:02:08\n",
      "49\n",
      "Train / Test performance: 3956.85 4206.29\n",
      "Elapsed Time: 0:02:11\n",
      "50\n",
      "Train / Test performance: 4024.33 4961.45\n",
      "Elapsed Time: 0:02:13\n",
      "51\n",
      "Train / Test performance: 4001.38 4267.47\n",
      "Elapsed Time: 0:02:16\n",
      "52\n",
      "Train / Test performance: 3951.68 4216.74\n",
      "Elapsed Time: 0:02:18\n",
      "53\n",
      "Train / Test performance: 3866.8 4128.41\n",
      "Elapsed Time: 0:02:20\n",
      "54\n",
      "Train / Test performance: 3937.09 3966.06\n",
      "Elapsed Time: 0:02:23\n",
      "55\n",
      "Train / Test performance: 3912.81 4215.17\n",
      "Elapsed Time: 0:02:25\n",
      "56\n",
      "Train / Test performance: 4014.89 4851.79\n",
      "Elapsed Time: 0:02:27\n",
      "57\n",
      "Train / Test performance: 3919.24 5059.38\n",
      "Elapsed Time: 0:02:29\n",
      "58\n",
      "Train / Test performance: 3963.39 4542.41\n",
      "Elapsed Time: 0:02:32\n",
      "59\n",
      "Train / Test performance: 3882.63 4089.49\n",
      "Elapsed Time: 0:02:34\n",
      "60\n",
      "Train / Test performance: 4120.98 4153.35\n",
      "Elapsed Time: 0:02:36\n",
      "61\n",
      "Train / Test performance: 3858.14 4098.32\n",
      "Elapsed Time: 0:02:39\n",
      "62\n",
      "Train / Test performance: 3934.92 4221.33\n",
      "Elapsed Time: 0:02:41\n",
      "63\n",
      "Train / Test performance: 3814.82 4091.11\n",
      "Elapsed Time: 0:02:44\n",
      "64\n",
      "Train / Test performance: 3806.7 4179.68\n",
      "Elapsed Time: 0:02:46\n",
      "65\n",
      "Train / Test performance: 3913.26 4197.77\n",
      "Elapsed Time: 0:02:49\n",
      "66\n",
      "Train / Test performance: 4051.1 4175.02\n",
      "Elapsed Time: 0:02:53\n",
      "67\n",
      "Train / Test performance: 4103.72 4605.82\n",
      "Elapsed Time: 0:02:55\n",
      "68\n",
      "Train / Test performance: 3849.94 4600.57\n",
      "Elapsed Time: 0:02:57\n",
      "69\n",
      "Train / Test performance: 3935.87 4161.96\n",
      "Elapsed Time: 0:03:00\n",
      "70\n",
      "Train / Test performance: 3884.32 4128.5\n",
      "Elapsed Time: 0:03:02\n",
      "71\n",
      "Train / Test performance: 4002.73 4053.21\n",
      "Elapsed Time: 0:03:04\n",
      "72\n",
      "Train / Test performance: 3935.47 4094.1\n",
      "Elapsed Time: 0:03:07\n",
      "73\n",
      "Train / Test performance: 3925.54 4484.07\n",
      "Elapsed Time: 0:03:09\n",
      "74\n",
      "Train / Test performance: 3944.85 4068.57\n",
      "Elapsed Time: 0:03:11\n",
      "75\n",
      "Train / Test performance: 3963.73 4067.7\n",
      "Elapsed Time: 0:03:14\n",
      "76\n",
      "Train / Test performance: 3945.46 4095.63\n",
      "Elapsed Time: 0:03:16\n",
      "77\n",
      "Train / Test performance: 3835.55 4041.3\n",
      "Elapsed Time: 0:03:18\n",
      "78\n",
      "Train / Test performance: 3879.67 4092.84\n",
      "Elapsed Time: 0:03:21\n",
      "79\n",
      "Train / Test performance: 3822.52 3872.61\n",
      "Elapsed Time: 0:03:23\n",
      "80\n",
      "Train / Test performance: 3948.92 4042.37\n",
      "Elapsed Time: 0:03:26\n",
      "81\n",
      "Train / Test performance: 3969.66 4186.49\n",
      "Elapsed Time: 0:03:28\n",
      "82\n",
      "Train / Test performance: 3906.81 3945.68\n",
      "Elapsed Time: 0:03:30\n",
      "83\n",
      "Train / Test performance: 3920.67 4184.29\n",
      "Elapsed Time: 0:03:33\n",
      "84\n",
      "Train / Test performance: 3857.13 4043.7\n",
      "Elapsed Time: 0:03:35\n",
      "85\n",
      "Train / Test performance: 3809.77 4010.41\n",
      "Elapsed Time: 0:03:37\n",
      "86\n",
      "Train / Test performance: 3907.55 4407.06\n",
      "Elapsed Time: 0:03:40\n",
      "87\n",
      "Train / Test performance: 3890.71 3956.56\n",
      "Elapsed Time: 0:03:42\n",
      "88\n",
      "Train / Test performance: 3772.73 3911.53\n",
      "Elapsed Time: 0:03:44\n",
      "89\n",
      "Train / Test performance: 3739.2 3996.25\n",
      "Elapsed Time: 0:03:47\n",
      "90\n",
      "Train / Test performance: 3785.77 3968.36\n",
      "Elapsed Time: 0:03:49\n",
      "91\n",
      "Train / Test performance: 3889.32 4152.85\n",
      "Elapsed Time: 0:03:51\n",
      "92\n",
      "Train / Test performance: 3848.7 3704.36\n",
      "Elapsed Time: 0:03:54\n",
      "93\n",
      "Train / Test performance: 3777.69 4046.0\n",
      "Elapsed Time: 0:03:56\n",
      "94\n",
      "Train / Test performance: 3717.11 3927.71\n",
      "Elapsed Time: 0:03:58\n",
      "95\n",
      "Train / Test performance: 3868.77 4434.09\n",
      "Elapsed Time: 0:04:01\n",
      "96\n",
      "Train / Test performance: 3871.36 4319.42\n",
      "Elapsed Time: 0:04:03\n",
      "97\n",
      "Train / Test performance: 3707.8 4067.83\n",
      "Elapsed Time: 0:04:05\n",
      "98\n",
      "Train / Test performance: 3917.63 4030.31\n",
      "Elapsed Time: 0:04:08\n",
      "99\n",
      "Train / Test performance: 3745.65 4541.21\n",
      "Elapsed Time: 0:04:10\n",
      "Running scVI on 3000 genes\n",
      "scVI will run without batch correction\n",
      "Will work on mode numeric for incorporating library size\n",
      "Will work on mode gene for modeling inverse dispersion param\n",
      "Will apply zero inflation\n",
      "1 hidden layers at 128 each for a final 5 latent space\n",
      "0\n",
      "Train / Test performance: 50877.4 41092.9\n",
      "Elapsed Time: 0:00:00\n",
      "1\n",
      "Train / Test performance: 22949.0 20887.7\n",
      "Elapsed Time: 0:00:02\n",
      "2\n",
      "Train / Test performance: 10330.5 17339.5\n",
      "Elapsed Time: 0:00:04\n",
      "3\n",
      "Train / Test performance: 5192.96 14545.4\n",
      "Elapsed Time: 0:00:07\n",
      "4\n",
      "Train / Test performance: 4761.45 8194.14\n",
      "Elapsed Time: 0:00:09\n",
      "5\n",
      "Train / Test performance: 4773.11 7433.03\n",
      "Elapsed Time: 0:00:11\n",
      "6\n",
      "Train / Test performance: 4542.69 6385.92\n",
      "Elapsed Time: 0:00:14\n",
      "7\n",
      "Train / Test performance: 4477.7 5033.12\n",
      "Elapsed Time: 0:00:16\n",
      "8\n",
      "Train / Test performance: 4395.23 4628.42\n",
      "Elapsed Time: 0:00:19\n",
      "9\n",
      "Train / Test performance: 4380.32 4913.89\n",
      "Elapsed Time: 0:00:21\n",
      "10\n",
      "Train / Test performance: 4357.6 4887.99\n",
      "Elapsed Time: 0:00:23\n",
      "11\n",
      "Train / Test performance: 4412.39 4588.49\n",
      "Elapsed Time: 0:00:25\n",
      "12\n",
      "Train / Test performance: 4602.15 4846.86\n",
      "Elapsed Time: 0:00:28\n",
      "13\n",
      "Train / Test performance: 4391.11 4566.85\n",
      "Elapsed Time: 0:00:30\n",
      "14\n",
      "Train / Test performance: 4440.74 5339.74\n",
      "Elapsed Time: 0:00:33\n",
      "15\n",
      "Train / Test performance: 4296.87 5396.13\n",
      "Elapsed Time: 0:00:35\n",
      "16\n",
      "Train / Test performance: 4429.12 4828.98\n",
      "Elapsed Time: 0:00:37\n",
      "17\n",
      "Train / Test performance: 4425.98 4759.7\n",
      "Elapsed Time: 0:00:40\n",
      "18\n",
      "Train / Test performance: 4066.56 4449.94\n",
      "Elapsed Time: 0:00:42\n",
      "19\n",
      "Train / Test performance: 4413.92 4623.56\n",
      "Elapsed Time: 0:00:44\n",
      "20\n",
      "Train / Test performance: 4235.47 4498.5\n",
      "Elapsed Time: 0:00:47\n",
      "21\n",
      "Train / Test performance: 4322.53 4577.07\n",
      "Elapsed Time: 0:00:49\n",
      "22\n",
      "Train / Test performance: 4247.77 4549.31\n",
      "Elapsed Time: 0:00:51\n",
      "23\n",
      "Train / Test performance: 4244.5 4395.58\n",
      "Elapsed Time: 0:00:54\n",
      "24\n",
      "Train / Test performance: 4354.53 4476.71\n",
      "Elapsed Time: 0:00:56\n",
      "25\n",
      "Train / Test performance: 4305.64 4555.83\n",
      "Elapsed Time: 0:00:58\n",
      "26\n",
      "Train / Test performance: 4184.27 4317.09\n",
      "Elapsed Time: 0:01:01\n",
      "27\n",
      "Train / Test performance: 4043.12 4562.64\n",
      "Elapsed Time: 0:01:03\n",
      "28\n",
      "Train / Test performance: 3964.18 4415.7\n",
      "Elapsed Time: 0:01:05\n",
      "29\n",
      "Train / Test performance: 4190.41 4244.9\n",
      "Elapsed Time: 0:01:08\n",
      "30\n",
      "Train / Test performance: 3998.95 4117.62\n",
      "Elapsed Time: 0:01:10\n",
      "31\n",
      "Train / Test performance: 4197.4 4191.2\n",
      "Elapsed Time: 0:01:12\n",
      "32\n",
      "Train / Test performance: 4047.87 4545.46\n",
      "Elapsed Time: 0:01:15\n",
      "33\n",
      "Train / Test performance: 4088.21 5146.05\n",
      "Elapsed Time: 0:01:17\n",
      "34\n",
      "Train / Test performance: 4110.74 4184.19\n",
      "Elapsed Time: 0:01:19\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "35\n",
      "Train / Test performance: 4123.87 4328.13\n",
      "Elapsed Time: 0:01:22\n",
      "36\n",
      "Train / Test performance: 3992.49 4338.34\n",
      "Elapsed Time: 0:01:24\n",
      "37\n",
      "Train / Test performance: 4038.86 4234.3\n",
      "Elapsed Time: 0:01:26\n",
      "38\n",
      "Train / Test performance: 3996.85 4603.12\n",
      "Elapsed Time: 0:01:29\n",
      "39\n",
      "Train / Test performance: 4084.51 4506.16\n",
      "Elapsed Time: 0:01:31\n",
      "40\n",
      "Train / Test performance: 4392.67 4387.71\n",
      "Elapsed Time: 0:01:34\n",
      "41\n",
      "Train / Test performance: 4044.75 4328.52\n",
      "Elapsed Time: 0:01:36\n",
      "42\n",
      "Train / Test performance: 4012.43 4214.02\n",
      "Elapsed Time: 0:01:38\n",
      "43\n",
      "Train / Test performance: 3865.53 4316.92\n",
      "Elapsed Time: 0:01:40\n",
      "44\n",
      "Train / Test performance: 3954.21 4423.58\n",
      "Elapsed Time: 0:01:43\n",
      "45\n",
      "Train / Test performance: 4088.78 4650.15\n",
      "Elapsed Time: 0:01:45\n",
      "46\n",
      "Train / Test performance: 3860.98 4393.1\n",
      "Elapsed Time: 0:01:47\n",
      "47\n",
      "Train / Test performance: 4170.73 4081.31\n",
      "Elapsed Time: 0:01:50\n",
      "48\n",
      "Train / Test performance: 4076.59 4261.13\n",
      "Elapsed Time: 0:01:52\n",
      "49\n",
      "Train / Test performance: 3975.09 4108.55\n",
      "Elapsed Time: 0:01:55\n",
      "50\n",
      "Train / Test performance: 4016.79 4139.68\n",
      "Elapsed Time: 0:01:57\n",
      "51\n",
      "Train / Test performance: 3890.91 4836.24\n",
      "Elapsed Time: 0:01:59\n",
      "52\n",
      "Train / Test performance: 3987.35 3993.36\n",
      "Elapsed Time: 0:02:02\n",
      "53\n",
      "Train / Test performance: 4102.84 4257.14\n",
      "Elapsed Time: 0:02:04\n",
      "54\n",
      "Train / Test performance: 4133.82 4470.54\n",
      "Elapsed Time: 0:02:06\n",
      "55\n",
      "Train / Test performance: 3874.42 4225.65\n",
      "Elapsed Time: 0:02:09\n",
      "56\n",
      "Train / Test performance: 3857.2 3933.64\n",
      "Elapsed Time: 0:02:11\n",
      "57\n",
      "Train / Test performance: 3916.07 4269.21\n",
      "Elapsed Time: 0:02:13\n",
      "58\n",
      "Train / Test performance: 3906.46 4540.21\n",
      "Elapsed Time: 0:02:16\n",
      "59\n",
      "Train / Test performance: 3926.29 4334.82\n",
      "Elapsed Time: 0:02:18\n",
      "60\n",
      "Train / Test performance: 3766.83 4026.44\n",
      "Elapsed Time: 0:02:20\n",
      "61\n",
      "Train / Test performance: 3950.98 4163.69\n",
      "Elapsed Time: 0:02:23\n",
      "62\n",
      "Train / Test performance: 3832.9 4588.57\n",
      "Elapsed Time: 0:02:25\n",
      "63\n",
      "Train / Test performance: 3884.2 4231.34\n",
      "Elapsed Time: 0:02:27\n",
      "64\n",
      "Train / Test performance: 3920.99 4078.71\n",
      "Elapsed Time: 0:02:30\n",
      "65\n",
      "Train / Test performance: 3919.08 4397.02\n",
      "Elapsed Time: 0:02:32\n",
      "66\n",
      "Train / Test performance: 3991.52 4321.27\n",
      "Elapsed Time: 0:02:34\n",
      "67\n",
      "Train / Test performance: 4002.37 4341.48\n",
      "Elapsed Time: 0:02:37\n",
      "68\n",
      "Train / Test performance: 3946.52 4448.52\n",
      "Elapsed Time: 0:02:39\n",
      "69\n",
      "Train / Test performance: 3845.27 4026.94\n",
      "Elapsed Time: 0:02:42\n",
      "70\n",
      "Train / Test performance: 4144.1 4108.5\n",
      "Elapsed Time: 0:02:44\n",
      "71\n",
      "Train / Test performance: 3961.64 4218.81\n",
      "Elapsed Time: 0:02:46\n",
      "72\n",
      "Train / Test performance: 4024.23 4413.12\n",
      "Elapsed Time: 0:02:49\n",
      "73\n",
      "Train / Test performance: 3873.99 4131.57\n",
      "Elapsed Time: 0:02:51\n",
      "74\n",
      "Train / Test performance: 3964.54 4233.49\n",
      "Elapsed Time: 0:02:53\n",
      "75\n",
      "Train / Test performance: 3897.89 4144.96\n",
      "Elapsed Time: 0:02:56\n",
      "76\n",
      "Train / Test performance: 3892.78 4125.94\n",
      "Elapsed Time: 0:02:58\n",
      "77\n",
      "Train / Test performance: 3743.04 4142.61\n",
      "Elapsed Time: 0:03:00\n",
      "78\n",
      "Train / Test performance: 3694.22 4150.37\n",
      "Elapsed Time: 0:03:02\n",
      "79\n",
      "Train / Test performance: 3982.36 4579.77\n",
      "Elapsed Time: 0:03:05\n",
      "80\n",
      "Train / Test performance: 3718.39 3862.82\n",
      "Elapsed Time: 0:03:07\n",
      "81\n",
      "Train / Test performance: 4055.47 3893.86\n",
      "Elapsed Time: 0:03:10\n",
      "82\n",
      "Train / Test performance: 3847.84 4127.75\n",
      "Elapsed Time: 0:03:12\n",
      "83\n",
      "Train / Test performance: 3942.9 4456.91\n",
      "Elapsed Time: 0:03:14\n",
      "84\n",
      "Train / Test performance: 3783.0 4107.19\n",
      "Elapsed Time: 0:03:17\n",
      "85\n",
      "Train / Test performance: 3824.99 4234.24\n",
      "Elapsed Time: 0:03:19\n",
      "86\n",
      "Train / Test performance: 3765.2 4090.61\n",
      "Elapsed Time: 0:03:21\n",
      "87\n",
      "Train / Test performance: 3857.46 3788.3\n",
      "Elapsed Time: 0:03:24\n",
      "88\n",
      "Train / Test performance: 4161.64 4260.73\n",
      "Elapsed Time: 0:03:26\n",
      "89\n",
      "Train / Test performance: 3890.61 4379.84\n",
      "Elapsed Time: 0:03:28\n",
      "90\n",
      "Train / Test performance: 3688.69 3905.66\n",
      "Elapsed Time: 0:03:31\n",
      "91\n",
      "Train / Test performance: 3847.98 3916.55\n",
      "Elapsed Time: 0:03:33\n",
      "92\n",
      "Train / Test performance: 3753.48 4033.05\n",
      "Elapsed Time: 0:03:35\n",
      "93\n",
      "Train / Test performance: 3705.86 3781.88\n",
      "Elapsed Time: 0:03:38\n",
      "94\n",
      "Train / Test performance: 3766.11 3977.16\n",
      "Elapsed Time: 0:03:40\n",
      "95\n",
      "Train / Test performance: 3837.56 3861.23\n",
      "Elapsed Time: 0:03:42\n",
      "96\n",
      "Train / Test performance: 3710.49 4185.6\n",
      "Elapsed Time: 0:03:45\n",
      "97\n",
      "Train / Test performance: 3747.73 3939.26\n",
      "Elapsed Time: 0:03:47\n",
      "98\n",
      "Train / Test performance: 3922.17 3795.76\n",
      "Elapsed Time: 0:03:49\n",
      "99\n",
      "Train / Test performance: 3836.83 3903.77\n",
      "Elapsed Time: 0:03:52\n",
      "Running scVI on 3000 genes\n",
      "scVI will run without batch correction\n",
      "Will work on mode numeric for incorporating library size\n",
      "Will work on mode gene for modeling inverse dispersion param\n",
      "Will apply zero inflation\n",
      "1 hidden layers at 128 each for a final 5 latent space\n",
      "0\n",
      "Train / Test performance: 50682.4 44718.7\n",
      "Elapsed Time: 0:00:00\n",
      "1\n",
      "Train / Test performance: 26093.4 26004.0\n",
      "Elapsed Time: 0:00:02\n",
      "2\n",
      "Train / Test performance: 9909.48 17476.1\n",
      "Elapsed Time: 0:00:04\n",
      "3\n",
      "Train / Test performance: 5680.36 11927.0\n",
      "Elapsed Time: 0:00:07\n",
      "4\n",
      "Train / Test performance: 4871.19 9694.47\n",
      "Elapsed Time: 0:00:09\n",
      "5\n",
      "Train / Test performance: 4438.71 5710.09\n",
      "Elapsed Time: 0:00:11\n",
      "6\n",
      "Train / Test performance: 5015.79 6533.98\n",
      "Elapsed Time: 0:00:14\n",
      "7\n",
      "Train / Test performance: 4702.44 5287.08\n",
      "Elapsed Time: 0:00:16\n",
      "8\n",
      "Train / Test performance: 4606.51 4768.62\n",
      "Elapsed Time: 0:00:19\n",
      "9\n",
      "Train / Test performance: 4426.92 4811.58\n",
      "Elapsed Time: 0:00:21\n",
      "10\n",
      "Train / Test performance: 4403.75 4390.72\n",
      "Elapsed Time: 0:00:23\n",
      "11\n",
      "Train / Test performance: 4535.37 4782.7\n",
      "Elapsed Time: 0:00:26\n",
      "12\n",
      "Train / Test performance: 4294.3 4720.29\n",
      "Elapsed Time: 0:00:28\n",
      "13\n",
      "Train / Test performance: 4419.39 4692.82\n",
      "Elapsed Time: 0:00:30\n",
      "14\n",
      "Train / Test performance: 4382.15 4913.0\n",
      "Elapsed Time: 0:00:33\n",
      "15\n",
      "Train / Test performance: 4428.28 4850.08\n",
      "Elapsed Time: 0:00:35\n",
      "16\n",
      "Train / Test performance: 4189.23 4687.1\n",
      "Elapsed Time: 0:00:37\n",
      "17\n",
      "Train / Test performance: 4072.73 4722.34\n",
      "Elapsed Time: 0:00:40\n",
      "18\n",
      "Train / Test performance: 4429.53 4779.53\n",
      "Elapsed Time: 0:00:42\n",
      "19\n",
      "Train / Test performance: 4404.65 4640.04\n",
      "Elapsed Time: 0:00:44\n",
      "20\n",
      "Train / Test performance: 4305.53 4556.38\n",
      "Elapsed Time: 0:00:47\n",
      "21\n",
      "Train / Test performance: 4319.86 4655.93\n",
      "Elapsed Time: 0:00:49\n",
      "22\n",
      "Train / Test performance: 4387.99 4480.04\n",
      "Elapsed Time: 0:00:51\n",
      "23\n",
      "Train / Test performance: 4063.54 4366.05\n",
      "Elapsed Time: 0:00:54\n",
      "24\n",
      "Train / Test performance: 4249.14 4253.1\n",
      "Elapsed Time: 0:00:56\n",
      "25\n",
      "Train / Test performance: 4149.29 4448.19\n",
      "Elapsed Time: 0:00:58\n",
      "26\n",
      "Train / Test performance: 4309.01 4788.69\n",
      "Elapsed Time: 0:01:01\n",
      "27\n",
      "Train / Test performance: 4079.12 4209.96\n",
      "Elapsed Time: 0:01:03\n",
      "28\n",
      "Train / Test performance: 4019.21 4140.5\n",
      "Elapsed Time: 0:01:05\n",
      "29\n",
      "Train / Test performance: 3934.9 4222.76\n",
      "Elapsed Time: 0:01:08\n",
      "30\n",
      "Train / Test performance: 4035.0 4617.98\n",
      "Elapsed Time: 0:01:10\n",
      "31\n",
      "Train / Test performance: 4237.23 5003.11\n",
      "Elapsed Time: 0:01:13\n",
      "32\n",
      "Train / Test performance: 4037.64 4586.57\n",
      "Elapsed Time: 0:01:15\n",
      "33\n",
      "Train / Test performance: 4133.74 4242.54\n",
      "Elapsed Time: 0:01:17\n",
      "34\n",
      "Train / Test performance: 4027.77 4587.78\n",
      "Elapsed Time: 0:01:19\n",
      "35\n",
      "Train / Test performance: 4150.88 4949.82\n",
      "Elapsed Time: 0:01:22\n",
      "36\n",
      "Train / Test performance: 3990.93 4368.65\n",
      "Elapsed Time: 0:01:24\n",
      "37\n",
      "Train / Test performance: 4004.05 4369.38\n",
      "Elapsed Time: 0:01:26\n",
      "38\n",
      "Train / Test performance: 3864.27 4391.32\n",
      "Elapsed Time: 0:01:29\n",
      "39\n",
      "Train / Test performance: 3949.99 4570.31\n",
      "Elapsed Time: 0:01:31\n",
      "40\n",
      "Train / Test performance: 4087.21 5149.69\n",
      "Elapsed Time: 0:01:34\n",
      "41\n",
      "Train / Test performance: 4139.99 4495.16\n",
      "Elapsed Time: 0:01:36\n",
      "42\n",
      "Train / Test performance: 4103.2 4359.19\n",
      "Elapsed Time: 0:01:38\n",
      "43\n",
      "Train / Test performance: 4121.67 4927.22\n",
      "Elapsed Time: 0:01:41\n",
      "44\n",
      "Train / Test performance: 4062.71 4456.34\n",
      "Elapsed Time: 0:01:43\n",
      "45\n",
      "Train / Test performance: 4246.02 4466.24\n",
      "Elapsed Time: 0:01:45\n",
      "46\n",
      "Train / Test performance: 3992.04 4474.41\n",
      "Elapsed Time: 0:01:48\n",
      "47\n",
      "Train / Test performance: 3884.38 4322.38\n",
      "Elapsed Time: 0:01:50\n",
      "48\n",
      "Train / Test performance: 4137.11 4280.25\n",
      "Elapsed Time: 0:01:52\n",
      "49\n",
      "Train / Test performance: 3993.16 4189.18\n",
      "Elapsed Time: 0:01:55\n",
      "50\n",
      "Train / Test performance: 3911.32 4454.55\n",
      "Elapsed Time: 0:01:57\n",
      "51\n",
      "Train / Test performance: 3855.86 4238.64\n",
      "Elapsed Time: 0:01:59\n",
      "52\n",
      "Train / Test performance: 3974.84 4155.79\n",
      "Elapsed Time: 0:02:02\n",
      "53\n",
      "Train / Test performance: 3875.53 4194.97\n",
      "Elapsed Time: 0:02:04\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "54\n",
      "Train / Test performance: 4290.16 4603.71\n",
      "Elapsed Time: 0:02:06\n",
      "55\n",
      "Train / Test performance: 3976.65 4023.47\n",
      "Elapsed Time: 0:02:09\n",
      "56\n",
      "Train / Test performance: 4048.04 4316.09\n",
      "Elapsed Time: 0:02:11\n",
      "57\n",
      "Train / Test performance: 3909.61 4365.85\n",
      "Elapsed Time: 0:02:13\n",
      "58\n",
      "Train / Test performance: 3874.02 3887.55\n",
      "Elapsed Time: 0:02:16\n",
      "59\n",
      "Train / Test performance: 3867.3 4062.63\n",
      "Elapsed Time: 0:02:18\n",
      "60\n",
      "Train / Test performance: 3846.94 4129.0\n",
      "Elapsed Time: 0:02:21\n",
      "61\n",
      "Train / Test performance: 4030.65 4335.27\n",
      "Elapsed Time: 0:02:23\n",
      "62\n",
      "Train / Test performance: 4060.32 4196.48\n",
      "Elapsed Time: 0:02:25\n",
      "63\n",
      "Train / Test performance: 3882.13 4162.33\n",
      "Elapsed Time: 0:02:28\n",
      "64\n",
      "Train / Test performance: 3938.74 4474.98\n",
      "Elapsed Time: 0:02:30\n",
      "65\n",
      "Train / Test performance: 3949.5 4606.8\n",
      "Elapsed Time: 0:02:32\n",
      "66\n",
      "Train / Test performance: 4033.79 4108.05\n",
      "Elapsed Time: 0:02:35\n",
      "67\n",
      "Train / Test performance: 3966.8 3953.19\n",
      "Elapsed Time: 0:02:37\n",
      "68\n",
      "Train / Test performance: 4010.93 3943.07\n",
      "Elapsed Time: 0:02:39\n",
      "69\n",
      "Train / Test performance: 4088.97 3975.47\n",
      "Elapsed Time: 0:02:42\n",
      "70\n",
      "Train / Test performance: 3939.63 4418.8\n",
      "Elapsed Time: 0:02:44\n",
      "71\n",
      "Train / Test performance: 3991.3 4085.71\n",
      "Elapsed Time: 0:02:46\n",
      "72\n",
      "Train / Test performance: 3859.63 3984.16\n",
      "Elapsed Time: 0:02:49\n",
      "73\n",
      "Train / Test performance: 3826.78 3929.49\n",
      "Elapsed Time: 0:02:51\n",
      "74\n",
      "Train / Test performance: 3906.41 3869.42\n",
      "Elapsed Time: 0:02:53\n",
      "75\n",
      "Train / Test performance: 3947.26 3896.17\n",
      "Elapsed Time: 0:02:56\n",
      "76\n",
      "Train / Test performance: 3750.03 4041.33\n",
      "Elapsed Time: 0:02:58\n",
      "77\n",
      "Train / Test performance: 3885.72 4326.94\n",
      "Elapsed Time: 0:03:00\n",
      "78\n",
      "Train / Test performance: 3980.2 4325.97\n",
      "Elapsed Time: 0:03:03\n",
      "79\n",
      "Train / Test performance: 3784.11 4006.99\n",
      "Elapsed Time: 0:03:05\n",
      "80\n",
      "Train / Test performance: 4013.44 4085.31\n",
      "Elapsed Time: 0:03:07\n",
      "81\n",
      "Train / Test performance: 3749.31 3957.9\n",
      "Elapsed Time: 0:03:10\n",
      "82\n",
      "Train / Test performance: 3814.95 4189.68\n",
      "Elapsed Time: 0:03:12\n",
      "83\n",
      "Train / Test performance: 3670.19 3889.23\n",
      "Elapsed Time: 0:03:14\n",
      "84\n",
      "Train / Test performance: 3840.68 4291.49\n",
      "Elapsed Time: 0:03:17\n",
      "85\n",
      "Train / Test performance: 3695.38 4185.33\n",
      "Elapsed Time: 0:03:19\n",
      "86\n",
      "Train / Test performance: 3818.43 4494.67\n",
      "Elapsed Time: 0:03:21\n",
      "87\n",
      "Train / Test performance: 3888.29 4239.08\n",
      "Elapsed Time: 0:03:24\n",
      "88\n",
      "Train / Test performance: 3758.96 4083.37\n",
      "Elapsed Time: 0:03:26\n",
      "89\n",
      "Train / Test performance: 3731.15 4208.38\n",
      "Elapsed Time: 0:03:28\n",
      "90\n",
      "Train / Test performance: 3992.82 3726.41\n",
      "Elapsed Time: 0:03:31\n",
      "91\n",
      "Train / Test performance: 4006.65 4095.61\n",
      "Elapsed Time: 0:03:33\n",
      "92\n",
      "Train / Test performance: 3817.95 3936.01\n",
      "Elapsed Time: 0:03:36\n",
      "93\n",
      "Train / Test performance: 3832.16 4109.06\n",
      "Elapsed Time: 0:03:38\n",
      "94\n",
      "Train / Test performance: 3803.69 4094.27\n",
      "Elapsed Time: 0:03:40\n",
      "95\n",
      "Train / Test performance: 3895.48 4102.44\n",
      "Elapsed Time: 0:03:43\n",
      "96\n",
      "Train / Test performance: 3903.99 5902.25\n",
      "Elapsed Time: 0:03:45\n",
      "97\n",
      "Train / Test performance: 3970.94 4064.05\n",
      "Elapsed Time: 0:03:47\n",
      "98\n",
      "Train / Test performance: 3663.55 3986.73\n",
      "Elapsed Time: 0:03:50\n",
      "99\n",
      "Train / Test performance: 4066.2 4163.23\n",
      "Elapsed Time: 0:03:52\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/romain/tensorflow/lib/python2.7/site-packages/ipykernel_launcher.py:9: RuntimeWarning: divide by zero encountered in divide\n",
      "  if __name__ == '__main__':\n",
      "/home/romain/tensorflow/lib/python2.7/site-packages/ipykernel_launcher.py:29: RuntimeWarning: invalid value encountered in subtract\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running scVI on 3000 genes\n",
      "scVI will run without batch correction\n",
      "Will work on mode numeric for incorporating library size\n",
      "Will work on mode gene for modeling inverse dispersion param\n",
      "Will apply zero inflation\n",
      "1 hidden layers at 128 each for a final 5 latent space\n",
      "0\n",
      "Train / Test performance: 50609.4 39747.3\n",
      "Elapsed Time: 0:00:00\n",
      "1\n",
      "Train / Test performance: 24142.7 18678.6\n",
      "Elapsed Time: 0:00:02\n",
      "2\n",
      "Train / Test performance: 9283.51 15586.4\n",
      "Elapsed Time: 0:00:04\n",
      "3\n",
      "Train / Test performance: 5412.33 15385.3\n",
      "Elapsed Time: 0:00:07\n",
      "4\n",
      "Train / Test performance: 4852.1 8915.97\n",
      "Elapsed Time: 0:00:09\n",
      "5\n",
      "Train / Test performance: 5035.4 7817.76\n",
      "Elapsed Time: 0:00:11\n",
      "6\n",
      "Train / Test performance: 4575.5 5397.38\n",
      "Elapsed Time: 0:00:14\n",
      "7\n",
      "Train / Test performance: 4880.64 5505.84\n",
      "Elapsed Time: 0:00:16\n",
      "8\n",
      "Train / Test performance: 4530.97 5027.36\n",
      "Elapsed Time: 0:00:18\n",
      "9\n",
      "Train / Test performance: 4529.24 5144.93\n",
      "Elapsed Time: 0:00:21\n",
      "10\n",
      "Train / Test performance: 4559.46 4696.4\n",
      "Elapsed Time: 0:00:23\n",
      "11\n",
      "Train / Test performance: 4431.92 5134.27\n",
      "Elapsed Time: 0:00:25\n",
      "12\n",
      "Train / Test performance: 4463.99 5145.84\n",
      "Elapsed Time: 0:00:28\n",
      "13\n",
      "Train / Test performance: 4267.2 4903.23\n",
      "Elapsed Time: 0:00:30\n",
      "14\n",
      "Train / Test performance: 4387.77 5112.96\n",
      "Elapsed Time: 0:00:32\n",
      "15\n",
      "Train / Test performance: 4487.78 4396.1\n",
      "Elapsed Time: 0:00:35\n",
      "16\n",
      "Train / Test performance: 4245.35 4780.17\n",
      "Elapsed Time: 0:00:37\n",
      "17\n",
      "Train / Test performance: 4417.46 4338.43\n",
      "Elapsed Time: 0:00:39\n",
      "18\n",
      "Train / Test performance: 4443.59 4648.56\n",
      "Elapsed Time: 0:00:42\n",
      "19\n",
      "Train / Test performance: 4178.93 4302.43\n",
      "Elapsed Time: 0:00:44\n",
      "20\n",
      "Train / Test performance: 4332.99 4253.15\n",
      "Elapsed Time: 0:00:47\n",
      "21\n",
      "Train / Test performance: 4321.15 4964.6\n",
      "Elapsed Time: 0:00:49\n",
      "22\n",
      "Train / Test performance: 4077.83 4686.34\n",
      "Elapsed Time: 0:00:51\n",
      "23\n",
      "Train / Test performance: 4316.99 4975.71\n",
      "Elapsed Time: 0:00:54\n",
      "24\n",
      "Train / Test performance: 4104.06 4187.56\n",
      "Elapsed Time: 0:00:56\n",
      "25\n",
      "Train / Test performance: 4151.75 4822.19\n",
      "Elapsed Time: 0:00:58\n",
      "26\n",
      "Train / Test performance: 3911.86 4440.0\n",
      "Elapsed Time: 0:01:00\n",
      "27\n",
      "Train / Test performance: 4133.12 4432.93\n",
      "Elapsed Time: 0:01:03\n",
      "28\n",
      "Train / Test performance: 4108.64 4674.77\n",
      "Elapsed Time: 0:01:05\n",
      "29\n",
      "Train / Test performance: 4378.66 4515.94\n",
      "Elapsed Time: 0:01:07\n",
      "30\n",
      "Train / Test performance: 4326.93 4401.89\n",
      "Elapsed Time: 0:01:10\n",
      "31\n",
      "Train / Test performance: 4249.15 4507.97\n",
      "Elapsed Time: 0:01:12\n",
      "32\n",
      "Train / Test performance: 4124.24 4485.85\n",
      "Elapsed Time: 0:01:15\n",
      "33\n",
      "Train / Test performance: 4230.85 4378.0\n",
      "Elapsed Time: 0:01:17\n",
      "34\n",
      "Train / Test performance: 4080.56 4427.51\n",
      "Elapsed Time: 0:01:19\n",
      "35\n",
      "Train / Test performance: 3973.2 4306.26\n",
      "Elapsed Time: 0:01:22\n",
      "36\n",
      "Train / Test performance: 4131.05 4270.32\n",
      "Elapsed Time: 0:01:24\n",
      "37\n",
      "Train / Test performance: 4086.73 4633.4\n",
      "Elapsed Time: 0:01:26\n",
      "38\n",
      "Train / Test performance: 4152.24 4261.11\n",
      "Elapsed Time: 0:01:29\n",
      "39\n",
      "Train / Test performance: 4106.41 4360.57\n",
      "Elapsed Time: 0:01:31\n",
      "40\n",
      "Train / Test performance: 3970.84 4281.79\n",
      "Elapsed Time: 0:01:33\n",
      "41\n",
      "Train / Test performance: 4238.84 4587.38\n",
      "Elapsed Time: 0:01:36\n",
      "42\n",
      "Train / Test performance: 4385.87 4186.52\n",
      "Elapsed Time: 0:01:38\n",
      "43\n",
      "Train / Test performance: 3866.32 4153.09\n",
      "Elapsed Time: 0:01:40\n",
      "44\n",
      "Train / Test performance: 4030.78 4091.87\n",
      "Elapsed Time: 0:01:43\n",
      "45\n",
      "Train / Test performance: 4012.91 4170.28\n",
      "Elapsed Time: 0:01:45\n",
      "46\n",
      "Train / Test performance: 4027.59 4803.12\n",
      "Elapsed Time: 0:01:47\n",
      "47\n",
      "Train / Test performance: 3970.27 4538.67\n",
      "Elapsed Time: 0:01:50\n",
      "48\n",
      "Train / Test performance: 3899.08 4222.28\n",
      "Elapsed Time: 0:01:52\n",
      "49\n",
      "Train / Test performance: 4045.59 4832.9\n",
      "Elapsed Time: 0:01:54\n",
      "50\n",
      "Train / Test performance: 3977.25 4576.18\n",
      "Elapsed Time: 0:01:57\n",
      "51\n",
      "Train / Test performance: 4045.29 4610.39\n",
      "Elapsed Time: 0:01:59\n",
      "52\n",
      "Train / Test performance: 3976.93 4498.15\n",
      "Elapsed Time: 0:02:02\n",
      "53\n",
      "Train / Test performance: 3921.81 4331.71\n",
      "Elapsed Time: 0:02:04\n",
      "54\n",
      "Train / Test performance: 4081.85 4500.68\n",
      "Elapsed Time: 0:02:06\n",
      "55\n",
      "Train / Test performance: 3942.87 3990.26\n",
      "Elapsed Time: 0:02:09\n",
      "56\n",
      "Train / Test performance: 4089.17 4307.21\n",
      "Elapsed Time: 0:02:11\n",
      "57\n",
      "Train / Test performance: 3960.23 4147.52\n",
      "Elapsed Time: 0:02:13\n",
      "58\n",
      "Train / Test performance: 3794.22 4066.61\n",
      "Elapsed Time: 0:02:15\n",
      "59\n",
      "Train / Test performance: 3899.59 4325.18\n",
      "Elapsed Time: 0:02:18\n",
      "60\n",
      "Train / Test performance: 3957.63 4553.05\n",
      "Elapsed Time: 0:02:20\n",
      "61\n",
      "Train / Test performance: 3992.21 4853.14\n",
      "Elapsed Time: 0:02:22\n",
      "62\n",
      "Train / Test performance: 3790.66 4097.95\n",
      "Elapsed Time: 0:02:25\n",
      "63\n",
      "Train / Test performance: 3978.57 4195.17\n",
      "Elapsed Time: 0:02:27\n",
      "64\n",
      "Train / Test performance: 3894.35 3888.23\n",
      "Elapsed Time: 0:02:30\n",
      "65\n",
      "Train / Test performance: 3768.09 4451.15\n",
      "Elapsed Time: 0:02:32\n",
      "66\n",
      "Train / Test performance: 3962.44 4062.32\n",
      "Elapsed Time: 0:02:34\n",
      "67\n",
      "Train / Test performance: 3941.67 4029.72\n",
      "Elapsed Time: 0:02:37\n",
      "68\n",
      "Train / Test performance: 3686.61 3979.22\n",
      "Elapsed Time: 0:02:39\n",
      "69\n",
      "Train / Test performance: 3883.81 4127.23\n",
      "Elapsed Time: 0:02:41\n",
      "70\n",
      "Train / Test performance: 3908.87 4380.22\n",
      "Elapsed Time: 0:02:44\n",
      "71\n",
      "Train / Test performance: 4001.71 3850.79\n",
      "Elapsed Time: 0:02:46\n",
      "72\n",
      "Train / Test performance: 3840.6 4074.52\n",
      "Elapsed Time: 0:02:48\n",
      "73\n",
      "Train / Test performance: 4038.94 4549.04\n",
      "Elapsed Time: 0:02:51\n",
      "74\n",
      "Train / Test performance: 3763.11 4114.3\n",
      "Elapsed Time: 0:02:53\n",
      "75\n",
      "Train / Test performance: 3918.74 4129.24\n",
      "Elapsed Time: 0:02:55\n",
      "76\n",
      "Train / Test performance: 3870.2 3922.26\n",
      "Elapsed Time: 0:02:58\n",
      "77\n",
      "Train / Test performance: 3902.29 4116.86\n",
      "Elapsed Time: 0:03:00\n",
      "78\n",
      "Train / Test performance: 3847.6 4163.52\n",
      "Elapsed Time: 0:03:02\n",
      "79\n",
      "Train / Test performance: 3913.52 4470.9\n",
      "Elapsed Time: 0:03:05\n",
      "80\n",
      "Train / Test performance: 3908.77 3972.47\n",
      "Elapsed Time: 0:03:07\n",
      "81\n",
      "Train / Test performance: 3679.35 4176.43\n",
      "Elapsed Time: 0:03:09\n",
      "82\n",
      "Train / Test performance: 3943.32 4173.75\n",
      "Elapsed Time: 0:03:12\n",
      "83\n",
      "Train / Test performance: 3796.48 4073.38\n",
      "Elapsed Time: 0:03:14\n",
      "84\n",
      "Train / Test performance: 3852.16 3865.84\n",
      "Elapsed Time: 0:03:16\n",
      "85\n",
      "Train / Test performance: 3692.06 4112.7\n",
      "Elapsed Time: 0:03:19\n",
      "86\n",
      "Train / Test performance: 4005.65 3940.84\n",
      "Elapsed Time: 0:03:21\n",
      "87\n",
      "Train / Test performance: 3733.07 3888.45\n",
      "Elapsed Time: 0:03:23\n",
      "88\n",
      "Train / Test performance: 3698.95 3947.89\n",
      "Elapsed Time: 0:03:26\n",
      "89\n",
      "Train / Test performance: 3754.37 3940.98\n",
      "Elapsed Time: 0:03:28\n",
      "90\n",
      "Train / Test performance: 3738.66 4196.91\n",
      "Elapsed Time: 0:03:30\n",
      "91\n",
      "Train / Test performance: 3895.13 4063.27\n",
      "Elapsed Time: 0:03:33\n",
      "92\n",
      "Train / Test performance: 3837.01 4054.3\n",
      "Elapsed Time: 0:03:35\n",
      "93\n",
      "Train / Test performance: 3751.99 4200.32\n",
      "Elapsed Time: 0:03:37\n",
      "94\n",
      "Train / Test performance: 3736.37 4301.37\n",
      "Elapsed Time: 0:03:40\n",
      "95\n",
      "Train / Test performance: 3858.3 3997.37\n",
      "Elapsed Time: 0:03:42\n",
      "96\n",
      "Train / Test performance: 3817.46 3983.37\n",
      "Elapsed Time: 0:03:44\n",
      "97\n",
      "Train / Test performance: 3778.99 4004.3\n",
      "Elapsed Time: 0:03:46\n",
      "98\n",
      "Train / Test performance: 3792.3 4245.25\n",
      "Elapsed Time: 0:03:49\n",
      "99\n",
      "Train / Test performance: 3837.52 3709.27\n",
      "Elapsed Time: 0:03:51\n"
     ]
    }
   ],
   "source": [
    "res = np.zeros((trials, len(list_cell_types), len(list_plots), 4))\n",
    "\n",
    "for t in range(trials):\n",
    "    # train the model for each random seed\n",
    "    tf.reset_default_graph()\n",
    "    expression = tf.placeholder(tf.float32, (None, expression_train.shape[1]), name='x')\n",
    "    kl_scalar = tf.placeholder(tf.float32, (), name='kl_scalar')\n",
    "    optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate, epsilon=epsilon)\n",
    "    training_phase = tf.placeholder(tf.bool, (), name='training_phase')\n",
    "\n",
    "    model = scVI.scVIModel(expression=expression, kl_scale=kl_scalar, \\\n",
    "                             optimize_algo=optimizer, phase=training_phase, \\\n",
    "                              library_size_mean=mean, library_size_var=var, \\\n",
    "                              n_latent=5, log_variational=True)\n",
    "\n",
    "    # Session creation\n",
    "    sess = tf.Session()\n",
    "    sess.run(tf.global_variables_initializer())\n",
    "\n",
    "    train_model(100, expression_train, expression_test, model.train_step)\n",
    "    # then perform the analysis for each cell type\n",
    "    for k in list_cell_types:\n",
    "\n",
    "        indices = np.where(c_train == k)[0]\n",
    "        data = expression_train[indices]\n",
    "        qc = qc_train[indices]\n",
    "        results = sample_posterior(data)\n",
    "        res[t, k - 1, 0] = normalize_plot(process_proba(results[\"dropout\"]), \\\n",
    "               process_proba(results[\"dropout\"], results[\"mean\"], results[\"dispersion\"]), \n",
    "               np.sum(data > 0, axis=1), s=\"number of gene detected\", plot=False)\n",
    "        res[t, k - 1, 1] = normalize_plot(process_proba(results[\"dropout\"]), \\\n",
    "               process_proba(results[\"dropout\"], results[\"mean\"], results[\"dispersion\"]), \n",
    "               np.log(np.sum(data, axis=1)), s=\"log average counts per cells\",  plot=False)\n",
    "        for i, qc_number in enumerate([2, 4, 5]):\n",
    "            res[t, k - 1, 2 + i] = normalize_plot(process_proba(results[\"dropout\"]), \\\n",
    "                   process_proba(results[\"dropout\"], results[\"mean\"], results[\"dispersion\"]), \\\n",
    "                   qc[:, qc_number], cell_qc[qc_number], plot=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "cell_info = np.unique(c_train, return_counts=True)[1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([2592, 1317,  961,  689,  589,  486,   78,   56,   49,   29])"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cell_info"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "collapsed": true,
    "deletable": false,
    "editable": false,
    "run_control": {
     "frozen": true
    }
   },
   "outputs": [],
   "source": [
    "np.save(\"9k_neurons_DE_statistics_multi_celltypes.npy\", res)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "res = np.load(\"9k_neurons_DE_statistics_multi_celltypes.npy\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "t_stats = np.zeros((trials, 6, len(list_plots), 2))\n",
    "p_values = np.zeros_like(t_stats)\n",
    "# get the t_stats from the regression value\n",
    "for t in range(5):\n",
    "    for k in range(6):\n",
    "        for i in range(len(list_plots)):\n",
    "            for j in range(2):\n",
    "                r = res[t, k, i, 2*j +1]\n",
    "                n = cell_info[k]\n",
    "                t_stats[t, k, i, j] = r * np.sqrt((n - 2)/(1 - r**2))\n",
    "                p_values[t, k, i, j] = 2 * scipy.stats.distributions.t.sf(np.abs(t_stats[t, k, i, j]), n)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ubuntu/tensorflow/lib/python2.7/site-packages/ipykernel_launcher.py:4: RuntimeWarning: divide by zero encountered in log10\n",
      "  after removing the cwd from sys.path.\n"
     ]
    }
   ],
   "source": [
    "# concat p_values across experiments and cell types\n",
    "p_values = p_values.reshape((trials * 6, len(list_plots), 2))\n",
    "t_stats = t_stats.reshape((trials * 6, len(list_plots), 2))\n",
    "scores = - np.sign(t_stats) * np.log10(p_values)\n",
    "#scores = -np.log10(p_values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "def finite(x):\n",
    "    return x[np.isfinite(x)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.patches as mpatches"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfAAAAFZCAYAAABnv33AAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzs3XmcHVWZ//HPt7OTgMgWlqwsioAbuyMqKCo4LArKvooi\nKgrOuAD6m6JARFARVFwiiogC4gKIg7IIKM6IGhBkU4clgUCIQFhCNtLdz++PU53cdO5SXd2dXvJ9\n53Vf3beqTtXJvX3v85xTp04pIjAzM7OhpW2gK2BmZmY95wBuZmY2BDmAm5mZDUEO4GZmZkOQA7iZ\nmdkQ5ABuZmY2BDmAD3OSQtKWxe/flvT/+mi/UyS9KGlE8fxWSR/oi30X+/u1pKP7an+DgaRjJP2h\n5vny92Y116Nf/iYGQn/UX9I2kmZKUl/ud00laaKkBySNGei6DDcO4H1I0uFFUOv+CEn/NdD1i4gT\nIuLMVttJmiVpzxb7ejQiJkRER2/rJel0ST/qtv+9I+KS3u7bmqv9m5C0u6Q5A10nWDXZaaTs33QP\nnQl8OYpJMorPw+Lis/yspP+WNLmmrj+Q9FKxfr6kGyVt3X2nxesbkj7Tbfm0YvnImv2FpJ1rttlS\nUsNJOySNlvQVSXOKesySdH7N+llFHTfoVu6vxbGmdVt+erF8l27LG74vjZL4iJgH3AIc36j+Vo0D\neB+KiB8XQW35AzgZmAd8t6f76/pADzaDtV62Zunq/enjfW4C7AFc3W3VvsXneRPS5/nr3dafW6zf\nDHgc+F6d3R8NzAeOKlGV+cDne1D1U4EdgZ2BtYHdgTu7bfMIcGjXE0mvBtbqvqOi5+GoHtS1jB8D\nH+qjfVnBAbwfSXo9cD5wSETMLZa9TNL3JM2V9Likz9d0Qx8j6X8kfVXSM8DpktokfU7SbEn/kvRD\nSS9rcsxPFft+QtL7u637gaTPF79vIOlXkp4rWg23Fce6FJgCXFtk8p+uaSEcJ+lR4OburYbCFpL+\nLOkFSddIWq841iotu65WvqS9gNOAg4vj3V2sX57NN3sNaupxtKRHJT0t6bMNXptdJD1Z+8Uv6T2S\n/lb8vrNS1+kLkuZJOq/J67y/pLuKbR8q/h9N39+ekDRd0u8kLVBq0X1DRS9Fs9ez5v/xx+K9nVuU\nHd3gOD8o6jge+DWwqVb0HG0qaZGk9Wu2317SU5JG1dnX6ZJ+KulHRb3vkfQKSacW79tjkt5Rs33d\n10rSq4BvA28o6vFcTV2/Jek6SQuBPWr/plu8L8dIerio1yOSDm/w0r8duDMiltRbWSz/GbBNg/WL\ngSuB13V7bcYD7wU+CmwlaccGx+9yCfAaSW9psV2XnYCrIuKJSGZFxA+7bXMpKwfko4Hu2wC8iZSo\nfBw4pNHfTg/9Cdhc0tQ+2JcVHMD7iaR1SR/0MyPi1ppVPwDagS2B1wPvAGq7nXYBHgYmAmcBxxSP\nPYDNgQnANxoccy/gk6Qvoa2AZt3g/wnMATYsjnUaEBFxJPAoRYsjIs6tKfMW4FXAOxvs8yjg/aQP\nfzvwtSbHh3TA3wBfAH5SHO+1dTY7htavwW7AK4G3Af9VBIHux/oTsBB4a83iw4DLit8vAC6IiHWA\nLUhfxKtQ6tr8IfApYF3gzcCsYvUPaP7+lnUZcAewAalLtyfjATqATxRl30B6TT7SrEBELAT2Bp6o\n6UF6ArgVOKhm0yOBKyJiWYNd7UsKFC8H/gpcT/qe2Qw4A/hOzbY/oM5rFREPACcAfyzqsW5NmcNI\nn4u1gZW6chu9L0Xw/Bqwd0SsDfwbcFeD+r8a+EeDdUhaCzgYuL3B+vGkVu6D3VYdALwI/JT0mrR6\nPxeRPhdntdiuy+3Af0j6iKRXS3XP398OrCPpVUVSeQjwozrbHQ1cy4q//31L1qGhiGgnvSb1Pt9W\nkQN4Pyg+PD8E7gXOrVk+EXgXcHJELIyIfwFfJX2QujwREV+PiPYimz8cOC8iHo6IF0ldZYeofjf2\nQcDFEXFv8YV8epNqLiMF2qkRsSwibus659fE6UW9FzdYf2nNsf8fcFCV1mcdZV6DPCIWR8TdwN00\n/qK4nKIbUdLapPfj8mLdMmBLSRtExIsRUfdLGjgO+H5E3BgRnRHxeET8veT725KkKaQW1f+LiKUR\n8XvSF2opEXFHRNxe/A3NIgXNsi257i4BjijqNYL02l3aZPvbIuL64gv7p6QE8YtFwL8CmCZp3V68\nVtdExP8Ur3v3VnLd96VY1wlsJ2lcRMyNiPsa7H9dYEGd5VcXPQHPkxLkL3Vb/8li/QJSMnlkt/VH\nk5LUDlJydki9XoxuvgNMkbR3i+0AzgbOIX1WZgKPq/4g0K5W+NuBB0jd/csVCcr7gMuK9+xn9F03\n+gLS62t9xAG8f3wG2BY4ultQnAqMAuYW3ZvPkT6kG9Vs81i3fW0KzK55PhsYSWo1d7dpt/Kz62zT\n5UukjPiGomvxlCbbNqpbs/WzSf/XDRps2xNlXoMna35fRGql13MZcIDSiNgDSN2lXfs+DngF8HdJ\nf5G0T4N9TAYeqrO8zPtbxqbAs0Ui1KXZe7mSotv6V0qnC14gteSqvg/XANtImk760n8+Iv7cZPt5\nNb8vBp6uGejYlfhNoPpr1exvsO77UryOB5Na9XOVBqGtMsis8Cypdd/du4uegLHAicDvJG1cs/7L\nxfpppP/nK7tWKA1424N0HhjSazoW+Pcm/xciYimp92WlQXqS3lRzmuO+YtuOiLgwIt5ICpJnAd+v\n0xN1KakX4xjqd5+/h9Qrcl3x/MfA3pI2bFbXktYGnuuD/VjBAbyPSdod+Czw3ojo/sf6GLAU2CAi\n1i0e60TEtjXbdG8FP0H6susyhfQBm8eq5pK+xGq3rSsiFkTEf0bE5sB+pO63tzWoQ6O6ddf92MuA\np0nd1ssHyxQtudovhFb77clr0FRE3E8Khnuzcvc5EfF/EXEoKYicA/ys6BLt7jFSF3u95a3e3zLm\nAi/vduza97LV6/kt4O/AVsXpgNOAMpdErfI+FK3cK0mt8CNp3vruiVavVZW/wUbvC0WvwNtJvU5/\np/Gg0r+Rkrj6B0+B8hek0xS71Vn/KHAScIGkccXiI0nftddKepJ0imws5U6LXEwKyAfUHOO2mtMc\nq/xtFT1RF5KSkW26rZtNGsz2LuAXdY53NCnBerSo609JidZhJeraUNFbtiWpd8z6iAN4H1IawXoF\nqVvwr93XRxrIdgPwFUnrKA3O2qLFQJXLgU8oDWqawIrzxe11tr0SOEbpOta1gKxJXfdRujRFpG7B\nDlI3I6TAuHnr//Eqjqg59hnAz4rW1z+BsZL+veg2/BxQe03oPFLXaqO/x568BmVcRvqSfTPpCwoA\nSUdI2jAiOlnRUuisU/57wLGS3la8h5tJ2rri+7uK4kt2JpArXR60Gyufh2z1eq4NvAC8WLQ0P1zy\n0POA9bXqIMkfklps+9FHAbzEazUPmNTDAVR13xel65D3LxKipaRz0fXeV4Abge0lja23Usn+pHP8\nDzT4v91ISjq7Lps6GshJA9u6HgcC71LNAMEG+2onfY4/02w7SScrDW4cJ2lk0X2+NmkcQnfHAW/t\n1sODpM1I4yX2qanna0nJ7FErb6qxtY+adSO7res6TbAzMKumt8v6gAN43/ogqVv3Aq16Lfi3i22O\nAkYD95My5J+RWgWNfJ/0pfl7Uua8BPhYvQ0j4tekUe83k7rHb26y362Am0hfZn8EvhkRtxTrzgY+\nV3RtfrLF/7nWpaSBSU+SWhgfL+r1PGkQ1UWkc24LSQPounQF0Wckdb/0BXrwGpR0Oemc8M0R8XTN\n8r2A+yS9SBrQdki98/1FF/KxpHO2zwO/Y0UPQU/f30YOIw1onE/6Al/e3Vni9fxkUX4BqaX5kzIH\nLM4XXw48XLz3mxbL/4cU8O7s4y/gZq/VzcB9wJOSnq5ffJX6N3pf2oD/IAXV+aT3vm5SE+ma5ZuB\n/buturb4u3iB1D19dJPz6JBOUX26SEimAhdGxJM1j1+SPqOHNtlHl8tJvTLNLAK+QvrsPU0a7X5g\nRDzcfcOIeCgiZtbZx5HAXRFxQ21dSQMAXyNpu2K7fyOdJlj+0IrxKN/qtu7iYvnhpCsLrA+p9bgl\nMxtokk4HtoyIIwbo+DeTBjZdNBDHX50kbUMavLdziYGd1oKkjUjJ1OvrDDy0XvCEHGbWlKSdgO1Z\ntVU6LBXjJHYa6HoMF8UVBqtc1mm95y50M2tI0iWkUy0nR0S9y6vMbIC4C93MzGwIcgvczMxsCHIA\nNzMzG4IcwM3MzIYgB3AzM7MhyAHczMxsCHIANzMzG4IcwM3MzIYgB3AzM7MhyAHczMxsCHIANzMz\nG4IcwM3MzIagYX83sra2thg3btxAV8PMzAbAokWLIiKGZWN12AfwcePGsXDhwoGuhpmZDQBJiwe6\nDv1lWGYlZmZmw50DuJmZ2RDkAG5mZjYEOYCbmZkNQQ7gZmZmQ9CQC+ASe0n8Q+JBiVMGuj5mZmYD\nYUgFcIkRwIXA3sA2wKES2wxsrczMzFa/IRXAgZ2BByN4OIKXgCuA/Qe4TmZmZqvdUJvIZTPgsZrn\nc4BdmhVYtAjWXbdf62Q27D333EDXwMy6G2oBvBSJ44HjB7oeZmb9SVLT9RGxmmpiA2GoBfDHgck1\nzycVy1YSwQxgBsD48YRbD2Y2HNUGaEkO2GuYoXYO/C/AVhLTJUYDhwC/HOA6mZmZrXZDqgUeQbvE\nicD1wAjg+xHcN8DVMjMzW+003Ltcxo8fH74bmQ0FVc9n+jyogbvQG5G0KCLGD3Q9+kPDLnSJIySO\nrLP8SInD+rdaZmuGKVOnIqllEAaWbyeJKVOnLl8eESs9ui8zs+GpYQtc4k/A2yJ4sdvy8cDvI9hh\nNdSv19wCt8FMEj//+xM9Lnfgq6fCsmU9Ljdx8mSefPTRHpezwc8t8PqGcwu82TnwUd2DN0AECyVG\n9WOdzNYoB269abWCt9zS4yLz9tij2rHMrE8p1zuBtSOLn3Vb/l7g+cjixlb7aBbAx0mMj2Cl5qvE\n2sDoKhU2s1VVaoFXDfpmNlj8F/DuOstvBa4FehXAvwf8TOKECGYDSEwjzUX+vZ7W1Mz6mFvTZkPZ\nmMjiqe4LI4unlatUl3/DAB7BlyVeBH4vMQEQsAD4YgTfqlpjM+sjFbrQHfTNBo11lGtkZNFeu1C5\nRgHjyuyg6XXgEXwb+HbRbU4EC6rW1MzMbLBRru8D+wD/iiy2K5atB/wEmAbMAg6KLJ5VLgEXAO8C\nFgHHRBZ3Vjz0L4DvKteJkcXC4rgTiv3/oswOms7EJvF6iR8BvwN+JzFDYsti3ZCaBMbMzKyOHwB7\ndVt2CvDbyGIr4LfFc0i3st6qeBwPveqN/hwwD5itXHco153AI8BTxbqWml0HfiDwU+Bm4JjicTvp\nvPgbSLOhmZmZDVmRxe+B+d0W7w9cUvx+CSsGm+0P/DCyiMjidmBd5dqk4nHbI4tTSPf3OAY4GpgS\nWZwSWZS6RrRZKzoD9oxgVs2yv0ncDPwdOK9Kpc3MzFabtRipXDNrlsyILGa0KDUxsphb/P4kMLH4\nvd4trTcD5lJBEfw/CmxTLJqpXN+JLJ4pU75ZAB/ZLXgDEMEsidkRnNbj2pqZma1Oi2iPLHasWjyy\nCOXq8xlylOstwI+A75O68QF2AG5WrncDZ0QWq8yGWqtZAF8mMSWClaZtkpgKLK1c6zWI56g2MxuS\n5inXJpHF3KKV/K9iealbWpf0JWC/yOKvNct+qVxXAXcDV7XaQbNBbBlwk8QxEq8uHscCN5AuQLcW\nPEe1mdmQ9EvSOWmKn9fULD9KuaRcu5JmTKvUfQ5M6Ba8AYgs7iINbju21Q6aXQd+tcQjwH8CHysW\n3wccFMHd1eprZmY2eCjX5cDuwAbKNYfUeP0icKVyHQfMBg4qNr+OdAnZg6TLyFoG2WaHVq6XRxbP\ndqvPekB7ZNHZagetrgO/GziqFxU0MzMbtCKLQxuseludbYM06KwvfBW4Qbk+CXRdS74DcE6xriVf\ny93HpkydymNN7vbU6Lz45ClTeHT27P6qlg03o0ZVmlVt4uTJrTcys34XWcxQrieAM4FtgQDuBz4f\nWVxbZh8O4H3ssUcf9c0prP8tW9ZwHIVvK2k2NEQWvwJ+VbV805nYzMzMbHBq2QKXeAVpuriJEWwn\n8Rpgvwg+3++1MxvmJk+ZUqn3ZfKUKf1QGzMbSsq0wL8LnAosA4jgb8Ah/VkpszXFo7Nnr3K5YaPL\nDmsfteMlJK306L7MzAYv5ZpeZlk9ZQL4WhH8uduy9rpbmpXQPeDUC0BroqqBuFGQ93wDZkPCz+ss\n+1mZgmUGsT0tsQVphBwS76XivK9msOoMdB50lfg1MFtzKNfWpNHnL1OuA2pWrQOMLbOPMgH8o8AM\nYGuJx0m3Ozuih3U1MzOzFV5Jug/5usC+NcsXAB8ss4OWATyCh4E9JcYDbREsqFBRMzMzK0QW1wDX\nKNcbIos/VtlHmVHoHaRJ10+NWN6NfmcE21c5oJmZmS33oHKdBkyjJiZHFu9vVbBMF/p9pMFuN0gc\nHMF8YM0daWRmZtZ3rgFuA24COnpSsEwAb4/g0xIHA7dJHEUxoM3MzMx6Za3I4jNVCpa5jEwAEfwE\nOBi4GNi8ysHMzMxsJb9SrndVKVimBf6Brl8iuFfiTcD+VQ5mZmZmKzkJOE25XgJeIjWaI7JYp1XB\nhgFc4q0R3AxMlZjabfWLvamtmZmZQWSxdtWyzVrgbwFuZuXr05YfE/hF1YPamse3WTUzW5VyCTgc\nmB5ZnKlck4FNIovuM6CuomEAjyArfh7bZzW1NZZvs2pmVtc3gU7graR7g78IXAjs1Kpgy0FsEidJ\nrFNMx3yRxJ0S7+htjZsc73SJxyXuKh7vqll3qsSDEv+QeGd/1cHMzGw12SWy+CiwBCCyeBYYXaZg\nmVHo74/gBeAdwPrAkcAXK1a0rK9G8LricR2AxDaku6BtC+wFfFNiRD/Xw8zMrD8tU64RdN1vJNeG\npBZ5S6UvIwPeBfwwgvsYmIlc9geuiGBpBI8ADwI7D0A9zMzM+srXgKuAjZTrLOAPwBfKFCwTwO+Q\nuIEUwK+XWJuS2UEvnCjxN4nvS7y8WLYZ8FjNNnOKZauQOF5ipsTMdt/41MzMBqnI4sfAp4GzSXf6\nfHdk8dMyZcsE8OOAU4CdIlhE6pvv1cA2iZsk7q3z2B/4FrAF8DrSf+YrPd1/BDMi2DGCHUeWudLd\nzMxs4MwjTaf6v8A45Sp1r5EydyPrBO6sef4M8EzFSnbtY88y20l8F/hV8fRxYHLN6knFMjMzsyFJ\nuc4EjgEeYsU05UEald7UoGufSmwSwdzi6XuAe4vffwlcJnEesCmwFdDyOjkzM7NB7CBgi8jipZ4W\nHHQBHDhX4nWkDGQW8CGACO6TuBK4H2gHPhrRszu3mJmZDTL3AusC/+ppwTL3A/8K8P1i9Hm/i+DI\nJuvOAs5aHfUwMzNbDc4G/qpc9wJLuxZGFvu1KlimBf4AMENiJOlOZJdH8HzVmq4JPHuYmZmVdAlw\nDnAPPbzCq8wgtouAiyReSRp9/jeJ/wG+G8EtFSo77HnKUDMzK2lRZPG1KgVLnQMvZjzbung8DdwN\n/IfEhyI4pMqBzczMjNuU62zSQO3aLvQ7GxdJypwD/yqwD+nOZF+IWD7y+xyJf1Srr5mZmQGvL37u\nWrOs95eRSQiYD7wugoV1NvFUplaaTxOYma0sstijatmmATyCkDgogjMbrPdgNivNYwPMzFamXGOA\nA4Fp1MTkyOKMVmXLTKV6p9T6vqRmZta/Jk+bjIp7O3d/AA3XTZ42ucWebQBdQ7pZVzuwsObRUplB\nbLsAh0vMLnYqICJ4TbW6mplZFXNmz+H8+ef3uNzJ653cD7WxPjIpstirSsEyAfydVXZsZmZmLf2v\ncr06srinpwXLXAc+G0BiI2BshcqZmZlZfbsBxyjXI6TLyFIvdxYte7nLXEa2H+mWnpuS5mqdSpqd\nbdve1NjMzMzYu2rBMoPYziRdn/bPCKYDbwNur3pAMzMzW25BnUepS3bKnANfFsEzEm0SbRHcItHz\nURRmZmaDkHLNIgXODqA9sthRudYDfkK6vGsWcFBk8Ww/HP5OYDLwLKn7fF3gSeWaB3wwsrijUcEy\nLfDnJCYAvwd+LHEBJYe4m5mZDRF7RBaviyx2LJ6fAvw2stgK+G3xvD/cCLwrstggslif1KX+K+Aj\nwDebFSwTwPcHFgGfAH4DPATs26vqmpmZDW77k+4URvHz3f10nF0ji+u7nkQWNwBviCxuB8Y0K9hq\nKtURwK8i2IN0m7NLmm1vZmY2qKzFSOWaWbNkRmQxo9tWAdygXAF8p1g/MbKYW6x/EpjYTzWcq1yf\nAa4onh8MzFOuEbS4vWirqVQ7JDolXuZpU83MbMhZlM5pt9hqt8jiceXaCLhRuf5euzKyiCK494fD\ngAy4unj+P8WyEcBBzQqWGcT2InCPxI3UnPuO4OOVqmpmZjaIRBaPFz//pVxXkW7UNU+5Noks5irX\nJqTLqPvj2E8DH2uw+sFmZcsE8F8UDzMzs2FFucYDbZHFguL3dwBnkO7PfTTwxeLnNX183PMji5OV\n61pSF/5KIov9Wu2jzExsl0iMBrYuDvKPCF6qUmEzM7NBZiJwlXJBiomXRRa/Ua6/AFcq13HAbFp0\nZ1dwafHzy1V3UGYmtncB3yGNPhcwXeJDEfy66kHNzMwGg8jiYeC1dZY/Q5q4rL+Oe0fx83dV91Gm\nC/08YI+I1BcvsQXw3+AAbmZmVoVy3UOdrvMufTIXOrCgK3gXHibNWGNmZmbV7NPbHTQM4BIHFL/O\nlLgOuJKULbwP+EtvD2xmZramiixm93YfzVrgtbOtzQPeUvz+FL6tqJmZWa8p167A14FXAaNJ138v\njCzWaVW2YQCP4Ng+q6GZmZnV8w3gEOCnwI7AUcAryhRs1oX+6QjOlfg69a5R80QuZmZmvRZZPKhc\nIyKLDuBi5forcGqrcs260B8ofs5sso2ZmZlVt0i5RgN3Kde5wFzK3WisaRf6tcVP38DEzMysfxxJ\nCtgnku76ORk4sEzBZl3odad36xJBy2nezMzMrLGa0ehLgLwnZZt1oVee3s3MzMz6V7Mu9MrTu5mZ\nmVn/KnWivK9JvE/ivuJe4zt2W3eqxIMS/5B4Z83yvYplD0qcsvprbWY2PEhq+rChocxUqv3hXuAA\n0k1SlpPYhnQ93LbApsBN0vLr4S4E3g7MAf4i8csI7l99VTYzGx4iVgxvkrTSc1u9lOsVwKeAqdTE\n5Mjira3KDkgAj0iXqNVJ9PYHrohgKfCIxIOkG6sDPBjBw0W5K4ptHcDNzGwo+ynwbeC7QEdPCpa5\nnWj97CBomR1UsBlwe83zOcUygMe6Ld+l0U4kjgeOBxg9uo9ruBq06sJytmxmNmy0RxbfqlKwTAu8\nUnYgcROwcZ1Vn43gmrL7qSKCGcAMgPHjG18KN1h1D9Du4jIzG16Ua73i12uV6yPAVcDSrvWRxfxW\n+ygTwNsj6HF2EMGePS0DPE66iL3LpGIZTZabma0xTl7v5IGugvWNO0hzrXR1uX6qZl0Am7faQbOJ\nXFZkB2LV7CBomR1U8EvgMonzSIPYtgL+TPoPbiUxnRS4DwEO64fj99rkKVM4cOtNK5UzM2vl/Pnn\n97iMg/7gE1lMB1CusZHFktp1ylXqjp/NWuC9zg4akXgP6fZpGwL/LXFXBO+M4D6JK0mD09qBj0ak\nbnuJE4HrSbda+34E91U9fn96dHbjW7y6K9zMzLr5X2D7EstW0Wwil5QdiLERrJwdqHf3A4/gKlKL\nvt66s4Cz6iy/DriuN8c1MzMbDJRrY9Ig7XHK9XpWNJbXAdYqs48y58ArZwdmfcmj880qGtP889No\n3cSpE3ly1pP9Vas13TuBY0hjus6rWb4AOK3MDpqdA1+RHYhK2YFZX/LofLOKltLk1lSNzdO8Pq+K\nJZHFJcAlynVgZPHzKvto1gLvdXZgZmZmq1KuIyKLHwHTlOs/uq+PLM6rU2wlzc6Bp+xAHBhBpezA\nzMzM6hpf/JxQdQfNutCPiCBlB2LV7CBomR2YmZnZqiKLrnuBnNP9MrKymnWh9zo7MDMzs6buVa55\nwG3F4w+RxfNlCjbrQl+RHXS7jMyspzzBjZnZqiKLLZVrCvAm4N+BC5Xrucjida3KlrmM7F6JlbOD\noFR2YNbFE9yYma1KuSYBbyQF8NcC9wF/KFO2ZQCPYEuJlbMD8VwELbMDMzMbJJpPo2AD51HgL8AX\nIosTelKwzO1EK2cHZmY2SFTp5HLQXx1eD+wGHKZcpwD/B/wusvheq4JlutBXZAdBj7IDMzMzayyy\nuFu5HgIeIjWUjwDeArQM4Gp17lHitaTs4M3AFLqyg2i988Fg/PjxsXDhwoGuBlD9XO9wP0fs18Ws\nnMnTJjNn9pxqhSu2wIf6Z0zSoogY33rLgaFcM4ExpCnKbwNuiywaDxqqLVvmzZGYQAriXdkBEUyt\nWuHVyQF88PPrYtZ7zT4PkhzABynl2jCyeKpK2baWOxczgT8C7wEeAN48VIL3ULLxlClIqvsAGq7b\n2JdZma2xWn1P2OBXNXhDuXPge0dQ+QBWzrzHHoNbbul5uT326IfamNlQULZ1PHHqxEo3Jpk4dWKP\ny9jqU+YyMgdvM7MhrNktQX0qaugq0wI3MzOzfqJc/wZMoyYmRxY/bFXOAdzMzGyAKNelwBbAXUBH\nsTiAvgngEqtmB9F652ZmNvh0H+DW/fma1qWuXHsBFwAjgIsiiy+uxsPvCGwTWc9f9DKj0C8Fvky6\njGyn4rFjTw9kZmaDQ0Q0faxJlGsEcCGwN7ANcKhybbMaq3AvsHGVgmVa4Ck7iEpXEZrZMLD7D3Yf\n6CqY9ZedgQcji4cBlOsKYH/g/tV0/A2A+5Xrz8DSroWRxX6tCpa6GxkpO5hbuXpmZmYDYS1GFrOd\ndZkRWcyoeb4Z8FjN8znALqulbsnpVQuWCeApOxArZwdBy+zAzIaHW4+5daCrYFaJjlV7ZDFoT/tG\nFr9TrokKgDZdAAAgAElEQVSk09MAf44s/lWmbJkAfnrVipn11sZTpqRJbhpoNNvUxMmTefLRR/ur\nWmY2fDwOTK55PqlYtloo10HAl4BbSfd/+7pyfSqy+FmrsmUmcvmdxMrZQVAqOzDrLc9QZ2b97C/A\nVso1nRS4DwEOW43H/yywU1erW7k2BG4CWgbwMqPQDwL+DLwPOAj4k8R7e1VdM7PVoNE9BDxXuHWJ\nLNqBE4HrSff7uDKyuG81VqGtW5f5M5SIzVCuCz1lB0WrW6J0dmBmNpBqL4nylKHWSGRxHXDdAB3+\nN8p1PXB58fzgsnUpE+XbunWZl84OzMzMrLHI4lPADOA1xWNGZPGZMmXLtMB/I1EpOzAzW502nrYx\n82Y3vutWw0GPUyc2veGHWX+KLH4O/Lyn5coMYvuUxIHAG4tFMyK4qqcHMjPrb/Nmz6PKlFNVbrVp\n1hvK9YfIYjflWsDKf7UCIrJYp9U+Ss2FHkGl7MDMrC+0GnDmc9s21EQWuxU/1666j4bnsiX+UPxc\nIPFCzWOBxAtVD2hm1lPd5+lek+futuGluBtZy2X1NAzgEexW/Fw7gnVqHmtH0LJp34zE+yTuk+iU\nVtwYRWKaxGKJu4rHt2vW7SBxj8SDEl+T8DUgZsPU5GmTm1765cvCbBjZtvaJco0EdihTsGUXusSl\nERzZalkP3QscAHynzrqHInhdneXfAj4I/Ik0iG4v4Ne9qIOZDVJzZs/h/Pnn97jcyeud3A+1Met7\nynUqcBowTrm6erUFvEQald5SmXPgK2cHonR20EgEDxT7KkViE2CdCG4vnv8QeDcO4GZmNgRFFmcD\nZyvX2ZHFqVX20TCAS6zIDlac8+5RdlDRdIm/Ai8An4vgNtLdYubUbDOnWFaXxPHA8QCjR/djTc1s\n8HFPug0hkcWpyvVyYCtgbM3y37cq2zCAR5CyA3F2BD3ODiRuov5Nyj8bwTUNis0FpkTwjMQOwNXS\nyj0AZUQwgyLJGD/e9zE3W6NU+cQ76NsAUa4PACeRbqJyF7Ar8Efgra3KlrkO/FSJVbODoGl2EMGe\nrfZdp8xSiluWRnCHxEPAK0gTzE+q2XS13i3GzMysn5xEulnY7ZHFHsq1NfCFMgXLDGKrnB30VDHP\n+vwIOiQ2JyUND0cwv7iEbVfSILajgK/39fEHnO+gZWa2plkSWSxRLpRrTGTxd+V6ZZmCZQaxrcgO\ngj0kSmcHjUi8hxSANwT+W+KuCN4JvBk4Q2IZ0AmcEMH8othHgB8A40iD14bfALYKt8100DczG9Lm\nKNe6wNXAjcr1LDC7TMEyAXxJBEskkBgTwd8lSmUHjRRTsa4yHWuzGd8imAls15vjmpmZDSaRxXuK\nX09XrluAlwG/KVO2TACfI7EiOxClswOzeupNuFG7zDNrmdlwp1zr1Vl8T/FzAizvfW6ozCC2FdmB\n6FF2YFaPA7T1mzFUGlE+cerEPq+KWQt3kK6ZqP2L7XoewOatdtDsOvBeZwdmZqvV0sYJoiQnjzZo\nRBbTe7uPZi3wXmcHZmZm1phyCTgcmB5ZnKlcU4CNI4s/tyrbbCKXXmcHZmZm1tQ3SVddvRU4E1hA\nGsy9U6uCZa4DX5EdBGdKpOwgaJkdmJlV5RuT2Bpil8hie+X6K0Bk8axylZoEvMwo9MrZgZlZVb4b\nma0hlinXCIpJgJVrQ1LMbanh/cBr7BLBR4ElABE8C/gWIWZmZr33NdK8KBsp11nAH+irqVSBZRIr\nsoM03Wmp7MDMzMwaiyx+rFx3AG8jDRJ/d2TxQJmyZQL4iuxAnAW8F/hc1cqamZkZFF3n90UWWwN/\n72n5MhO5/Fhi5ewgKJUdmPUJz/duFXWf9a/7c18XbgMpsuhQrn8o15TI4tGelm8awIuu8/siqJQd\nmPUJ3+TFKnKAtiHg5cB9yvVnYGHXwshiv1YFmwbw4rae/5CYEkGPswMzMzNr6v9VLVjmHHjKDsTK\n2UHQMjswMzOz+opz4KdHFpW6DMsE8MrZgZmZmdVXnAPvVK6XRRbP97R8mXPgp0fgE4pmZmZ970Xg\nHuW6kZXPgX+8VcEy58A7JV4WQY+zAzMzM2vqF8Wjx8p0oafsQKycHQQtswMzMzNrLLK4pJj7/BXF\non9EFsvKlC0TwCtnB2ZmZtaYcu0OXALMIs21Mlm5jo4sft+qbJmJXC6RWDk7CEplB2ZmZtbUV4B3\nRBb/AFCuVwCXAzu0KtjyZiYSuwP/B1xIujPZPyXe3JvampmZGQCjuoI3QGTxT2BUmYJlutBTdhCk\n7ECUzg7MzMysqZnKdRHwo+L54cDMMgXLBPBRXcEbIIJ/SuWyAzMzM2vqw8BHYfnA8NtIvd0tlQng\nMyUqZQdmZmbW1EjggsjiPFg+O9uYsgVbqZwdmJmZDVXKdTrwQeCpYtFpkcV1xbpTgeOADuDjkcX1\nFQ/zW2BP0iXbAOOAG4B/a1WwTABP2UGQsoM0O1up7MDMzGyI+2pk8eXaBcq1DXAIsC2wKXCTcr0i\nsuiosP+xkUVX8CayeFG51ipTsOUodFJ2MK7m+Tjgpp7Vz8zMbNjYH7gislgaWTwCPAjsXHFfC5Vr\n+64nyrUDsLhMwTIt8LERrMgOghclSmUHZmZmA2otRipX7bitGZHFjB7s4UTlOoo09us/I4tngc2A\n22u2mVMsq+Jk4KfK9QRpIpeNgYPLFCwTwBdKbB/BnQASpbMDMzOzAbWI9shix0arlesmUtDs7rPA\nt4AzgSh+fgV4f19WL7L4i3JtDbyyWNSnU6mm7ED0ODswMzMbzCKLPctsp1zfBX5VPH0cmFyzelKx\nrGodlgH39rRcmalU/yKxcnbgqVTNzGyYU65NIou5xdP3sCLI/hK4TLnOIw1i2wr48+quX5kWOEXA\n7nF2YGZmNoSdq1yvI3WhzwI+BBBZ3KdcVwL3A+3AR6uMQFcuAZMii8eqVK5UAO9rEl8C9gVeAh4C\njo3guWLdytfWBdcXy/cCLgBGABdF8MWBqHt/mTh5MvP22KNSOTMz63uRxZFN1p0FnNXL/YdyXQe8\nukr5ppeRSUiiPyLEjcB2EbwG+CdwanG82mvr9gK+KTGiuPb8QmBvYBvg0GLbYePJRx8lIuo+gIbr\nnnz00QGuuZmZ9cKdyrVTlYJNW+ARhETl7KDJfm+oeXo78N7i93RtXbAUeERa6dq6ByN4GEDiimLb\n+/uyXmZmZqvZLsDhyjUbWEgaLB6RxWtaFSzThX6nxE4R/KWXlWzk/cBPit+bXVv3WLfluzTaocTx\nwPEAo0f3WT3NzMz62jurFiwTwFN2IFbODlL3d0MSDa+ti+CaYpvPkgYA/LhHtW4hghnADIDx44m+\n3LeZ9b9JUydx8nonVypnNpREFrOV67XAm4pFt0UWd5cpWyaAV8oOImh6bZ3EMcA+wNsilgfZZtfW\n9dk1d2Y2uD02q/GgXEnLx4aYDXXKdRLphim/KBb9SLlmRBZfb1W2zHXgsyVWzg6CUtlBI8WI8k8D\nb4lgUc2qdG2d6H5tnYCtJKaTAvchwGG9qYOZmdkgcBywS2SxEEC5zgH+CLQM4C1vZiJxEqmLe6Pi\n8SOJj/WquvANYG3gRom7JL4NEMF9QNe1db8BPhpBRwTtwInA9cADwJXFtmZmZkOZSJdNd+kolrVU\npgs9ZQdByg5E6eygkQi2bLKu7rV1EVwHXFf1mGZmZoPQxcCflOuq4vm7ge+VKVjmdqKVswMzMzNb\nlXJNB4gszgOOBeYXj2Mji/PL7KNMCzxlB6LH2YGZmZnV9TNgB+X6bWTxNkh3/OwJNRrNKTE9gkeK\n37cHditW3RbBXytWeLUbP358LFy4cKCrAVQfPbsmj7rdeMoU5j3W82mCJ06e7Fnqhqk1+fNgPSdp\nUUSMH+h6dKdcfwV+CnwY+Gr39UXLvKlmLfCUHYjfRlApOzDrrWZB2F/kZjaEHULq0R5JGtTdY81a\n4M2zg6BldjAYuAU+fPl1WTP5fbeeGKwt8C7KtXdk8esqZZsNYjuENGCtKzvo/jAzM7NeqBq8oUkL\nfPkGYu8IKh9goLkFPnz5dVkz+X1vbNmyZcyZM4clS5YMdFVWq7FjxzJp0iRGjRq1yrrB3gLvjZYB\nfKhzAB++/Lqsmfy+N/bII4+w9tprs/766yOtGVf7RgTPPPMMCxYsYPr06ausH84BvMxlZGZmNgQs\nWbKEadOmNQzeHZ0dXP1/TxJaRtvIlxg9qoMJY9pYd8xYNhw9gY3GvIxRbUMrLEhi/fXX56mnnhro\nqlSiXGsB/wlMiSw+qFxbAa+MLH7VquzQeqfMzKypZi3vFztS13pbjIFlY3hpGcxf1Ml8dfIQi1C8\n1HTfHSMW8r4tG06kOWCGeG/DxcAdwBuK54+TBpD3PoBLrMgOgg9KpOwgWu/czMwGj4XtS0GdELXj\nl9sg2pZPr/nYRlfWLdvWOZpNnnlXv9dxDbRFZHGwch0KEFksUl4uIykzlerFwFJWzg4+X6maZmY2\nYBZ2LK1ctq1zNNG2rA9rY4WXlGscpNtqK9cWpJjbUpkAvkUE5wLLAIrbfw7p/gozszXRko5llb+8\n22IUqL3ysadNm8a4ceOYMGECL3/5y/n3f/93Hqswy+IwlJHuvjlZuX4M/JZ0u+2WygTwlyRWZAei\ndHZgZmaDx5KOdqq2v9Q5GrV19ur41157LS+++CJz585l4sSJfOxjPb8zdXt79SRiMIosbgQOAI4B\nLgd2jCxuLVO2TABfkR2IHmUHZmY2eCzpaIeoFsDbYjRtvQzgXcaOHct73/te7r//fgCWLl3KJz/5\nSaZMmcLEiRM54YQTWLx4MQC33norkyZN4pxzzmHjjTfm2GOP7ZM6DBbK9WZgW2AB8AKwTbGspZaD\n2CK4UeJOYFdS6nZSBE/3or5mZjYAXuropGoLvC1GM6JMk6+ERYsW8ZOf/IRdd90VgFNOOYWHHnqI\nu+66i1GjRnHYYYdxxhlncPbZZwPw5JNPMn/+fGbPnk1nZ98kEYPIp2p+HwvsTBqV/tZWBcuMQu/K\nBBYUP7eRIILf97SWZmY2cF7qCCoH8M5RjB7Ru+FP7373uxk5ciQLFy5kww035PrrrycimDFjBn/7\n299Yb731ADjttNM47LDDlgfwtrY28jxnzJgxvTr+YBRZ7Fv7XLkmA312P/DK2YGZmQ0eyzqqlx0R\noxk9YkSvjn/11Vez55570tHRwTXXXMNb3vIW7rrrLhYtWsQOO+ywfLuIoKNjRWU33HBDxo4d26tj\nDyFzgFeV2bBMF/rK2YEonR2Ymdng0dFZvQ98ROcYxvQygC/f14gRHHDAAXzoQx/i9ttvZ9y4cdx3\n331sttlmdbcf4hO1NKVcX6cYJE4al/Y6St6+u8q7WTo7MDOzwaOzo3oAb4vRjGnrmwAeEVxzzTU8\n++yzbLvttnzwgx/kE5/4BP/6178AePzxx7n++uv75FhDwExSr/YdwB+Bz0QWR5QpWOYceOXswKqp\nl23WLvONHMysiogRla8DV4xm7Ijezb697777MmLECCQxdepULrnkErbddlvOOecczjjjDHbddVee\nfvppNttsMz784Q/zzne+s1fHGwoii0uqli3zbsys+b0duDyC/6l6QGvNAdrM+kVn9RZ0W+coxo0Y\nXbn8rFmzGq4bO3YsX/jCF/jCF76wyrrdd9+dOXPmVD7uYKVc97CicbzSKiAii9e02keZc+CVswMz\nMxs8FNUDuDpHsdbI6gHcVrFPb3fQMIBLNM8OgpbZgZmZDR7R1k50RnFDEyFGom6d6qMYhRAdxb8u\nbTGKCSPWmJHg/S6ymN3bfTRrgfc6OzAzs8HjgC2nsnhZB4vaO1i4rIOFL7Wz4KV2XlzWwZJlHSzr\nDDb+13vqlm1vW8T4EcPvOuyBply7Al8nDQ4fDYwAFkYW67Qq2zCAR9Dr7MDMzAaPNonxo0cyfvRI\nNhzoyliXbwCHkO4BviNwFPCKMgVbXlMgsavEXyRelHhJokPihV5V18zMzACILB4ERkQWHZHFxcBe\nZcqVGYVeOTswMzOzphYp12jgLuU6F5hLyTlaSm0UQcoOgo4ISmcHZmZm1tSRpFh8IrAQmAwcWKZg\nmRb4IomUHYgeZQdmZmbW1A7Af0cWLwB5TwqWCeC12cEn6EF2YGZmNlQp1/uA00kjxHeOLGbWrDsV\nOA7oAD4eWVxfLN8LuIA0mvyiyOKLLQ6zL/BV5fo98BPgN5FFe5n6lWlJ70C67vuFCPII/qPoUq9M\n4ksSf5f4m8RVEusWy6dJLJa4q3h8u6bMDhL3SDwo8TV1v3jRzMysb90LHAAr3z5bubYhjQ3blnRK\n+ZvKNUK5RgAXAnsD2wCHFts2FFkcC2xJGmd2KPCQcl1UpnJlAvi+wD8lLpXYRyrVam/lRmC7YjKY\nfwKn1qx7KILXFY8TapZ/C/ggsFXx8Hl4MzPrN5HFA5HFP+qs2h+4IrJYGlk8AjxIutX2zsCDkcXD\nkcVLwBXFtq2Oswz4dbH9HcC7y9SvzFSqx0qMImUUhwIXStwYwQfKHKDBPm+oeXo78N5m20tsAqwT\nwe3F8x+S/oO/rloHM7M1zUvRzpjF02hr62TtpVPZpHM6W7MVrx2xNbuMehVvGrM1E9qG2WxrazFS\nuWrv6TEjspjRy71uRopdXeYUywAe67Z8l2Y7Uq69gYOB3YFbgYuAg8pUolRrOoJlEr8mTa06jhQ8\nKwfwbt5P6vfvMl3ir8ALwOciuI30wtTOZl/7YpmZWQmz25+G0c/Rece3eX7cHJ4fN4d/rnUPvxr3\nGzpGzSE0l/ozaCcjlm5I+5h/rb4K94VFtEcWOzZarVw3ARvXWfXZyOKa/qvYckeRYuCHIoulPSlY\n5nailbIDicYvSnBNsc1nSXc4+3Gxbi4wJYJnJHYArpbYtvV/Y5VjHw8cDzDac++bmQHwaPtTSGsT\niyfB4kkAdBYPAHbfo2n5jpHP92v9BkJksWeFYo+TBnR3mVQso8nyRsc/tOt35donsvhV2UqUaYGv\nyA6C0tlBBE1fFIljSPOtvy0ipXzF/pcWv98h8RBp0pjHSS9El6YvSgQzgBkA48c3SSfNzNYgczqe\nRrFO5S/Fts610thq+yVwmXKdB2xKGpf1Z9LNvrZSrumkGHUIcFgP9nsGUDqAtxzEFsGhEVwdwVKp\nb25wIrEX8GlgvwgW1SzfUEp/HhKbk16UhyOYC7xQTOsqUlKxOro2zMyGjSc656P2tSuXH9kxoXLZ\nadOmsdFGG7Fw4cLlyy666CJ23313ACQxfvx4JkyYwAYbbMChhx7Kc889V/l4fUG53qNcc4A3AP+t\nXNcDRBb3AVcC9wO/AT5aTIPaTrrk+nrgAeDKYtvSh+xJ/Xo6orxH2UET3wDGADcqVff2YsT5m4Ez\nJJaRenVOiGB+UeYjwA9I5+B/jQewmZn1yJPxTK8C+JjOljfIaqqjo4MLLriA0047re76u+++my23\n3JIXXniBgw46iNNPP53zzz+/V8fsjcjiKuCqBuvOAs6qs/w64LqKh/xQTzbuaQDvk2uvI9iywfKf\nAz9vsG4msF1fHN/MbE30VOd8or16K3qtznV7dfxPfepTnHvuuXzkIx9h3XUb72udddZhv/324+qr\nr+7V8YYC5RpLaqDuBoRy/QH4VmSxpFXZnk6J2qPswKwvSVrp0X2ZmTX3DPPpXFY9gK/dsX6vjr/j\njjuy++678+Uvf7npds8++yxXX301u+66a6+ON0T8kDQhzNdJvdPbAJeWKVhmFPrK2YFI2UHQMjsw\n60sRHo9o1hvP8CzRiy70daP3dxE/44wzeOMb38hJJ520yrrtt9+etrY2FixYwFZbbcXFF1/c6+MN\nAdtFFrWztd2iXPeXKVimBV45OzAzs8HjOZ6Fii1wxQjWjw16XYftttuOffbZhy9+cdUpwu+8806e\ne+45lixZwoc//GHe9KY3sWTJsG8r3qlcy7salGsXYGaT7ZcrE8C3i+C4CG4pHh+Enl+bbWZmA2uB\nnoeKLfC2GMX6Wq9P6pHnOd/97nd5/PH6VwOPGjWKD3zgAzzyyCPce++9fXLMQWwH4H+Va5ZyzQL+\nCOykXPco19+aFSwziO1OiV1rpjEtnR2YmdngsbjtucoBXDGCjfoogG+55ZYcfPDBfO1rX+PVr371\nKus7Ojq4+OKLGTduHJtvvnmfHHMQq3xfjzIBPGUH4tHi+RTgHxL3kE5Lvqbqwc3MbPVZOuKFygE8\nQkzsowAO8F//9V9ceunKZ2Nf+9rXIom2tjZe+cpXctVVV7Heen13zMEosphdtWyZAO67fpmZDQPt\nIxZUPgceBJu2VR+FPmvWrJWeT548eaXz2x6k2nNl7kZWOTswM7PBo3PkguotcHUyaWTvR6Fb3+mL\ne3tbA/WuTa5d5ozTrJzun6Xuz/1ZKqdt2Tp07tb89tQTmMAIRrCEJSytuf1F6CWmjuj9KHTrOw7g\n/chfKmZ9w5+lvrFkzGPMWbyUWUuWMGvJEh5evJj7Fy3iwcWLeXzpUp7tWMyLY5+EcXNg3OMw/hEY\nPwvGPQGjXmDyyN5N5GJ9ywHczGwNMaqtjenjxjF93LiBror1gZ5OpWpmZoPYmthbsSb+n8EB3Mxs\n2Bg7dizPPPPMGhXQIoJnnnmGsWPHDnRVVjt3oZuZDROTJk1izpw5PPXUUwNdldVq7NixTJo0aaCr\nsdppuGdq48ePj9obyJuZ2ZpD0qKIGD/Q9egP7kI3MzMbghzAzczMhiAHcDMzsyHIAdzMzGwIGvaD\n2CR1AosHuh6FkUD7QFdiEPLrUp9fl/r8utTn16W+cRExLBurwz6ADyaSZkbEjgNdj8HGr0t9fl3q\n8+tSn1+XNc+wzErMzMyGOwdwMzOzIcgBfPWaMdAVGKT8utTn16U+vy71+XVZw/gcuJmZ2RDkFriZ\nmdkQ5ABuZj0macRA18FsTecAvgaSpIGuQ38Zzv+3wULS0cD3B7oeA0GSvzMrUGGg6zHc+I9xDdL1\n5RM1Ax+G04dK0ojwoI5+I+koSaOA64A3S9ppoOu0ukhqk6SI6Cyer3k3n65IUlsUJPkW1n3IAXwN\nUvPlc7ikT0hafzgFvIjoKBL9kyTtJGniQNdpuJB0KPAD4NiIeAr4FvD5Aa3UaiBpHKTPThGAdpP0\nQ+AdA1y1IaPme+dUYIakAyWtO8DVGhYcwNcgkqZJ+jZwKPBvwPclbTTA1aqsq1VU8/zNwB3AVsDb\ngG8PVN2Gg26tpcXAU8BukqZHxLnARpIOK7YdNj05XSQdDnxF0suK5x8HvgJcDVw7kHUbzLr/LUha\nW9KFwMbAhUAGHCtpzEDUbzhxAB+mug8ykjQVOAV4TUTsExHvA14C9h1q3YFFK3tETatoUrFqO+Aw\n4CzgrUB715evlVe8vmcBv5H0qmLxjcANwBTSawyQA5+RNHY49eTUJC5XA9sAXacKJgKXAdcDr5f0\nZncJr6yru7z4vSuQjyAlgBcBBwOLgFsiYunA1HL4cAAfhopzdR3F7wdI2jwiZpO+hDsldc2XfBmp\nK/AVA1TVHpE0puZ8Woek9SWdB8wsujrfRPo/XQn8vEhSFnjEdHmSXlV8Aa8LTAW+LKmru3gm8GXg\nHZJ2jYirgdnAp4qyw6IVHhFdNwR5DSnJPUHS+sBNpP/rhcB/AGeTkkUrRESnpHGSvgKcKGlLYBkp\nEboKeCIido2IuyRN8qDA3vGLN0xIeqOkIyANUpO0i6Q/AkcAl0n6DHAnqfWwb7HdVUAn8D5Jaw1Q\n1UuR9ArgS8D44vmJwO9JQeZ/I2IxKXBPAPaJiO8URT8BvG7113jokbQ3cKakLUizel0DPAx8CNgB\neB9wN6n7+Lii2GnAJyVNHA6t8KL3YYyk7wBnAD8G9gTeFxG3ADtHxDERcQRwLvDsAFZ3wNXpLt+V\n1NIeD2xKSqQXAv9HunLhG8V2p5K60t2N3gsO4MPHs8BtNR+otwM3RMQBwPGkLuWNgT8BEyXtU2x3\nBvCTiFi0uitcRs3I+X8CnwbGSZoCLCX9H48DOoquzJuBPwIXSTqhSGC2Bx4ckMoPAd16J/4J3Asc\nHBF3A08Cj5H+ZrYgBfF3kQawbS3p8Ii4F3hTRMxbvTXvG917Z4okpBPYCPhYRFwCfAA4oEhsnpY0\nRdI3SC3w21d3nQeDItFpq03aJO0MnAN0RMQJEXEqMFLSUcXyicD1kv4A7ApcUCTeVpED+BAmaYea\nL6B/AlsClxbP1wbmShodEX8jtb7fHxE3APOAN0oaGREPFF/Cg07Xee7i91GkL9UbgY0j4rsR8QRw\nALA4Itoj4nlSa/FnpC67z0XE4cVy60bSXsBvu55HxEOkYL25pF2An5OSvmeAy0mj0B8uWlQXk85r\nUvx9DUk1p5o+Vpxu2gJYi5Qgji8+Pz8jtShPLNa9ndQtvH1E3DowNR9YxWmsTklTJeWSdoiIPwO/\nICXUWxabfgT4IvBURJxE6rE5Nf5/e2ce99lY9/H3Z2YYMjFZki1meiqVJTuJLBWPypJJq0iobCWT\nLFmSbRATD8mSUD1kKyFaLC1kizJPjCVRSZJ17PR5/vhcZ+4zP/dMasZ937/7d71fLy/u8zvnvM45\nzrm+1/VdPl97M9tThkvYZbCoBry7WQc4XCnx2Y3EI1cs7uapwLJk1gsZgMeU/z4N2K8V6xuSlDj3\nGEmTgW1t3wv8CPiQ+spQniXJVvNIOg5Yz/bZtnez/TOo4hszw/alwFskvae1+XrgNmAr4F7iMl8e\nWIYMxj8vx37T9vkDesFzgH5cvitLupZ4ahYCTgamAQ8Sr9VCZdfrgI2BRYDTbe9u++leyq/o/I5K\nGOtc8qx2KXHvbwNzAStIGm37KuAukjeA7Wtt/6IcX3UbZpM6sHURjduqtel+Mqh+EjjN9p3kAzqE\nrJbGAgdI+hQx2r8AsP3HoW68ASS9A7iKDKjfL5uPJjHttcvf6wITyUryOeDK1vGN+/2fA3PFQ5u2\n8c2ziBQAABswSURBVGplT+8BHNtst/0g8CuSObwSmTCJxL/n7ubM4U6DUZ7HGGBXMgF+B/Hc7Efi\ns28EJku6iHh/trN9R/PttJNFhzN6sYjNfOWn+YD1SNhleQDb/yCTnXWAt5T93gPs3XneXnh2LzfV\ngHcRLbfVspI2Iv//vgP8w/bDxWCdSNyeGxF38uXAqsABtk8YrGv/d2gZmrFkxfcV4s5cmcz2Twa2\nkbQUif1PBT5p+/NlVSSohrtNZ7wSeAHA9hnAw5L2KvuNBG4hMe+xth8g2deXdLPxhukendGSDpU0\nAVjG9s+Bx8gE8DpixD8OLGB7e+Ac4PwSirmm43zDfvXYGO6SGLuspD2ALct78lHgCuBzwETbe5RQ\n12kkubTJMp9m+x/VEzbnqe1Eu4RilATsC2xNMrK/Q1zIPwcm2/5e2XcCKfcZ301GTNJ8Jb7a/L02\ncCAwN8lifR1wj+1tJf2KJBHdaPuvZf8RlHnOgF98lyDpYJIZfFNZkb6glBX+mLwvj5T9vgccX1yg\nXUkxPu0V91rEg/MjMvHb3fb4kguwvu0vlhj4j4C7bW/Ucb4R3fQ9zSnKivtEYEkSUvg/4NNkHNrN\n9hvKfq8mZXYHA/OVHJXKy0gVIRiidBqjMgNelLg1N7T9p9a+Xydu5O9JehfJjD0AmEvSc0Nx0Oln\ncF0b2FDSEcCzZdb/K0lbkxn8NEkrAZ8oh3y8JF01x4+sLrk++nm+KxIhn6dI6KVZkY6wfYOks4Fv\nSJpKVOymkrrvrqRtbJXSpieABUglw9/I5O+uMjFegOQC7Ab8N3AoJbu8/RyH4nc0p+l8bwqbAKNt\nr1+SG3cgao6nANtJOop4AzcGzixJo4/O4nyVOUQ14EOU1uCzFpGw/AuJyY2y/acyK37a9gu2z5S0\niaRbSTxqJ6f8ZcjRDKzNR10SXZ4BxgFr2X6645AHSSnK9mTW/0OYnjE9fYCoxruPDuPV/PeKwBbA\ncrb/2s+E57NkcrgxyRL++YBf+BykhJoWBjYFtgE+Q+5/TfI9ndn6Rs6W9ChRCfvfElboKeMj6TW2\n7++Y9DXvyCjiBcP2tZKWIyG6i4mmxMpEmnmj9sKi7N8Tz2+wqC70IULnilvSgiS5aBxx6b2VDDB/\nJYauMWDjSJnP88Cytn8zCJf/kugwLKsR6dNFbB+h6CJfAuxs+7bWMSOJWtx2wOG2bxyES+86FAnZ\nr5DV9g9IPffXSRnYfi/FOHWTy1jS2Mb939r2baKL/0XbVyp1ypcDCzcTRUlfAy6zfUnHsb1kvEeR\nRL57SNLoO9yqMJC0IfkGf2j7l4os82UkF+XEjrBXDWMNIDWpYAggaRlgtY6XflXgSttrk4Sj9Uhm\n8LHAqZI+JOkA4CLgv2w/OZSNN0xfFS0o6TDy8d8HbFRcl68nJUzuOOYF4HLbH7B9o0J9b1t0Po8S\najgPuJVM7s4mcrmnEA3vt5SQTL/PsZuSAJVGGccBm0uau2xrPIvHkMqEEUo993VEzvNbkiZLuoHo\nJVzdOt+LWu4OV1r3+jxR3PsWSeZ7sv07iXk/BOxTJt67EHGk5TrONz3hbUBuoFJd6EOEMcDWkjYn\n5RcfBd4LjJX0QbKKWsH2s8DBJU75NlJv+c4miWuo0emmlbQ4qQeV7beWbfeR1fV2wJbAhWV7O/b4\nTNnWrAjrAEFfYmM/hnYBUnZ3JZH7vJXEtEcQpbo9gW1mZqC7ZQBWhIgel9TUqr8ZuLkYJMqE73oS\n1/4NSfj8FJnMvIuEmq5rn7MbJi2zSyvs1L7X1xAxqF87+gDTn4Xt+0uezdMkifZqkkR7OhmDKPt1\nxXsznKgGfBAoA++IlnGbCqxBRCL2tH2PpAuI6tgajdtY0vuBv9k+R9L5QznuWwaJRuVqGaf2/D5J\ntwATmv1s3yLpyyQ8MDcp47m6v8GgFwbXl0prgmNJywI7kzKos4DFyIToI8Ak2z8ox4wh4ZjFJS3s\n1Hx3Lbafl/RKUp2wEvBnSVNtP1WM+/P0GZrVJf3UkQy+ufwzq0nQsKUVptucJKjdSISeLiBSp6va\nvqHs0xj7x4CvKe2Ilyfv2U2U1XplcKiuyEGgSbqStIgiVrIgabpxFdFansdpnHAJsIOk9SSdRMQQ\nnivnGJLGu+2ClLS6pJ8AR0r6TonLHgdMkbRt2X+k7Uec5iOnU+Q5q5t81pTnO0rSpqTE5z7gQ8Rt\nfB9xne/VMt5nAJvb/o3tnbvRePcTKlgF+CW519+SCctKMN24j3BKma4CtgXm7Ti+v5XosKQJi7T+\n/hgZTy4mIj37ELGaM4iQDZJew4ubjYwr+x5je4/iFawMEnWQHCCUbmGLtv6eSFbYG5J67jvIQLMF\nWVFAVlFNN6gmeW0Gl99QoWW4/ylp/nKv+5N69K2Jctphtv9O9JI3lbRQmcg0nqD/o8TVemFQnQNs\nTQbbn9s+jBiwpYlewHnAZyV9t7iYp5FkNqC7JkhN3oP7EiAb+dKlgDttH2V7d6I0uIGSfQ55DhBl\nws+1k62gN1y+s4jprwucWiZ4O5P3Y03SMWxJRX3u28RgT8f2bbbfb/uHL/vFV/4lXfMRdzPqq518\nZ/l7fmBh8sH8mGSYv5sMsPMTd9+OwAa2jyDxygM8hOVPW4PrjuQ+FiMrQkjM7EzgE4poyDlkpb1r\nOfb5MujuSpppVAr9Je21DNi5xI05UtKrnfrbn5EysBNJ6dT5wEdt71TixV2ToNbQrJIlvbV4onZS\nqjQM3KskgULi/tsBa7ZDOOX4B7pp0jKnaH2Xu0g6XumbAElCW7x4+5r8iJXKO7QNCbVMsH3roFx4\n5SXRcy/0QNIMtLavBX4PvEkp+1oRWAv4BhGN+IjtMx0d4ZNJjGknsupmKLqpOg2LpNdIOpA0wfig\n7ZttTyNG/CDb+5HJyullYD2MlDUB0zW4N7J98kDex1CnZbwWV8p5GgEW2X4cuJS4Pt9WDrmGGLXR\nth+wfa5L1ye9WE51yNLPpGUr4Guk5/tqJFQwgnQHa3TxryUhppVIIt8MdNOkZU4haZykw0ly7NXA\niYqozx3k2W1Wdr2PrMKxPcX28bYf6cVJTzdRk9heRspAO4a48N5IBto7yGp0EWCK7XVheinZOo4o\ny40eon1yJS0JvMH25SSBajEnC/4hUuY2liL6oMhSjgKekfRmcu+LldXilLJPO9v84YG/o6FD8ywk\nvQ/4hftkTQ8iSX5HKvXyz7ae2fmK2M++JZ9iS9JG9ZnO89IF2fuziEu/ksigPkM8VufavqA8jy0l\nbQGMJxPgY4fipHegkTQv8cK8C/iA7TvLAuILJOdmbuCLkj5AQlcf7zi+p5L7upFqwF9GirvyiyQR\n5D0kgevtpN55f+DAMuiuDOwIHF8+miFnvJUmBaOB1Uls9V6ijLaFpOtIFutJRCt5HaJodVeJv25D\nYm47257YPm+3rAgHgmK8FyBa0gcB5yk63SvbfmOzX+MGlzSX7edIDsU44rFZ3fb9necdqHuYHdqT\nuZKctzZwvdOPezyRC55KPFZTiofrPCLOsgkpgbqtHN81IjSzi/qREW7GEUnnAUsQj9+dtg9W+gi8\ny/Z3Jd0MLG37x53n7Zb3ppepBnwO0B542pQB+fXAFWU1fiiJ877P9iRJryJShIsBW9r+/cBe+Uuj\nxOy3J4Iq55fB9QAiSbksuYejyCBxC/BWSVNs30JWTa8BHm2SiPobcHqd1srzUUVB7L1loH0AWFTR\nm36CJG7dDhxO1PewfbOk35De73O3zzcY9/KfUr6XRUhW9AfJxORwSfMAdxOZ4H1s36H0vD8a+LLt\n64kISVsJrFeMd7tcc1Hbfys/jSTvx3UkT2JZSW8qMe2vA3tL+n6Jf08tx9fvssuo8Y3ZRP00pW+t\nkEaSpgjzK1KPfyaD8IclbWL7ZNsTnVaFQ8p4N/cA4NSA/g/wQHH1H0xcbo/afs6RXbyOrJDOJG70\njdSne36f7SdaOQF1kOigGK9XlSStNYmQz3aOut5xRAv/JqIFv2kJXbTV1M4C5iEDd1esnmYSX92b\nJHweZvskMnHcnHxHvweOLROcC4hS4fWt8/WcElh5B1aX9GMy2TlW0hgnMXRUeRaXEe9Zk0PxbZKg\n9mTHuep32WVUAz6blJX13JJ2kLRB2dbEJ18A7iIurJ3KIa8kIhJDNt5b4tyNUlrjpRlFYmc72r6T\nZJovqGQDA3yVlLlNI4lVl3auguoA0Yf6MsnbfBqY3/aWZDK0WVk1nWH7GNvfBxYnpYUPwfSyPdn+\ng1PfffeA3cRs0I6vStpO0keVSoSjgD8Tr8O8tq8kk5JNbR9Amq78EHi77aPa5+wFw928N61FwgIk\nTHcIUUnbikyAGolUiifsHuDVxcOB7antSXqlO6kGfDZRMoN/RVybu0s6sHxUDZeS+PDbFe3lscAu\ntq8Z+Kt9yWwEXAEzCGI8SYRlxkh6J8kIXoHEw5ckterXl2MucElSq7yYDrfnYq2fRpMscpxOWVOA\nz0gaI2lVSTeRbPP93EpS60bDVVaOb5G0H2kRuwrRRXieiIusQJlEkmfyWDnudttn235Y0sheM0Kt\nSXAjsPIWYpxfSybVJ5NvE5hhovgt4ABHh6E5V9e9N5UZqd3I/g1mkiyyG3AbMXjXEM3lz5eVaHu/\nscAr3CVN7hUN6bNsf7VJliqDwUTgVaTT1ftILPZSskLf0/ZDg3bRXYSkFciq6VkS0z6EZAzPCxxX\nDNT6xKhtDPwOWNx2IwHaVUland9O8dzcSTqBfbhsO5wYpM1J/LvR596QuHxvah3fdTH+/4QyQWl7\nK9YnDY1uAE4lCnT3EBGbbV0qOSSt76g5dp6vq96byqypK/CXQDt2K2kJSe9VGnNAXJr7k1X4sbZ3\ntD2tc2XgyIV2hfEu7EpKk+Ypxnt0GYAvIbHWrWyfRbwLh9re3vZDM4lr9jSd7vKSgHUE0QH4PDHQ\nnycD8irAjpLGk9X2VcDjTk13Vxpv6Fs5Slq/JFs9BEwCxkuaqxjkvUhXumVITP9+4De2X9c23uV8\nvWC8Rzj8U6nnXoV0JfwsMeDbkvDc4aSpyCOSlpN0CfDOkvw3A9323lRmTR1sXwKtwed9ZKX9CeCM\nEk/6I4lnb2b7DEmjJU0ihr1rsf1rUp7z9fJ347K9g2hPLyFpPtt7274XutOwDASt92fxMrG7A/gY\naQRxPnmntiHVCF8iHo7vEY38TRwhoPb5hvwzlvR+pea4+XuCpCnE6Bwn6fW2JwHzEY32xiDfQNzl\nl5Lval6lYUnPUQz3aKVt8BWkMcsqjgbDhcSDsTXJG3iQJJB+C/iR7X1dep5Xhi/Vhd4PajWlLwPu\naPJx/JOUrfxe0tFkgP0MMXKjSVefDxJN710dpayupbg57wbG2/6HpMlEp32S7V+29usJd+ZLpfN5\nSHoXydy/GXiV7a3Ke/Ut4BzbF0n6PrCg+4R9pncL67aJkaQ1ierXJFLPvgiwF/E6TCPZ5OeScMzG\nwGTgBOJxaLw7j0jajKiqHe1UQgxrOv8/S1qaJI4uZXszSW8joYW3OuWG6wMfBi6yfaGkhYAnGsPd\nbe9N5d+nrsA7aLJji/Fejhivp0mCSNMoAuBAImqyIil1uYAY9M/b3rbbjTdAcXNOBu4uCVTLkAS8\nX0JfJmw13n2oJVcqaT711dB/mhixCZK2Bv6LvC93KloBdxPFuqXKOR5UoRsGYfVVK0A8C/eSb2Ol\n4qHZnbSK/SmJ4S4EbOGUIN5AvqXDbb/bRYHO9g9sH9gjxrsd5166bH4AeIRkj89r+2oiF/uV8vtN\npIb7tSVM87Dtp1shvyH/3lRmj7oCB0qs6JnWwDuGZHIuR2ooF7S9S1klXUBUxp6V9Gngi7bHzezc\nw4Fy3ye4qDXVmf2sKR6cA4ioyrUkm/oZInN6TnEdo4izvJbENXe1ffagXPBsIunjpJRpG/f1kf4a\n0fS/hTyLJ8nqcX/btyr97scCnySllecAq9p+TH29vHsKSUsQvYVFgKfJhO850qDlZtunKRUf15Gw\nw3WS5u+FCU6lf3p+BS5pDzJ4TCiGG+LK+63tNYh+9MYl3n0KMIGsxHE6Pk0qC6Vh+yxtb277x819\nVuM9c4qb80RgMdt7k9XlriS+vY6jwLegpLUcWdl9gDc2xrsz4a1LWBQYAxwkqdE7uIusFN9EmtQ8\nS7xX20lah2Tb/wB4yPZvSZb91tBXvzycmcn/588Bd9h+O2n88wViwH8LrCZpvCMGdQpJXmtElmYQ\nXqr0DsPW6PwrJK1R3MIrk4HmAfKxQNp+ri7pQlLW8m7bf7d9EYnhfVJpooDtE5tM0YG/i4GjnRE7\n2NcyxHmcKKn9DaAMuBeRVebCJT7c9EOfy/adLjXNZf+uEbtpGY0fkKzxG4EPSlqXaAmMICGYD5fV\n5ZdJ6OBQ4Cjbkxt3OfAh28cP6A0MAurrz90kNi7djCVkUnNz+f0I8vzWIMI1I0i/d2zvb/uC9nlr\nGKs36VkXuqTDgFttn9HPb+uRJJvtbF9Ytm1I6i2fAxZpXIWV3kMz0YxuktckfQlYkuh2PyRpPuJG\nXoII/nzD9ncG9qpnn5ndd/ltRxLXFll9HwOcX0JP5wE3OY00FnB6TjfH9aRHR9KbgNOJtPLvgV1I\nNvntJET3WMmV2NT2B5SmR3+w/afWOWryaI/TkytwSXMTpaf2x/AKSUdKmkh0p79DpCxfV7KvjwYW\ntn1PNd69SStpr1k9TSgxyYbme/ofUs+8donnPmF7TxLnXrcx3t0UdikhpG9KekXH9mYVfjHxMjwH\nfB84jcRqIVrujxaD82g5rmcSrdQnfzpS0iilLGwnUha2GZnYTSTVLBsTD984ood/CYDtq9rGu2yr\nxrvH6ZoBZA4zlpSr3AUx3qTcZQ3SXesNRFf4MSKSYGBtpza60qO0khw3kPQTErM8VenH3Qj9jCxu\n4e8CO5PYd0OjktU1xkvpKY0jwfkIiXdPp3kmtv9CEvZeC6xhe1+SW4LtK20f1zY43RQq+E/pnPAB\nI0t8/xVkrPldiWHvRQz6o8SoL01KDO+0fdpAX3elexjWLnRJry0lLO1tjZvzHNJNa/syoL7C9uOS\n9gZG2D6k7D+PqyBCT1IG4BGtFfdIIh/7VWAH25crLWJfAM60fXvbJVyyzI90X4vHrkHSokQZ7m7g\nBtunS/pf98mejnYR92myxoux34usxCd13nevunwlfYxM5i4nIj7fJUI1h9LXavhIYHnbG5djXulS\nitqrz63yrxm2K3Cl1/YpSoejGX4q/94NeF+JLc1TjPfCpHb1iWbnarx7k2bQLIPr/GVAfYFkSz8D\nvLnseibwapIlPLejnjUKwGkV23XGu9Ak4u0HfE4RCfmLpDdI+hxwhKTdYXrDm5G2nyKG6eL+7nu4\nGyFJq0iaJGltSXOVbRuQSd/HiMfvIOIBPJeo0i1SDj8EeFzS2PIsH5c0ohrvyqwYVga8xJjGK6IH\nDxM1qBHtEotmgLX9VyJbuQNwvlK3ehVwre3Jg3IDlSFDy13+BdJl7SBJ+9n+A1EUW78Y9VtJA5sN\nyeRvhjKobopz98MNpH77YqKI9koi0LIUUVlbpuSMQMJM2L7G9s8G4VoHjTLBOxU4iYQZliRucshE\n6DJgC7IK/5Kja38CEfLZqlQjPGL7A+XfL0DGqmq8K7OimweXGVDqb/9ABCUOLZvXKh+LlWYAM4j7\n2z6Z1OhOJrrC69k+ZiCvuzI0UBjRse0zJB9iZaJ69SVFNe0sogff1DyfC/yadKWbgW6Ic8+Ey0lo\n4FJgRds/JUb6nSSWb+BK0k2tm+9zTrATKS9d0/ZhwPdamfZ/JC0+xwCrOf0SxikKfZOJ9GlTvtrt\nE77KADNsYuCSNiYxpCNLgtFEYBNSczsReIrMivex/UCvlq9UXkxH3Hop4ClHyvRVpD/1EWTVOQ0Y\nZ3sNSf9NSsM+bPvuwbr2lxul6cgdZfJ7FPmWJgDvBtYh9dvXzuocw5kSLrmYyMBe0dq+EanrnkIW\nFD+1fZKkzYE9SAe/Hw3GNVeGD6P+9S5dw3PEXb4BqaUcRdxXJ5NszkmKStRngX2r8e5tSsLVUrZv\nL2GV0SQOuSHwO0nX2j5BaRixkO33SloAeFjSBNvnSnqkbbyHY7zS9h3l30+XuO4Str8t6XJ3V3vc\nl4uFiHjPVIDyjA4m+vc/JAJRhwInSHovMD9wWNt4D8f3pjIwDCd3zZVECWpn4BHbN5LY5ArA/GVG\nvDildKzSuyi93O8Djm+FVbYFptleiSRv7VH2ez3wQMnKfj+JZy4Life2zzucB+Fy/2OImBGN8VZ3\nSr/+28zsPkuy3mKkK1jDN20vBPwZeLvTx309YKLt9WxfUs5ZmwFVZothY8Btv1DidB+xva+ia748\ncA0Rltik/H3hIF5mZQhQjM81ZDX0qbJ5fuBZSRcRA715a78nyQRxC+ATtg8e8IseZIqhOsgdOuXu\njXrumYnYNEZ9H+ALkhaz/ZztqWX7PEBT8/+s7dvbx1XDXZldhpMLHQDbz5Qs82mS7iLuz+sl3dQ5\n+FR6A0UtbQ/SCezqUhJ1G0lc3FjSWSQZa2fgU7YvK8dtClxle09JZ9ieUraL5I/0VBimZZh6glLN\n8pTtv0tqRGymh0xKiaFsXyXpfOAoSfeRapYvEEXHr3WetxcmPZWBYdiswNuUutRFiETh35ttg3tV\nlUFkHZL78BVJy9v+B1kZLQb8rPx2KnlXFpX0ekkHkthlUxrWGO+RpT68p4x3LyFp0eKJOULSNmXz\nq5t8B/U1H4G+RdDuRP/9KVLzfZztj7hD/rRSmZMMmyz0Nk1SiGbRfKHSW5QBeUmiU74A0ZjeHjgP\n2BvYk9Q5bwasSuqd97b94KBccGXQkLQZsCnx2lxBSuf2JnXemwDjgD82Jaezqmip1S6Vl5PhugJv\nRCWq8a40HAAsA/wEeAfJOH+eNNz4KYnvXudoeH/U9g6llKwnkrQqL+Ili9j0Z6DV1za0Gu/Ky8aw\nNOCVSielKuEKIrrxIeB+YDwRK7mMZJqPL96bBxthlzoJ7En+LRGb/qiGuzIQDEsXeqXSH5IWJMpY\nq9meKul1tu8qSY81R6IyA1XEpjLUqQa80lOU5LQtbS/fz281XlnpF0nfAI4oE77Fq4hNZShQXeiV\nnsL2gaSr1kLtJjflt2q8Ky+i10VsKkOXugKvVCqVf4GkN/ZaHXxl6FMNeKUnqSWGlUql26kGvFKp\nVCqVLqTGwCuVSqVS6UKqAa9UKpVKpQupBrxSqVQqlS6kGvBKpVKpVLqQasArlUqlUulCqgGvVCqV\nSqULqQa8UqlUKpUu5P8Bg9WkhETFcjsAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f05e14ac4d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "list_plots = [\"% Gene detection\", \"Library size\", \\\n",
    "               \"Mean reads per UMI\", \"% Mapped reads\", \"% Genome not gene\"]\n",
    "\n",
    "data_1 = [finite(scores[:, i, 0]) for i in [0, 1, 2]]\n",
    "data_1b = [finite(scores[:, i, 1]) for i in [0, 1, 2]]\n",
    "\n",
    "props = dict(widths=0.2,patch_artist=True, medianprops=dict(color=\"black\"), whis=[5, 95], sym=\"\")\n",
    "fig, ax1 = plt.subplots(figsize=(7, 5))\n",
    "ax1.hlines(y=0, xmin=-0.5, xmax=2.5, color='blue', zorder=1)    \n",
    "ax1.set_ylabel('p-value for correlation with library size QC', color='b')\n",
    "ax1.tick_params('y', colors='b')\n",
    "ax2 = ax1.twinx()\n",
    "\n",
    "bp1 = ax1.boxplot(data_1, positions=[0, 1, 2], **props)\n",
    "bp1b = ax1.boxplot(data_1b, positions=[0.2, 1.2, 2.2], **props)\n",
    "ax1.hlines(y=0, xmin=-0.5, xmax=2.5, color='blue', zorder=1)\n",
    "for patch in bp1['boxes']:\n",
    "    patch.set_facecolor('lightblue')\n",
    "for patch in bp1b['boxes']:\n",
    "    patch.set_facecolor('c')\n",
    "    \n",
    "data_2 = [finite(scores[:, i, 0]) for i in [3, 4]]\n",
    "data_2b = [finite(scores[:, i, 1]) for i in [3, 4]]\n",
    "ax2.hlines(y=0, xmin=2.5, xmax=5, color='green', zorder=1)\n",
    "ax2.set_ylabel('p-value for correlation with alignment QC', color='g')\n",
    "ax2.tick_params('y', colors='g')\n",
    "bp2 = ax2.boxplot(data_2, positions=[3, 4], **props)\n",
    "bp2b = ax2.boxplot(data_2b, positions=[3.2, 4.2], **props)\n",
    "for patch in bp2['boxes']:\n",
    "    patch.set_facecolor('lightgreen')\n",
    "for patch in bp2b['boxes']:\n",
    "    patch.set_facecolor('lime')\n",
    "\n",
    "ax1.set_xlim((-0.5, 5))\n",
    "ax1.set_xticks([0, 1, 2, 3, 4])\n",
    "ax1.set_xticklabels([list_plots[i] for i in [0, 1, 2, 3, 4]], rotation=30)\n",
    "\n",
    "\n",
    "m1, = plt.plot([], [], c='lightblue' , marker='s', markersize=15,\n",
    "              fillstyle='left', linestyle='none')\n",
    "m2, = plt.plot([], [], c='lightgreen' , marker='s', markersize=15,\n",
    "              fillstyle='right', linestyle='none')\n",
    "m3, = plt.plot([], [], c='c' , marker='s', markersize=15,\n",
    "              fillstyle='left', linestyle='none')\n",
    "m4, = plt.plot([], [], c='lime' , marker='s', markersize=15,\n",
    "              fillstyle='right', linestyle='none')\n",
    "\n",
    "#---- Plot Legend ----\n",
    "\n",
    "\n",
    "\n",
    "plt.legend(((m2, m1), (m3, m4)), ('Ber',\"NB\"), numpoints=1, labelspacing=1, fontsize=12, loc=\"lower right\")\n",
    "\n",
    "fig.suptitle(\"Zero distribution vs cell quality metrics (BRAIN-SMALL)\")\n",
    "plt.tight_layout(rect=[0, 0.03, 1, 0.95])\n",
    "plt.savefig(\"9k_zero_qc.pdf\", dpi=300)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running scVI on 3000 genes\n",
      "scVI will run without batch correction\n",
      "Will work on mode numeric for incorporating library size\n",
      "Will work on mode gene for modeling inverse dispersion param\n",
      "Will apply zero inflation\n",
      "1 hidden layers at 128 each for a final 10 latent space\n"
     ]
    }
   ],
   "source": [
    "tf.reset_default_graph()\n",
    "expression = tf.placeholder(tf.float32, (None, expression_train.shape[1]), name='x')\n",
    "kl_scalar = tf.placeholder(tf.float32, (), name='kl_scalar')\n",
    "optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate, epsilon=epsilon)\n",
    "training_phase = tf.placeholder(tf.bool, (), name='training_phase')\n",
    "\n",
    "# getting priors\n",
    "log_library_size = np.log(np.sum(expression_train, axis=1))\n",
    "mean, var = np.mean(log_library_size), np.var(log_library_size)\n",
    "\n",
    "model = scVI.scVIModel(expression=expression, kl_scale=kl_scalar, \\\n",
    "                          optimize_algo=optimizer, phase=training_phase, \\\n",
    "                           library_size_mean=mean, library_size_var=var)\n",
    "\n",
    "# Session creation\n",
    "sess = tf.Session()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "sess.run(tf.global_variables_initializer())\n",
    "result = train_model(model, (expression_train, expression_test), sess, 125)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# Get a simple plot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "indices = np.where(c_train == 1)[0]\n",
    "data = expression_train[indices]\n",
    "qc = qc_train[indices]\n",
    "results = sample_posterior(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAU0AAAFACAYAAAA4bi4aAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzsvXuUnPdZ5/l53mvduquvat3vliLbcWLHjuMkJAFiMBMI\nbGBJYMLADovPnIXAhBNmGRgyEBaWyWYnZ/FkGDwMZzMJM4HMwGJI4hAnk5gkwvElsSNLliy3rEur\n1epr3d/7b/94q1qlVl+qJbVU3fp9zil3V9VbVU+3pa+e33MVpRQajUaj6QzjZhug0Wg06wktmhqN\nRrMKtGhqNBrNKtCiqdFoNKtAi6ZGo9GsAi2aGo1Gswq0aGo0Gs0q0KKp0Wg0q0CLpkaj0awC62Yb\nsFqGhobU7t27b7YZGo1mg/Hss89OKaWGV7pu3Ynm7t27eeaZZ262GRqNZoMhIqc7uU4fzzUajWYV\naNHUaDSaVaBFU6PRaFaBFk2NRqNZBVo0NRqNZhVo0dRoNJpVoEVTo9FoVoEWTY1Go1kFWjQ1Go1m\nFWjR1Gg0mlWgRVOj0WhWgRZNjUajWQXrbmCHRqO5MRwbL/H4kQnG5hps68vy0J0jHNpSvNlm3XS0\naGo0miv43AtjPPLlV4gSxUDeJghjHn2yzsNv23PLC6cWTY1GcxnHxks88pVXQGAgb+NHCScuVjmw\nqcDjRyauu2iuN492TWOaIvKQiBwXkZMi8uuLPP9xEflO83ZCRObW0h6NRrMyjx+ZIIwTejMWIkLG\nNnEtgwtlj7G5xnX9rGPjJR598hSlRsiWYoZSI+TRJ09xbLx0XT/nerJmnqaImMAngAeBc8DTIvKY\nUupo6xql1Afbrv8AcPda2aPRaDpjbK7BYN7BjxIytgmAaxnM1EIe2Jedv+56eIiPH5mgmLUpZm2A\n+a9r4dFeL9bS03wjcFIpNaqUCoDPAD+6zPU/BfzXNbRHo9F0wLa+LFuKGfwowQtjlFKUvQjLEB66\ncwS43EOseQH/9Vun+Yk/Osy7/vDv+dwLYx1/1thcg57M5b5bT8a67h7t9WQtY5rbgLNt988B9y92\noYjsAvYAX1lDezQaTQc8dOcIjz5Z5+BIgfGSx3QtwDYNPvD9++a9v5aHeLHc4BuvzGCbBnnX5ELJ\n4w++cByAd921bcXP2taXpdQI5z1MgIoXsa0vu+RrbnYMtFsSQe8D/ptSKl7sSRF5GHgYYOfOnTfS\nLo3mluPQliIPv20Pjx+ZwLZMHtg3dIUwjc012FLM8HdHS9imQcY2UEqRKEXetfjk4TO8665tKwpc\nKtCngNTDrHgRpUbIe+/bDlwpkAdG8jxxbJJi1r4sBnojs/prKZpjwI62+9ubjy3G+4BfXOqNlFKP\nAo8C3Hvvvep6GajRaBbn0JbisiLU8hBrfkTeSeOecaJwLYMe12Si7M0f4ZcTuHaBbgnje+/bzqEt\nxUVf/8iXX+Hg5sJNjYGupWg+DdwmIntIxfJ9wE8vvEhEXgP0A4fX0BaNRnONHBsv8anDp/n22Tn8\nMEapNCnihQkoRSNKcC2T0ak6m4sZPn34NKOTVYI4oTdjs39TnmLWvkzglvNEF0sSRYlivOSxe6gw\nb9eNjoGuWSJIKRUBvwR8ETgG/IVS6kUR+YiIvLvt0vcBn1FKaQ9So+lSjo2X+Ojjx3lqdAbHEAqu\nRRAn2JZB2QupBjEZy8AQ8MMYQfHESxdRSmEKvDpV42+eH+fpU9McbZYTrVRutFiSaCBvM10LLnts\npRjo9WZNY5pKqc8Dn1/w2IcX3P/ttbRBo9FcO48fmWCmFlDIWPNlSEMiAPRmbMpeiB8l5B2TN+0d\nYGzOAwV+lDBVDbAMwTKE0akaZ2cb/MZfvoDAsuVGiyWJNvdmKDfjnovFQG8E3ZII0mg0Xcqx8RJf\nOjrB2FydvGMyWHDJORZu08sME8V779uJ0RRRgNGpGjnXZKoaIECiEqp+BMD2PocXz5eZrYe8Zf8A\ncEkU24/aiyWJTNPgA9+3jxMTtStioDcKLZoajWZRjo2X+PTh0/z9yWn8MMYEgkhxfs5ja18GQwTX\nMunNOlS86DKP0LVMXMtEKUUQKeYaIaYhFFyT/rxL1Y/pz9kcPV9h08FLR+v2o/ZySaJ33ehfRhta\nNDWaW4iVSoBaz794vsS52QYqUfRlLTzLYGwuRJoZ8omSRyFjs2coz3vu2coTxyaBSx7hUMElaaYp\nlFI0whhDhJHetGi+kLHYO5Tjm6/MLHvUXimLfzPQoqnR3CKsVALU/ny5EeJHMefnGmRsk4JrMVRw\nqQdp1tyPEx7cO8D7H9jFoS1F9g4XLvcIf/AAwLyn6pgGvRkT0xD8KOGOrb04lslb9w9SzNq8eL5E\n2YsoZi0ePzIB0HVi2UKLpkZzi7BSn3f781NVn1I9xBAhjBOiWNEIYvpyNnduLXKh4uPH6jKBW0zk\nfu89d112zM8Br99RxLFMSo2Qh9+2B4AzM3W29+foyVg3pWB9NWjR1GhuEVpdPO20Ei/Hxkv83dEL\noKAna1PxIgxJS4vKXprAEYG5esi3z85xz86+y7zVdx4ank/OOKYggB+r+RBASzxb3mgxa8/HJz/+\npRPramiHFk2N5hZhqT5vxxQeffIUjpm2QvphTC2IsE0D20g7fBSKMFaIwD07++aLy4tZm9mazyNf\nfoU37RvEMuCp0RkUcP/e/iu8xsVEcDkx70b0jiCN5hbhoTtHKDVCSo2QRKn57+fqAaOTVSpeyNic\nRyOIcU2DRhBR8UOiBDb1ZHjHgWG29eXYOZi/7H3HSx5RopqxyTJVP2S2HvC1E1OEcTzfBbQU2/qy\nVJrebIsbXbC+GrRoajS3CK0SnmLWZrzkUczavPPQMEfOV1BKMVRwGSo4XCh7eFECCnb259g1kKUR\nxFT8iLt3FK8QuOlawEDeZqrqcWamjlKQtQ0aQcyzp+fwo2hJr/HYeImpisdXXrrIV49fZKLcmBfz\n1hi6bkMfzzWaDchSpUULj8gf/9IJ+nPpcV1E6Ms5zNVDHNNgSzFDzrUoeyGFjMXWYob3P7DrioJz\n2zTY3Jvh5MUaGcsEFHECWSed+H70fIV3HNy0qI2tbP1b9g9w9HyFb74yw1v3D3ZtEgi0aGo0G45O\npgu1GJtrcGhLD985m/Z7u5ZBECdEccJdO4oMFdJYY6LSQRmQepFPnZpGEO7eUeQD37ePJ45NMlX1\nGczbnC/7oBTbChlQitlGtKjXeHk232bTwUsx124VTNCiqdFsOJYrLWp9bXmgjik4lsk9O/s4OVmj\n6kWIUhgifPHF9PrNvRn2DecpuNa8GL/z0Mh8Mfre4QIPDxf4148dZaYasK0v22ydBER46/7BDZEA\naqFFU6PZYCwlRi+eL3Fmpn6ZBzpR9kmUYtdgnvv3DHB6qsaFcoMgiDENwTKFMzN1Jise9+zsZ6SY\nXVSMP/jgAX7n3bfPi2p7h8/PPLBrUTuvZmp7N6BFU6PZYCwlRmUvHRh8bLxM2Qvpzdhs7nUpuBbF\nrM3YXIMLFZ/hvIvKQzWICKKEjG1QzDq8OlNn/0jPZZ/VEuOPf+kEY3MNcrZBEMWMl1Lxu293H48f\nmeA/ff3VK9o2Fw7kODNd4/hEle39WT7+pRNdu8pXZ881mg3GUqVFhsBLFyp4YUyPa+GFMS9dqHCx\n6vPBBw/wsf/5dewcyCEG9OVsdvTn2DdcYPdgHtMQBLkic356qsa52QanJqucma7x1KlZnjs9x/ce\nTFdkPHFscsl5me3Z/JculDl+Id2tfmhLb1ev8tWiqdFsMBYrLXr4bXtIFPN7zNu/lhqXhHBbXxbX\nMvGjZP6x2XrATC3ACyP+4ZVpTk1W58X4xESVzT0OJy5W8aOEgbwNAo985RU+dfj0fGzVEJn/vr1m\n89CWIh988AC3bynypn2D9GQtvnVqhqdOTTM6WeXTh0/f0N9dJ+jjuUazAVms+6Y3Y1Guh3hhTJwk\nTFZ8qn6EZRr8wn9+mtu3FDkwkufImMupqVo6nSiIGC/7FLM29+8doO7HHJ+oUg9j7thaZMdgllI9\nxLUMEqUYm2vghTGJgsOj0/zwXVsvs2GpRM/YXAPbhG+fKTX3DKWe8N+fnObYeKmrjulaNDWaW4Q7\nthbJ2SZHx8ucnW3Mb4+0DIMz03VytsmZmTrvuWcr3zw5zbfPlphtRGwtZrlvT39aftQMaY6Xfcbm\nGpQaUdObtRgv+fMT2pWCmWrAmenaZft8lkr0bOvL8tXjF3EtY34yvIjQn7O7rgddi6ZGs87pdA/4\nQ3eO8NHH55iqBRRck0YQEyWACIlSXCj7HNrSy4mJGr/3nrsA+NBnn2dLMTM/lX2q6vHShQpxAvfv\nGSAIY14aL1NuhLh2Gu2LEhjMOwA8f3aOV6frBFGCYxkM5B3+xUMHF7Xtr749Rn/WSvvfowQ/Snj9\njmLXlSAtGdMUkYqIlBe5VUSkfCON1Gg0i7PScrKFTFV86n5IPYiJVVo2lLGEmh9T9sIrjs8L+8JP\nXqwhIgwUHAwR9gwXuHtHES+MCSOFYaSCaRjCnqFc2o4JKNKBxO0rMdo5tKXIW/cPgghVP8a109rR\njG11XQnSkp6mUqpnqec0Gk13sNKMzBYtcW2ECYN5lyBW1IOIOFHYtkEjjNmRyV1xfF5YFjRV9bEN\ng/3Dl4Z2ZGwTBGpBhBcJOdvk3t2DvHi+zEhv5rIWylIjXPK4/TPNFs2FdZ43cmlaJyznaQ4sd7uR\nRmo0msVZbM3tYsmWTzV3kM81gvk2ScdMxbIeJBgCm3vdKwZlLMzEDxZcDowUGO5Ji+dPXCjzjVem\nyVomBTfdCzRZCRibrTNbD7l965V1nUsdt5fK+ndTPBOWj2k+CyhgMX9aAXvXxCKNRtMxixWyn5mu\nMV72+dBnn2dbX5YDI3m+fnKa/qzFSI/LWMkjUarZZw6mIdy2Kc/uocK8YLaK1Vsx0g8+mK6v+NwL\nYzzylVd4cbzMYN7hlckqgrBjMAfATC2g4kW8PFnjew8MY1tpUmeq6nHyYi3tTy+4S2bEu3En0EKW\nO57vuZGGaDSa1bNYV81zZ+a4e8elyeqPfPkVspYBIhQyFttFmKz4BHHCa7f38Tvvvn1eqI6Nl/jY\nF08wVfXxo5iXJyocGSvxoebOnyeOTXJgU4ELZY+ZWkjZi9jZnyHnpFKScyySJEn3ngNfeekiWdvA\nC2OyjoVtGGzucbt6ncVKiGpujFvyAhEB/jGwRyn1uyKyE9islPrWjTBwIffee6965plnbsZHazRd\nSXv2/MxMnc09LnuG0zKfyYrHF45cIEkSLNNkMO/Ql0unrU9UAu7fO8CmgosCgljx0niZi+UGIgZ+\nFBPGCUmi2L+ph7u2F7Et8zKv9jPfOkOcKPZtulRWdHamxmw9ZFtfjowtjM02iBLFcE+6Lz1RYJvC\nHVt7+f1mlr4bEJFnlVL3rnRdJyVH/x5IgO8DfheoAP8duO+aLNRoNNeF9iNtq0QIUsF87swcjin4\nShgqOExVA6p+SD1M2NKbYajgcHh0BgHu29PP6GSVRphgGemUorR9El6drlHyIt6yfwC4JJqv31Hk\n709Op2t4XZOpasBkJWC44DCQt/GjhBjoy1qUGhE9GZucZeCHMV/vwsL1TuikjfJ+pdQvAh6AUmoW\ncNbUKo1Gc1W0lwidnKzNF4v7UcLFio8p4IcJBcfEtgy+dPQiNT/CNIQjY2W8KEGAWKWL1BKlml5o\ngh/GPH5kgsOj00xWPCYrHhfKPv1Zi+mqz7ELFS6UPSwD+gvOfKtm1jKYrqd98K3WTdoK19cbnXia\noYik45gBERkm9Tw1Gk2X0R7jrDRC4iRhth4yXHCJEkUjjKl4IY5lYBqCUgoEpqo+iQLLEIJYgWrO\nwwRAYQg4llAPEsqNkG+enCJKwLEM7tpe5PhEDdc0qAYR9SDmpfEKWdukmLMpuCbTtZCCI5cVrt+9\ns/sK1zuhE0/zD4G/AjaJyO8BXwd+f02t0mg0V0V72Q4Cc42Qzb0Zepv3If1SD2IytonRLCav+hFV\nP5xfv7sQU6Dix2zqcYnihDOzDc7PNah4IS9frGEaUA1jGmHcbM8kLWfyIyarIRkrXaVR8SMytskb\ndvXhWt1XuN4JK3qaSqk/E5Fnge8n/X3/mFLq2JpbptForoltfdnmAI2IqWqMIYIhaW94PYi4UGoQ\nNJM9LaH0orRmU5F6mkL6vdl8MIgTSo0IyxAMSa95dbpOj2viWAYNwDAMXAPCOKERJmRsk7t29DHc\nm+36wvVOWNHTFJE3AWNKqU8opf4dMCYi96+9aRqNZrW0t1Ue2tLLSI/LRCXAjxK8KCZKEgzDwDEN\npmsBhmHQl7PJOyZ518IQIe+Y2KYxL5gAfqSwDbhYCdJ96KZgGkLGTr8v+xGotE2yN2NhGoJtpiVO\nP/TaETb35dZF4XondBLT/CPgnrb71UUe02g0XcDCtso37hng3OwYUZzQl7NTDzJJ+86Vgv3DeYI4\njTPuHcrx9KuzzNaDtCzIACVCHCsSYK4eEipQSdDc/wMTZY8kSZMctSCm4FqEzY6jRCnm6gF/+/w4\nu5q70ltF8uuZTmKaotqKOZVSCXo6kkbTlSxsqxwqZCg4JojgRQmNIMYyDRzLQCUJM/WQjG2ydyjH\n6FQd25C0DVBBkEDYFEyAsKkCQZIey6MkvbWeT1Ra01n1IoJYETafrwUx52frfPTx4105iX21dCJ+\noyLyy6TeJcD/BoyunUkajaYTFhsJt1hbpeuYuLFCRMg6RjM2CTnXZudAjl2DeY6eL1FuBEzVAlQC\npgHxMjUyC5+yjXRYiGEYuFFCLYgR0my8iKIWxozNNrpuNubV0Imn+c+ANwNjwDngfuDhTt5cRB4S\nkeMiclJEfn2Ja35SRI6KyIsi8l86NVyjuZVZaiTcgZH8FfuBCq5Nf87GMYUkrSZqJnEUkxWfMIo5\nO9tgth6StU0sI/UQO0VIr79YCSg3QmxTcE0h75pkbAPHMkHBTD24osTo2Hi6lO1Dn32ej3/pxLrw\nRDvJnl8E3rfaN27Wdn4CeJBUbJ8WkceUUkfbrrkN+JfAW5RSsyKyafF302g07Sw1Eu7ERI13Hhrm\nk4fPMFH2GOnN8D/dvYU/e+ocm3szxEpxZrpOECsyljA21yBjG0TNOGQcJ4TJpQRQJ7SuFUApRc1P\nMI00ZiqSfjXNNDbaXmLUEv72lcLroSd9LWOTbwROKqVGAUTkM8CPAkfbrvkF4BPNLqOWQGs0mhVY\nuNt8suJx8mKVs7MNihmbAyMF7t8zQMWLeOlCjTu39jA253FuJhXMgmuiVFqv+ezpufn3WY1YLkQp\nyDsmfhTiRyDEWGa6tsIyDIpZ67Kxc53OAu021lI0twFn2+63jvbtHAAQkW8AJvDbSqnHF76RiDxM\nMySwc+fONTFWo1lPtMcuWz3mfhgzV/OZrvqcmqoy3OPSl3NohGkh++beDGdm6hQzJl6kqPpRW9fP\ntSGAawmxErb0uoyXfESEMFZYhiIRhWUa/N9/d4I7thZ56M6RK4Qflp+32S3c7Cy4BdwGvAPYDjwp\nIq9VSs21X6SUehR4FNIpRzfaSI2m22hvlzx5sYofxoyXGnjRpb8e5+bS/vB9mwr4YVoCpJTCC6Ea\nRNfmVrYhpJsugzhd4ZtzLHpzLl4QcXa2AShEKaaqPpMVnzPTdV44N8eWYoaKFxFEMScna1S9aH76\nUTezomiKiAv8OLC7/Xql1EdWeOkYsKPt/vbmY+2cA55SSoXAKRE5QSqiT69ouUZzC9Nql3z8yAQT\nFZ+aFxIu4jb6MRwbr5JxTMI4BqDeXLG7+Lae1aMAMYQ+xyLXHA6yZzDHt16dTdcGeyGRgjiMydkm\nc42AOFFkLYNGlPDyhQphkhDECqXAtYyunn7USfb8r0ljkRFQa7utxNPAbSKyR0Qc0mTSYwuu+f9I\nvUxEZIj0uK7LmTSaDji0pcgHHzzAj71+G7GCZImMd1p3mXBmpkHZi4ia4nq9jmyWkY5+CxNohGmR\n/AtjJSpeusAtjNNMfRQrqkFEECUUXJNTM3WylkE1jAliRdY2GOl1ma4GfOrw6etk3fWnk+P5dqXU\nQ6t9Y6VUJCK/BHyRNF75p0qpF0XkI8AzSqnHms/9gIgcBWLg15RS06v9LI3mVqO9RrPuhVS9aFkR\n9CKFSFpqBFyXWKbZ7D13LJMfed02zkzXeOrUDM+fLVEP43RiUpQ0u5Au1X76UUI9iLGtVDh39WfJ\nOJekyAsivn12bukPvsl0IprfbMYZv7vaN1dKfR74/ILHPtz2vQJ+tXnTaDQd0F6qY5vw0kQVy5BF\nj+ctEpWKnHD9juWxSo+qpsB01Wf3UIFj4xVKjRDTEFzLSD3gKCEh7TIyJD1+T1YD3vXazXz7bOkK\nsU87NK+Xldef5bZRfldEXgDeCjzXLFJ/oe1xjUZzE2gv1RmdrNOTsRjucVd8nVKpp2cusXv8ajAN\nGMg7PHdmjsmKh9G8/wO3jzBSzOJaBo51SagzVtrCmXEM3v/ALu7eUaTqx3jNkXJeGFP1Y+7e0Z3x\nTFje0/zhG2aFRqPpmPZSnbIX0uNahEly2VSixWiFPKNErXhtJxiAbQibi1m8MG5Oik+3Tw73ZLhn\nZx9lL6TqQSEjbClmiRMuy5AroB6kszwtQ4gShSFp/3u3JoOW9DSVUqeVUqeBLcBM2/1ZYPONMlCj\n0VxO+0qL3ky6h6fcCDsWwYSld3N3ikl61K5Hiu+eS73M87MNhgouA3lnvo0z75gEcYJKoOKlLZZD\nBZe37B/k0SdP4Vgm7zg4TF/WYa4R0Z9zeMfBYRzL5NEnT3VlW2Un2fM/Ih0H16I1Gk6j0dwEHrpz\nZL6/fO9wjooX4YXJqouur9bTFNKsbask1AsTSo2QRhjznnu28hNv2MbTr87wN8+PM10NyNvp+uDZ\nWkgjiEmU4hsnp+dDDCO9WfryDjsHcvTlHEaaw4qL2e7cIdTJ7/mK0XAicrOL4jWaDc1iE4yA+cey\ntsHFcoNXpxsomjt9bpBtC8U2AepBwqEtWb55cpp6mGAZwv7hPBfKPn6UsKM/gyGCa5vsGszz1Klp\n3nnoUktl1YvocU3KXjj/WLd2B+nRcBpNl/G5F8Z45MuvECWKgbxNEMZ89PE5DBF2DOTYUsxweqrG\nd8fK3LOzj52Def786TNMV4KbZrMCjp4vM17yefD2zYTN/vZYpUvZTk2lmzFFhH3DOQSh4kXz/eaF\njEV5wUi7ihd15Q6hTkTzn5EuV/tXpL+bL9PhaDiNRrM6jo2XeOQrr4Awvzf8xMUqcZKQcyxGel2e\nOlVmdLKKF0T83dEJHMvAQIhvot2ttb8ztYCejEUhY+GH6UzN1mzNnCMIwlOjs7xmc4FSI/UqezIW\nm3tcxucaHBwpkCjV1TuE1mw0nEajWT2PH5kgjBMG85f2hgOcnm6QKI9T0zWylkHNC+f7zF1LqPrR\nzTQbRTrl3RDF6aka+4fzPHdmjjhRJEmCZQpxAgN5iySBvpzDzzywaz7csGe4wA/eOcKJidp8SOK9\n923vyux5J73nGeDngTuA+ZEkSql/uoZ2aTS3JGNzDQbzDn6UzAtmFKcdNKYh9GRMQOYF0zSg6sfL\nTlm/UZiSZvO/fXaOu3f08fodRb54xMOxTPKuSd6x6M+77B3KEcSKQ1uKV4jiu26S7auhk+P5p4CX\ngB8EPgL8Y0Cv8NVo1oBtfVnCKOb4RFqw4loGE2Uf1zLSBExzoG/zW1op2m4Y/aWA/rzD9r4Mo1M1\nYqUwTWF7b5Y37OpnuCf1uRau41hvdFJytF8p9VtATSn1SdJ/DPQKX41mDXjozhEMw+DgSAHXStfs\nhonitdt6ydomZS9ith6k7YvN18h17PC5Fjb3uuRdi1dnGkxWfe7fM8iDt28ijBVPjc4wUW5warLK\nP7wyzdHmmovF6jC7fQVGJ6LZqgGYE5E7gSKg11JoNGtAa+Tb7qEChiE0wphGEPIPp2apNMtxHNPA\nstLET9pTruYF9GbiR4pEKS6UG5TqIUfHywjCA/sGMAz4wpELfO3lKZRSDBWc+fUW7aK41O6jbhLO\nTo7nj4pIP/BbpKPdCsCHl3+JRqNZicVqMVtxvtHJKn/53DlcM/VrlIIgUmQsaIQxYazmpwz5NzNt\n3sZMLcALY6JYkXdN/DDmuTNz7BnMYYqBIOwayIII3z5T4g27+uYL2FuxzfWwAmNFT1Mp9SdKqVml\n1NeUUnuVUpuUUv/hRhin0WxUVvKo/sNXR2kEMRfKHnGSZsgNAS9WxM1pRrHqjlhmi1il8zQtw8AE\nMraJaxk8P1YCActMi9tbj5+8WLuigH3h3nboviL3FUVTREZE5D+JyBea928XkZ9fe9M0mo1Lu0c1\nXfU5Ol7mu+dK/OvHjvK5F8YYnaohkkqiSLOcp/m3VXWZWLZjSjqQw48TZmo+jilUGhFKKbYUM/jN\n3cCuZVD2wisK2Nv76lt0W5F7JzHN/5d0WPDW5v0TwD9fK4M0mluBlkfVvhStP2cxXfX52BdPECcJ\npUZE3Jz6AxDEzX3lN9n25YgTRcY2MUQYL3mMlz16shav2dzDa7cV8aMEL0xHwTmmQakRXrahsr2v\nvrW3feE1N5tORHNIKfUXNP9fKaUiuKnNBxrNuqflUaXj1AwytkkQK3K2yWTFx2mOWE8SRRAr4qZr\nabZNX+82Wl1BfpTgmEIYJ8zUQvYO5Sh7EY5l8vrmnMy5RsSdW3uv2HHeSoQVszbjJY9i1u66Peid\nJIJqIjJI80QgIm8CuieVpdGsQ1rbJGeqAY4JF0oeXhRjIIRxjEJQShEtPIeLgWOCF3WXv9kScgMI\n4wQvVFimsK2YoeLFDBUsgigmSuAdBzfNJ70WY7Gi926iE9H8VdKs+b7mfvJh4CfW1CqNZoPT8qg+\n9NkXOD1dI++YDOUdzszUgfSY25JFg7YBwnGC6rKA5nDBoepH+GG6DyiIE0wR8o6FYxsEccKOgRzF\nrM0HHzxws829ZjrpPX9ORN4OHCT1wI83V+5qNJpr4NCWIq/fXqQRxPRkLM7PNVCkKynadbHdp4y7\nTDAhnR7yv3YjAAAgAElEQVSv2kVeQW/WwhDh/KzHtv5s12XAr4WO5mI245gvrrEtGs2Gp7020zGF\nfxidJlGK83MNZmrpaLcu1MVl8aN0fYZJKvCt+lGTSxPiz0zXGC/7fOizz19Wk7oe6SQRpNForgPt\ntZmWAU+NzjBTC3FMgzhRKJXWMmasLs30LIFlpDFNJbCt6NLjWtT8CFBs63OZawQ8d2aOzT1u13b5\nrAYtmhrNDeKyLZJTdQoZi+Eeh+laSNVP95b7kcK/IvvT3SiV3hzTIJ+x2T/Sw5a+DLuGCvhRGlK4\ne0cfe4YLGCJdvcqiEzo6novINmBX+/VKqSfXyiiNZiPSvkVyquKl7ZCRwgsjolgh0t2F60thmQam\nAZYIp6ZqOJZBb8YmZwllU0h8xYVmveZQIf3513OMs5N5mv8GeC9wlEv1mQrQoqnRrALHFJ48MUnF\nCxkveRikR9ogSpMoosBqCud6KYQ2gCRJMBDENnBNgzBOiJOE586UuHtnkd6MRdmLePb0HG/Y1cdQ\nIdN1XT6roRNP88eAg0opf62N0WjWI0sN3lh4zUTZp+JFVL0IlSTU2pSxtYfcECHstpqiZVCkLZGx\nSgvxe7I2rm2wqSeNXTYixf5NBZ47MwfAyxNVbNPs2lUWndBJTHMUWL8TQzWaNaTTUWaPH5lgx0CO\nB/YN4McJ3gJXsiWTUXK5YHZjSki4ZFfeNejN2sSJIowTbFPY1JOh7IX0uCZVL2K4J8M9O/vozVhM\nVPyu7PJZDZ14mnXgOyLyZWDe21RK/fKaWaXRrBM6HWXWimcaYlNwLSpelB5tF6zeXdjn020+Z8vL\nUjRLi5I0eeVYQhApxksed2ztBdIJ7b3N38dwTwbHMnlgAxS4dyKajzVvGo1mAe3JnRaLJTm29WXn\n1zy4loHAlS2S6wTTEOJEsbnHIZ9xqAYRUWIAMUopnjo1w1DeZc4LObCp+7dLrpZOOoI+eSMM0WjW\nI+1i2GKxJEer1xxgqOBybrZOvF6yPVyKuQLsGshy20gPJy5W6cvZ9Ocdan7Eq9M1EqWIEkXGMdmV\nsylkLMZLXldvl1wtS4qmiPyFUuonReS7LHJKUErdtaaWaTTrgHYx7MlYS3pUrV7zTx0+zdmZGsE6\nEkxIC9h7sw5v2TfInuECB0byfPivjzJV9ck7Fl4YYxkGm3pc+vMOb9o7OP+PyXo/ji9kOU/zV5pf\nf/hGGKLR3Gw6yYIvpCWG7a9bzqNqhAmOZV7mua0X3ri7n11DeV48X+LMTJ3dg1lePF+hHvjEiWK4\nYGMYwv5NeWB912Iux5KiqZQab349fePM0WhuDq0seDFrX5YF7yTLu9Ios5YY/93RC1S9kLE5b90J\npmXAdD3k5e9eoB5G3DacpxYk7BrMUfFCpqoBc42IN+zKzRewr+dazOVY0zZKEXlIRI6LyEkR+fVF\nnv85EZkUke80b//rWtqj0SxFexb8erb6tZckzdV8xuY8kvWmmEAjgtGLVcI4xhQ4cr5CnCT05Rx2\nDOS5baSAbRjNn687J65fLzpqo7waRMQEPgE8CJwDnhaRx5RSRxdc+udKqV9aKzs0mk7oNAu+Wj59\n+DRHxua4WPEpN6J152G2ENKsuSFCb8am1KhT82MG0pM4lmGwazBHGKsNl/hZyKpEs7nKd4dS6oUO\nLn8jcFIpNdp87WeAHyVtx9RouopOs+Cdcmy8xKcPn+a/P3euOR6tG8vUO8exhOEed77cyDKEWpAu\nTPOjBD9KODhSYPdQYcMlfhbSSe/5V4F3N699FrgoIt9QSv3qCi/dBpxtu38OuH+R635cRN5GurDt\ng0qpswsvEJGHgYcBdu7cuZLJGs2q6TQLvpDPvTDGJw+fYaLsMdKb4Wcf2Mne4QKPPnmK0ckqhgix\nUgRxsm69TIAwUrx0oYxrGtiWQdY2mKnHvHyxytZihoMjBQzD2JDH8YV0EtMsKqXKwHuA/6yUuh94\n53X6/L8BdjfLl74ELFoTqpR6VCl1r1Lq3uHh4ev00RrNJa5modfnXhjjD75wnHIjZFPBodwI+YMv\nHOeRJ16mmLUJ4gTHMkgStS7jmO0kpBPl62FCzY+o+DFv2Flk50COih9TcK113Rq5Gjo5nlsisgX4\nSeA3V/HeY8COtvvbm4/No5Sabrv7J8BHV/H+Gs11ZbULvT55+Ax515o/0ltGQiOI+fLxSV63vYgp\ngmnIuhdMuNRrrgDTMNhSzBAr4R0HN82HNW4FwYTOPM2PkO49f0Up9bSI7AVe7uB1TwO3icgeEXGA\n97GgHbMpxi3eDRzrzGyN5uYzUfbocU0Aan6UjnsTBSjKXsRMLaDmh+v6WN5CNW+2AXnXpD/nUPbS\nVWEbtR5zKTppo/ws8Nm2+6PAj3fwukhEfolUcE3gT5VSL4rIR4BnlFKPAb8sIu8GImAG+Lmr+ik0\nmjVkqaL3kd4M5UZIMWswUfbwwpgwTjANwQvj5sqH9VjGvjiGpD9JkqTJn95M6mFv1HrMpRC1wuw+\nETkA/BEwopS6U0TuAt6tlPo/boSBC7n33nvVM888czM+WnML0l703p4gevhtexidrPIHXziOZQjn\nZusYhqAUbClmCKKYqWqIiKI3YzNVW/8LXAXI2gamYTDc43L/3n5cy5r/faz347mIPKuUunel6zo5\nnv9H4F8CIUCz3Oh912aeRrM+WK7o/V13bePXf+ggc/WQqDkiLYoVM7WA2XpEnCgMMRBZ3+VGLYzm\nj9GbtXhg7wBhzLqfjXk1dJIIyimlvrXgf3y0RvZoNF3FUkXvR8dL/MZfvsDXjk8y54UY0txJLlD1\n02kcCgiihMlqcOMNv44IkHMMijmH3oxNb9bm995z687r6UQ0p0RkH83AjIj8BDC+plZpNF3CYkXv\nZ6ZrvHyhwqhRY7YRIM2dPkJ6dGv3KBYOFV6vRLHCkjSo2ZtZs0bCdUEnP/0vAo8CrxGRMeAU8P41\ntUqjuUksTPocGMnzxLFJ4FLR+/GJKo5lknVM4jlwLaERqbTzR9KsZ7xOcz+WgGEwP7pOgIydtk9O\nVn163IjenM2x8dItdSRvp5Ps+SjwThHJA4ZSqrL2Zmk0N57FJh09cWySdx4a5sREjaPjJUqNiLl6\nQJQoNlsutinECViiiBQkCCLrcA9vk0iBpaDgGLxmSw9VP0EpxVTVJ04gVsLmHrfjCVAbkeWGEL9f\nKfVpEfnVBY8DoJT6t2tsm0ZzQ1lq38+JiRoP3TnCmZk62/pyoBTHL1R4+WIV0xD8KMEg3ZljGYIB\nxIlaN7ppNKuiWqGErG2lyZ59w8zUfP7H8UkSBVnHIGub7BkuUGqEV+xBulVYztNszi+h50YYotHc\nbJabdPSpw6cZnawSxgo/jPCjiCQRUAmmQJik8cyMLRza0sdzr84SrJNWoIVmxqrZKulFDBUyFLM2\n2/uy+FGCa6fF/LdaQXs7yw0h/uPmeLeyUurjN9AmjeamsNSkI8cUvn5ymv6sRcE1ma76zalFCj9O\nPbWcJWQcC8c2qHgR0TraXb4QP0gQB87O1GEgR49rUfbS9FZr0+StVtDezrJ1mkqpGPipG2SLRnNT\neejOkfnhue2DdAXoz9kggojgRTGJohnLTI/kXqSYq4dMV3xOTlTWdb+5Etjen2Ok103DFTkbFBzY\nVGCw4G7oAcOd0Elx+zdE5N+JyPeIyD2t25pbptHcYBZOOgqimJxt8PVXpomThFI9wAtj4kSRKEUC\nWGZzihFpTFAEovWsmIAhgh/GfOdsiYfuHOE//pP7+Pj7Xsee4ULHE6A2Mp2UHL2++fUjbY8p4Puu\nvzkaTXdQagScnW5wYKTASI9L2YsQSQfwQuplmgKmAX5bYWacXBkj7HYMAaVSwW/VYRazqWfdniW/\nVUVyIZ2I5s+3pq+3aE460mg2FO0lR6V6CAInLlbZO5Sj7EW4tolCkXcs6kGMIel2yXaNXG/1mQL0\nZS3KjQjDEFzLYLDg4MeKe3YWcSzzls2SL0Unx/P/tshjn13kMY1mXdNeclTxI3ozFq5lMFULecOu\nPkyBV6fq9GYt8k6aRTZljbcTriEGUMyaKCX0ZGyyjslg3qE/73LPzj6GezK3dJZ8KZar03wNcAdQ\nFJH3tD3VC2QWf5VGs35pLznqzdh4YYxrGVSbpTeWabB/U4F3HNzE3z4/xrk5DwDLBC9cf+ssso7B\n2w9swjCM+d3t13NP0kZlueP5QeCHgT7gR9oerwC/sJZGaTQ3g/aSo/2b8jx7eq45NzIdfzZbD3nz\nvgEgTfpkbYPSOt8wWXAt3rx/kMePTHB0vDQfx901lO94T9JiLDWDdCPQyTzNB5RSh2+QPSui52lq\n1oqFszPPTNd44VwJgKofMlsLUSLYzS6g9SqWlgHFjEV/PsOuwRyJUuwazM//zMcnqmzvz3LH1uJV\nid1yM0i7WTg7nafZSSLopIj8BrC7/Xql1D+9evM0mu6jVXLU8pAKrsVwj8uZ6RpT1XSIsFJqPoO+\nHhEgY5kMFjL05R2mqj4Ad23vA2D3UIH+fFqfebWreJdqR90oCaVORPOvgb8HniCdgKXRbFjaS2ve\n/+hhnj9Xwo/SrmyDdTuHYx4FJM2VwvuH8zx3ZvaKnezXmvxZrh11I9DpEOL/fc0t0WhuMMvF3f74\nay/z9dGZy65f77MxW5PXw1hhNVP+s7WQIE44PDrN/uE8wz2Za07+LNWOulESSp2I5t+KyD9SSn1+\nza3RaG4Qx8ZLfOyLJ5iq+vhRzMsTFY6MlfjQD6ZH0k/8j9EV3mH9MZhPN0gKwlQ14KnRGWxTyNgW\n5UbIc6fnODBSwDQN3nvf9qtO5jx05wiPPnkK4LKY5tUklLqRTkrMfoVUOD0RKYtIRUTKa22YRrOW\nfPrwaU5N1QDmtyqemqrx6cOn+dTh01T9jbXRxTXBsQx63NRPaoQxhYzF2w8O89bbhihmbcIk4ULF\n5+G37QHg0SdPUWqE87NFH33yFMfGSyt+1sJ21I3WdtnJEGI9Gk6z4fjmK9PUgoiyF+JaBv05h4Jr\ncnh0mihhQZRv/ZN3bdJCGSFjm+RsA9cy+PaZEoWMxf7hPG/cM8B4yePQliIf/9KJa0rmbOS2yxU9\nTUl5v4j8VvP+DhF549qbptGsDcfGS0zXA5JE4ZhCFCeMlzwaQZS2S1oLUyPrl9ZSNKUUjSAmUoqt\nRRfXNil7EQXXxA9jnjszx+mp2nzccWyuQc+CXUAbKZlzLXRyPP/3wAPATzfvV4FPrJlFGs0a8/iR\nCTb3ZNL1Dc0BG/Ug5NXpBmGcrnfINIftrmcMAdsQejMWCdCXt/lHd27m7p393LU99QL9KMFtZoVO\nTFTnx71t68tS8S4PUWykZM610Ilo3q+U+kXAA1BKzQLOmlql0awhY3MN9g2nRd2z9YCZWkAcKyxT\nMEV4dbqBZcq67Cm324y2BLKOSRjD228b5t2v28bvv+cu/FixczDPG3b1kbHN+T77HYPZ+SP1UrNF\nb9UZmu10kj0PmxPcWyt8h1n/1ReaWxjXFJ6bqLG5mGFsroEXxOlStDgha5uQKMpetC7/kIdNow0g\nn7HJ2iaNMOZC2cNpes+tkqChQoahQlpPubBEaGGh/7a+LO+9b/u8qG7kNsmV6EQ0/xD4K2CTiPwe\n8BPAv1pTqzSaNUTRWk1rkiRJs+A7fcKP4nRv+XqvYgfqQYSQ9sjP1EIe2JcerTstCVoqmbPY1s5b\naTvliicQpdSfAf8C+D+BceDHlFJ6NJxm3RLEivv29BMnCi9K1dEAEKgF69G/TGlPXiVAGCnqQYxr\nm1iGzB+tr7UkqL1N0hCZ//7xIxPX/4fqQjrxNFFKvQS8tMa2aDQ3hNbxNOdaFFwLL4gIFazjXWhA\n6hzbBjS7PkkAwwDXMvnA9+27TBSvpSRoo7dJrkRHoqnRrCdWirc9dOcIH338OC9PVGgE0XwccCMg\ngGMJpiFEseLObX38zrtvv67H5o3eJrkSWjQ1G4qF8bZTk1U++JkL7BjMcvuWdNTZ6GSVV6dq1Pzo\nssTJRtDOMAFHFFEEjm1w23D+un/GRm+TXImORFNERoD7mne/pZS6uHYmaTRXT3u8bbLi8cK5OSp+\nzFTN52LZ54mjFzg356GUwmwGAaX5H1HrP/+jSPcUZW2Dt+4fxLbM656kWSmzvtFZUTRF5CeB/wv4\nKumfr0dE5NeUUovtDtJobirt8bbnz5WY8yIcM90iOVcPODvTQJHuKg/b5mIa0lzBS/duk+zPWdT8\niCBOPWNT0lmNibqUyELBnqG0JbJVTgSrm2XZSTnRRm6TXIlO6nd/E7hPKfWzSql/ArwR+K1O3lxE\nHhKR4yJyUkR+fZnrflxElIisODVZo1mO9k6WibKHYwqCYAhcrPg0K4tIlJrPNivSziBDLglmt7VR\nOmaaqMo5drNvHKLmD+MY0JtNk1pDPTb9OecywVxNkqYV3riaQR23Cp2IprHgOD7dyeuaBfGfAH4I\nuB34KRG5fZHrekgnKT3VkcUazTK0OllenapS82NmawFlLySKU6FsIXK5MCouZZ27DUOgP+dycHMv\nP/vm3XzPbUPkMw6mcek4HiWKu7b1snuwwEwtvOz1q0nS3OrlRJ3QiWg+LiJfFJGfE5GfAz4HdDJb\n843ASaXUqFIqAD4D/Ogi1/0u8G9otmlqNNfCoS1F3nlomOMXqliGzO/yrocR0tRMU8AyjCsSPwtF\ntBswBWwToiRhsuLx5WMTfOOVaZIk3b9ezNr05Ry2FrNM1UJyTlqTebXtj3pQx8p0Utz+a8AfA3c1\nb492OMl9G3C27f655mPziMg9wA6l1Oc6tlijWYETEzXetG+QH3ndFrYUszjNgRQJYBqp5xYllyTT\nJK1v7BahbCdWkCSCH8RMVXxemawShDH1ICaI0u0zAlT81Lu8UPb5wPfvu+rCdT2oY2WWTQQ1j9hP\nKKW+F/jL6/nBImIA/xb4uQ6ufRh4GGDnzp3X0wzNBqSVDDLE5s37Bjk5WeNiqcFkLcAAan58mUAm\ngCnNLEqXYUpad9kIEywTBIPerE3Vj5vxV4VjmdSCmB39Fr1Zm3fdtY13XeXn3erlRJ2wrKeplIqB\nRESuJk02Buxou7+9+ViLHuBO4Ksi8irwJuCxxZJBSqlHlVL3KqXuHR4evgpTNLcSLW9pqurxwrkS\no5NVSl6EodKu80LGIu8aWM3zuCLdm9NttGR871CBnGPSl3UYLDiYhmCbaejBjxKGe1wObu7h9q1F\n7th6bRntjT51/XrQSZ1mFfiuiHwJqLUeVEr98gqvexq4TUT2kIrl+7g0kxOlVAkYat0Xka8CH1JK\n6aXmmmvioTtH+NgXT/DShTI1L8IyBaWgESmylmLXUJ6cY1EPIr47lm5u6TbJbHmYIORcC0PSnvnh\nHofxkodtGvhReoz2w4RdA7nr5hHeyuVEndCJaP4lV3E0V0pFIvJLwBdJw0Z/qpR6UUQ+AjyjlHps\nte+p0XTCoS1FRnpdjowpxBBc22RzzuHkxSphkjBTC8g5FjnHwjboyjZKpdJs/o4+Fy9Mj+JJnGAa\nQjFrMlH2iROFaxn05Sz2DBduqfFsN5NOdgR9UkSywE6l1PHVvHlzg+XnFzz24SWufcdq3lujWY6L\nlVRUTONSTrwnYzFbC5iuBnhhjGUYSJf2T5oGbO51GerJ4EcJxaxDkiQ0gpjJSkDWthjusbhrWxHT\nNLRg3kA66Qj6EeBjpNPa94jI64GPKKXevdbGaTSr5dh4iU8fPs3Tr84QRAkZyyASODNTJ4pj4mYl\ne81PiJLurc20DaEv51D1062Rb71tkOlaQKkRkSgYKDjze8pLjXBVHT+aa6OT4/lvk9ZcfhVAKfUd\nEdm7hjZpNFdFq5tldLLKSI/LeMmj5sfEcYIXK6JYkbHAEIN6mGB16T4Lk7TbJ+dYvGnvIJBOVr99\nS5axuQb37xnAkMs9aF1HeePo5I9N2EzatNOl/z5rbmVa3SxlL6TqpzWMsVKUg4Q4URgCW/tzBLFa\nkw6g69F6mbWE3qyFAFNV/4oCdV1HefPpxNN8UUR+GjBF5Dbgl4Fvrq1ZGk1ntA+XOHq+zK6BDOVG\nRJwkhLHCNgUVK/KOiRclXCwH833n1zNjnrWFRAlhlHTkUQjNgRuqff2GQW/Gxo8SRnozDBZSb3nh\nFCFdR3lz6cTT/ABwB+AD/wUoAf98LY3SaDqhdRw/NVnlzHSN8bkGXz52ET+MqHgRQZQKZ6Kg7McE\nsaLiR6g1GAEXxakna1tX+puOubgPGqu0E8k1pdnaKSgUBddiU6+76CxMXUd58+nE03wD8GGl1G+2\nHmi2Pz63ZlZpNB3w+JEJZqoeR85XSJQiVjFhAmFzz4+a/8/lXO/YkgBJAmJCnFyanuRaQpQo4gWz\n5tpXUtiWwdtvGwaBV6cbKBS7B9IQgm2ZDBSsKxaX6TrKm0snnuYXga+IyKa2x/5kjezRaDrmxfMl\njpyvNDcumsQJ86PfbmSxumUABtjmpb9OirQYPU4uHzdnCIgIOcdgz2CWB/YN0Zd3ODVdB+CeHX30\n5x12DOT0pKEupRPRPE46hPhrIvLm5mPdNm5QcwtS9iKiJMGxhDCO8SN1U/5gOqbBtmIWBAxJj9qL\nDTPucQyGCy5DPS53bC0SxIqTExWeGp3BMQTLgMOjM3zl+EW88PJkj86Qdw+dHM+VUupvReQ48Oci\n8qd0X9eZZoPRyfTwYtbCMoSaHxPECSh1w/9gCuDHiju29nB8oorZI5yeqWMoRZRcEs+sRVr25IfE\nicIPE8I4YUsxQ8Y2ydhm+n4iVP2QY+MVRnovZcR1hrx76MTTFACl1MvA9wBvIx0Rp9GsCZ1MDz82\nXqLUiLBNg1oQEcUJ5k2ou0xLlxRfOT7JZMXHMQ1MkUtHckn39dimSaJANbNQcaJwLAM/jNN5n0HE\nudk652br+GHMxYp/1TMxNWtLJ22Ud7d9XwN+UkT0fDbNNbOUN9k+PRyY//qpw6cZ7slwdLzE2ekG\nm3sdilmbqp+KZrJU5ucqsaS5UgKwTSFJFEsNQ/IjhR/FvHyxgpCu0DUVzQRV+l62Kdimwc7hHK/b\nUeTIWJkLJY9SI2SqGmAZgmUIgolrCWEUM16KbrnFZd3OVa3wVUqdud6GaG4tFq7abc8Qj801sAw4\nOl5msuzhN4/e1SDmew8MU6qHIHChErBvOE/VT0uM/ChhJdHstD7TtcAUg6g5zcM1hcYKG9cECGKw\nDIVtGCQoBCFjGdSDmB39GfKuTaQUJy/W2Fp0mSh7TFR8TEmHdASxoi9nc9e2IkM9GT744IGOfp+a\nG0eXNpJpNjrL7aJxTOHpU7PM1nzKXkQQJkxXA0TB8Ykqk1Wf3oyFaxlM1ULefmCYgXw60GKlRFCn\nfqgfQb1t/FE9TFArvLj12YkCxzIwRNLNl6bgWkIQKwxD6HEtvDDm+ESNe3b20Ze1EIQE2N6f5c37\nBtk1lNeJny5lSU9TRH5FKfX/iMhblFLfuJFGaTY+7at2W7QyxK6Zik25EWEaIKT3neYRebLiM1cP\nydgGWdvkgb2D2KbQk7HxwphGmFz3hJBrpVstvSjB4NKxvZ2WxAqwcyDH2ZkGvZbB5l6XszN15hoh\neTcV+9Z1fTmHH339dkqNcD4MAWmvuU78dCfLeZr/S/PrIzfCEM2txXI91H6sODiSp+xHzNUDakE0\n3wY5XQvSVbuJYqYWcG62wee/e56yF/EDd4zwY3dvo+Be/wOUFyqGe1x6M1ZHgnziQoWyFzJbD3i5\nOcfTAM7PNZis+Li2yX17+gliNb9BUyd+1gfL/ek6JiIvAwdF5IW223dF5IUbZaBmY7KcULimcHyi\nRq9r0ZdzyDsWYawIonTwhmMazXW8gmMK5UZE1japN4d0GIa5JjZP1wL6cjaubWCsEAcIY5XWa4oi\niBMaYZrdt0wD1zbZP5wnY1ts68vq1sh1xpLHc6XUT4nIZtKOID07U3NNLJYpf/htey57rJUhbg2w\n6M1azNRCBIVlpAMxTAPiJB2btqvHJeeYVP2Y2zblOX6himsbhFG0kjlXhR/GbOrJECeKsTnviqSS\nNG+mAQXXxjbTUqJI/v/23jxOrqu69/2uM9RcPbdagzXLsi0bjLHwAATMbMIN5BFyw2AgNwOBcHl5\n5ELuzUs+FwJJICQvTl5eEkIIgUBeIOSGxEAwMRiFyZM82xISskZLrVYP6qG6azrnrPvHPtWqbnW1\nqrurW9P+fj716TPsc86uU1Wr99lrrd+KUBWmqkpH2iHpCk+fGGdLb25aaMOmRl48zOs9V9WTwPUi\nkgBqbrx9qlqd5zCLZQbzecrrvcN7+8e48579/OCZYXJJFxDaUko5jMinjaLPi7Z28+TxMUrViJPj\nJVyBfMqM0IYnyxQqIcVq60PcBTNKdOKa6T25BMMTFcK6Nq5j0jlL1RDXMXnnZsQphBhBjlX5JOUw\nwhHHjiYvUppRbn8p8HfAYcx3Z72IvFNVv7vMfbNcIjSKu6xXG683rH35JKfGS0yUQ9rTHld0Zlid\nT5JLeQyMlxmZrBrj5QrFSshYscqR4cllq/VjDKbwlpuu4MFDo4wWqwThTGeTK2Z06YjR8Awi5YrO\nNIeHJikHEY4I2YRLoErSc7lubZs1mBcpzcyY/zHwalV9qaq+BHgNcOfydstyKXF8tEg+NfP/8+xc\n6nrD2pNLMDJl0g2LlZDxYpVHj42ysTvNyfESYaQUygHjxYCpSkgp0CUZTAESjpkfna3sZqpCOtxx\ny3oODRU5MjyF7xjD6DrmB5RwIeG5ePHoMp/0UDXnXd2eigPWTfxlwhEKpYATY6UZGU6Wi4dmgtv9\n+oJqqrpfRPz5DrBY6lnXkT4rpObI0CQnJ8p84MuPs64jzZ7+Ma5e3QbA0KRJn5woBUxWAjakfVbn\nE/zDg88yVqziiBKGSiB6lijGQqipDqV9l3zKY2CiPON8Trz/BZs6yKeSjEyOkUm4gKJqVI0cMaNK\nRwXPXI8AACAASURBVIzgcTrhs21VjrTvcHi4iOc6rG5P4bsOruOQTHhcu66dyXLAh+7aw4auTMPc\nesuFSTNGc7eIfBr4Qrz+NsDWJrc0ze3X9c1QGz8yNMmjx0bZ2pPh6PAkjx0dZXCixP6TE2STPqcm\nSvTlk/Tmk6z3M9yypZuvPv4sx0+XcB2hGkaEgNMCg+m5wpqOFKcny2cZYBOY7nB6KuD4aJFKENGT\nS3ByvGzCnuLhZDrhsqk7S7Ea0p1L8hu3XzXDAH7gy4+zpj01XddnqFDiRycnCCO4eXPXWXqZlgub\nZozme4D3YspcAHwP+Itl65HlvNCMqtBSzvfKa3rZPzDJ8dEiJyfKbO3JcHh4imI1ohyETFVCCqWA\nLb1GWu3YSJGuXILr1rXx8OFhDg6ZR/laLjecKROxGNtZ09zM+C6dmQSHh4yepWBENmrq7mEUUSib\n+NEfD0wAsKY9xcC4yRl3HWFjV4adm4zxm8vwJV3hu/sHqYQRbSmfqXKAiNCVO5MNBdiKkhcJ55zT\nVNWyqv6xqr4xft2pquWV6JxlZWhGVWip5/vW3kFuv66PP/rZ69nQlWGwUGG0FABKNYzwXBAHRotV\n0gkXxxWSrsNwocwDh05Pn7teMGMpYsO1keZYMTCjvjhHUuPttQoVlRCqYcT2vixd2QSFUoArxnD2\n5pKsbkuxtjPdMLZyb/8YJ8ZKFEoBCUcoVgIODU1SqoZsqytnYfUyLx4WJdhhubRoxrvdyvOt60jz\n7T0DJDzBcx0iNTqSCU9QhBdu7eHxZ0d59nSRo6en5h1R1kaGC53brK9GWYjrBrlAyNlVKn1H+OdH\nTvCmG9fxgwPDPHpsFEG47ape7rh147z36O6nBtjYnWVNe4oDg5MUSgHZpIfvCr35M2mkVi/z4sEa\nTcu8eeCLPV9NpahQCsilPLb0ZDg+OsWd9+xnT/8YU9WQlAqe46CqlAMTwD5RCvjhgSFSCY+r+vI8\ndWKcMGqcS+6wcIM5F4rJHfcdZnji057QkUlwaGiSHx4Y5vffuDAp2dq9dcSfNpID40V++MwIY8Wq\nrSh5EWKNpmVO7/ZSRj4JV3jg4AiuI0yUq/SPFfnRiTGySZd1HRmuXt3G/pMTnBgtUgmNepCRRjMV\nHUemqnQBW3oy8xpMgITvUg7CcyoQzaYWWaSY+UsHYzQdMeFBTiwevKknR8p3UVUePbbw6Yq57m3K\n93jxtm7a0/5Z2VCWC59mgtu3Ax8ENta3V9WXL2O/LCvIbO/2UkY+e/vHeOLZMU6MFYkUsgkX3xUm\ny0ZdshKEjEyG5FM+jlOiVNXpR2wRIYyMAc2nfY6OFBtOumc8ARGSvkOlGjZo1ZjZ6Y8IiELKd/E9\nxXfM1MGzp4sEUYQoJBMue/vHGhq3uZxpje6t9ZRfvIie41+0iDwOfBJ4GM5kjanqw8vbtbnZuXOn\n7t5tI55aTSu85zUH0JPPjjE6VWayEhJGSkfaNznYEaR8E6+YS7pGL7NsNDB9V0j5Dr7jECp0Znwm\nysbpMlGsMhVrsdVGiO1pj6lKSDWcvy6QK8axU43r9TgwI/Ux4ZrH+yiChCes6UhzdV+e7x0YwhWh\nVA1BzP7V7Ul2buqeNnj19yzpCifGSmzszp5lHIGWRiZYlgcReVhVd56rXTOP54Gq/mUL+mS5gGmF\nYETNAdSVSzA8WaYzkyCMlEiVamRUzMuBkklA/1gRo1AZx0s6Qtr3iFSpVCOmKiGqSqCQTvp0ZYXh\nqSrVICICMgmXShCZNtGZkaMTe4xq05L15XNFwHGEhMOM/PSk5+C5LqtyPr7n0ZU1ZTRGp6qImNIV\njmO0L+tL6dbn0393/yCFUjA9f1nv/Hr/q7bP6VW3hvTipJk0yq+KyK+KyBoR6aq9lr1nlouOWrrk\ntt4sjgiVIMIRGC9VSbgOniNxYbGIamiqOIIJI6rFa06UAsIoohxEpH1jGKvVgOGp6nQdoLTvgAhX\ndGZozyTozSfIJsxXOYrTF2sjTADHMSLCvmuM9LqODG1JF0fAEYf2dIJV+QShCjesb6ccKi/Z3kM6\n4ZJNunRkfDb2ZAijMw6y2crzlTAil3Q5MDg5fT8aOdNaHeJlWVmaGWm+M/77wbptCmxpfXcsFzM1\np0dvPsVNmzr54cFhJsoR1VDJ+EI64VIJI8aLwXQYUS1fWzExk5mESz7pMlEOODFapT76x6EWcK5M\nlgM6Mz4JzyEIInzXpSPtMF4KiBSSLqxrTxOJcHqyykSpStJzSCVcJkoB3bkk2WpIKYhoS3skPZee\nXJI7bt3I3U8NMFassrU3R6kakoqVizK+O+0gmx1x0JbyKVYCCnXCyo2caa0O8bKsLM1Uo9y8Eh2x\nXPzUnB4jhTJHT09RjechM76D4zgkXIfrr+jg3n2D+I7G8ZlnHpMVKFVNdpAnMFuDo7ZeCZRqGLDv\n5ATomdITCTcOShcII+HI6SIdaZ/efIKEZ0aZQaSEqrxwWzeFUkD/eHnO/O9PffcQq9uS/OjkhCnY\nprCxKzPtIKsZ1prB27Yqy33PjJBPmSmG+ZxprQ7xsqwszXjPfUwq5UviTbuAv7KamhY4e27u6tVZ\nvvJoPwPjJXJJl85sYnruMZXwGJ6skPQcSlUl4zsEYUS1zpVTy/iZTxIzgjnTgSqxh8cT8FwTpD5R\nCqiESi7h0pZJANCW8vBdF8eJ+J3X7zhrdFdTUr/7qQGmqiFjxYC2lMfm3txZhhWMwfNdl809Wfra\nkvSPleYNI2p1iJdlZWnGe/5pwAc+F296OxCq6i+d8+QitwN/ikm2+LSqfnzW/ndj8tpDoAC8S1X3\nzHdO6z2/cKjNzYWhEQQemTQ1cTb3pNl3ssBk5Uz8pCvgOqbiYi7hIiJ4rkyX3o3UzEHWcr4XQ+14\n14GU58QOIjMR0J1NsHNjJ/sHCqzvTrNjTTvb+7LT+fCLccYs1plTrx1qw5AuHJr1njcVcqSq159r\n2xzHucB+4FXAs8BDwFvqjaKItKnqeLz8euBXVfX2+c5rjeb55+tPHOdz9x1l38kJXDFOllVtaZKe\nw57+cYqVcEaOeA1XTKiQKozGlSZVmbPtYql5yX1XiOIhaaTQlU3w1ps3Thu28224rPf8wqOVIUeh\niGxV1WfiE29hZqhbI24CDqjqwfi4LwJvAKaNZs1gxmRZ/CDDskJ8/YnjfPwb+8gmPVxRCuWQ0aIy\nXgroyCSoBHMbTDCebVVTQ9wUsxBaXWy3JrZRK2zme2YutTubnGGYluKMaYXBszWBLl6aCTn6IPAd\nEdklIv8B3Av8tyaOWwccq1t/Nt42AxF5r4g8A3yCM/Jzs9u8S0R2i8juwcHBJi5tWS4+d99RskmP\n9rSP6zgEkRnJFSsRgxPl6XnFuVBgohwQBBGRSeiZVhNqJdVaVlHKJ+G5dKR9tvflpuMroTk1+bmw\n4UKWZqThvg1ciTFo7wOuUtXvtKoDqvrnqroV+O/Abzdo8ylV3amqO3t7e1t1acsiGBgvkU+6TJYD\nJsvBdPB4BARNPGdrHHg+Pdd5rlq4C0CAnnySRGyJJysBPVmfF27rYWNPdoZBnK/u+nzMjs+sLdcb\nZMulTUOjKSIvj/++EXgdsC1+vS7edi6OA+vr1q+ItzXii8BPN3Fey3mkry3FUKFC/1hp2nlT41xl\nejzHzDXWTGsUV2tsxFzmtBbEXt/Gd42nMeU7bOzK0J7xyaZ8NnRl6M6l6M2nzjKI89Vdn4/FjlAt\nlw7zjTRfGv/9qTle/6mJcz8EXCkim+MSwG8G7qpvICJX1q2+Dvhxk/22nCfeeesGhgsVwsgUFoMz\nOd3zUStYVomNZE0Dc77B6WxRDQczDVC/LS7ZAwJr2pKICKvyKVAYL1YZK1bmNIi1sKL2tCn/20hE\neDaLHaFaLh0aOoJU9UPx4kdU9VD9PhE5Z8C7qgYi8l+Bb2IGAp9R1adF5CPAblW9C/ivIvJKoAqc\n5kz2kWWFada58brnruPz9x3h6EiR8ZJJk0z5rqkcOVu9N8YTWJVPMlEOCFUpV6OmPeYmHVIII6OG\nFGGUk/raktMxmOVqSNqDfNrEYWaTHus6UpwYK8Vpkv6cMZOLcca0UhHKcnHSTMjRI6r6/FnbHlbV\nG5e1Zw2wIUetZ77wGzhboac+G+bHA+M8cOg0E6XqmVEkM+v3OGICyqcqIUKtLviZ689X58cR6M76\nOOLgOjBZCVnbnubV166ebnPfM0P0jxXJJn2SnkPSM+mUKNz55utb7qW24UKXJksOORKRq4FrgfZZ\nc5htQGruoywXI43Cb75w3xGmqtG0kk/NU/zKa3r51t5BRgplhgoV0r5LGEVUi0EcIwmuOISq01qa\nE3FOuBPHUMIZw5n0hdKsFKBpg4upKa4KxWpIynNJeDMnA1a3pRgvBVzVl6N/rMTwZAXfdXjfK7Yu\nizGz4UKXN/PFaV6FmbvswMxj1pgAfnk5O2VZWRrlQn977wg3be46y5juH5jkldf08mf3PkM1jFjb\nkWJNewe79g8RBKHxjouQ8WKhi4RDGJkCZWDk2SQOPBfm9rrXtgQKw4UyCc/kr2/szpBNejNKRbiu\nw/tevpX9A5P4nsutW3vs6M+ybMw3p/mvIvI14L+r6u+vYJ8sK0yjXGhF5/UU37LFlGwYKpQ4cGqS\nlCecrihr21P05pPsHygQREo+4TE6dUaxqFyJcJ36x/f5C/GWAqUahty6uZ3f+qkdwMwpg9p85eta\neE8slkY0M6f5oKretEL9OSd2TrP1NJrTTPsOhVLAoeFJ+sdKgNGy7EgnGC8H9OWTVIKQfbFxRNUo\noPsOvgPlQOnM+FRCpRKEFOo8374jRHUKwfM5hgTIJV02dme5/gqjd2nnEi2tppW553diBDu+BEwr\nrKrqI0vt5GKwRnNh1DstEq4wOlXh8HARRblhfQdvj0vQzuXcODhY4KNf28tkOSTlG1HhQiWkPeXj\nCEyUqpRnZQA5QF9bgkoI3dkEY0UjhuW5wnChRC1ax3eMsIbrOERRRKAyZ8ymAyCQT7pEKvTmk9y8\npZOk51mRC0tLaWXu+fPivx+p26aALax2gVM/gvQc+O7+IUanKqxpT5FJuDxwcIT+sRK/cftVczo3\n7n5qgLaUT6UaUiibejwuJv7Rc+Usg1l7yB6ZqqIKpUqVYlya13Vk2mA6QBiZx/L2lM94OSDjuYwV\nK2fVHJc43bIaKfmkRz7lcXBwilu2dE/3sfbXerMtK0EzIsQvW4mOWFpPvVd8T/841TAk6btMVkO6\nckkQYWSy0lCk4vhokUoQIo5DW8qE8ZSrakrd6tyhQop5LE96UA0FV+KA9rpRZCrhIAiOmBAi82Qf\nkXSds0r2Rgop11QT6skl4nAiM3rNpzyePjHG0ZGpszz8s0egNkzI0irOmXsuIu0i8sc1wQwR+X9E\nxH7bLnD29o9xz54B7j84xP0HhxmaKBFEStIzj9lgCopVgqhhCuC6jjTFqqkW6cU1fmoDwbmqQNav\nh6HRshSRs75kYagkPMeUraiEcRiSQybpkvJMWFIthTLtO2xdlWdTTxbPdSgHEa4I9x0c5htPnmRP\n/zhRFM2bC25FNiytpBmVo89gwoz+c/waB/52OTtlWRo1I+G7QtJ1KMXq46pmFFiLcywHEQnPaZgC\nePt1fTiOUA0jqmE0Ixf8XAk9tbK4keqMJPLa6NRobiqOwM4NHQAkPY/tfW1ct7aNzT1ZbtrUyc1b\nutm5qYvnrGujUAoYmihTKAeMF6tmTlSEH52cYKhQmr7G7FxwK7JhaSXNGM2tqvohVT0Yv34HW1Tt\ngqZmJK5d2zZd8bEr4xGEEeVqSNZ3KcVFwLqyiYYiFdesaeflV/Wyui1FpGZe0W22E3U1flK+gxuP\nHn1Hpsv6grC+M831G7qMFmcYUomUdMLj1q1d7Fjbztr2FO1pnyCCm7d00ZVN4DrG8O3c1MmajjQi\nwoFTZ6pAzs4FtyIbllbSjCOoKCIvVtXvA4jIiwD7bbuAqQWrO+Lz/A0dHBicpBJEdOUS7FjTxuHh\nIpVIuXlL17T3vBF33LpxOivo/oNDnBovM1GqnlNsI1KTNx5GRs0om/SYLAcgiuc4tKc9XIFM0uOe\nPQMMFyrkki6v3rG67hxK/1jA+1+1fXrbB778ePzezPBVFR45MspQodywoJmtyWNpJc0YzfcAn4vn\nMQUYwQprXNDUG4nevJFGq63XG6BmqC8y5oiDKqzvTHNyosLUPIrDDkaBqBqZNMhc0iOdEEYnAzwH\nU3Y34eKIkE46nAJGSwGDEyV68yY7aS7DNtsA9uZTbO/LcXKi3LCgmRXZsLSSZrznjwHXi0hbvD5+\njkMs55lzGYmFepJr+4YmShw/PcVYKSSbcHGBiUaGU0Acl02dKZK+S7ESMDIZsqE7w85NnXzjyQEm\nSiEd2QgCyKXM9MHTJ8Z5yfZkQ8M213tzXWfOqpL1/a8Z/tlZRBbLQmkmuL0b+BDwYswc/vcxcnHD\ny9+9s7HB7c3RyDAupqBY/TEDY1Pcd/A05WpAwnephhHFSnSW9qURHHboa0+RdF1CjQgjuHVrFz25\nFF99/DiFUkAEbO3NsW1VlkiVx4+Ns2Nt27zG3IYPWZaDVga3fxH4LvAz8frbMNlBr1x89yzLTSMl\nnvkUjXryqTkNUf0x7el2RITdR04zWQ7IJjwSbkSxGpF0hUCVIFTSvktHxqccRDgijBWrvGrHKnpy\n5tG7J58inwyoRDodqG7a9J1zCsGqDFnOJ80YzTWq+tG69d8VkZ9brg5Zlpc9/WOMTVWZKAe0pfzp\nEd4PDozw8qtXnRUgDnDPngEijWhPJ+jO+hwcmmJ1W2r6HAcGJlBVHMeBENpSLq7jUA2V9ozPK6/p\n49t7T5H0znzdtvVmeeDgCLmU19CBY7FciDRjNP9dRN4M/GO8/iaMGrvlImNv/xjHhouUg5ByEHFy\nrMSBUwVySZfOTKKhnqbvCqiJ93zgUIHubIJSpIwXjdH0PQcNIxAhm3BMGV1HKFZD1qcyTJQCbljf\nPp2Hnk95JDyXDd0Z1ran+NHJccaKAW0pbzp20o4kLRcqzRjNXwb+L+AL8boDTIrIrwCqqm3L1TlL\na7n7qQH62hI8dPj0meJmCqOTFa5enZ/Rtl5P89q1bTxydJSkJ4RRxGixQhhBV9Yn5btkEh5huQoK\nYRgROQ7lwGT6rG5LNlSB/43brwLgU989xLqODPmU1zANslXY+VDLUmnGe54/VxvLxcHx0SJDhQqu\nI0SqcXE0IXTg4NAkV9cZj3o9zfp4T891qAZKR8anM2Nq8uSSHuUgojvrMzJVJeU5FKsRV/bl2NST\nm2GY6hWV/ub7hzk6MsWatuSC5lgXS71Da748dYtlPpoZaSIirwdeEq/uUtWvLV+XLK2mZqT2nBjn\n4OAkmYRLOmFye4JQKUrAkeEpvv5EP11Zn9VtKVzX4Yb1HUyUghnxnmvakuw7WSDpO5yeLDNeCimU\nqyRcl5PjRmH9Jdt7GwbNzzZcjx0dZWyqQi7lTTuJykHQcI51KcatkROskWCJxTIXzQh2fBz4NWBP\n/Po1EfnYcnfM0hrqxSquX98WO12qVILIGMxqiGLq9kxWquztH+cHzwxz9eosb79144za4IcGC+w7\nWaAj6zNZDjh6ukipGuLFtcyD0MRvfv/AMJ+/78icghiz88C7comz0iD3nJigM9P6XHGbTmlpBc3k\nnv8k8CpV/Yyqfga4HWxlgYuFeiO1Kp9mQ1cGETFB4Y7gu+ZR3XMdVuVTXLOmjd58kq880g8wXRt8\nb/84+08VuGp1js09GUYmTe3z8VLARDFkslwliKBQCuhMezx9YnxOJaHZhmtbbxaU6TTIsWKV01NV\nrllz9hzrUo2brVluaQXNGE0wxdVq2OeYi4jZRuqmzV2s7UiTTri0pT2TI+4IffkUKd9lqhIyVqxw\nfLTIh+7aw8HBAgAnRkskPYdqGHHfMyMUygGOmGwHxTiUBKZrm1dDnXN0ONtw1dIgu3NJ+sdKtKd9\nXrytm5Q/c0TYCuN2+3V9HBuZYte+U3zz6X527TvFsZGphoIlFstcNDOn+THgURH5DuZ38RLgfyxr\nryxzshjPby1XuxKEHBicpFAK8F1hXUeaq9e00Zae4sToFL4nHBwsMF4KEFFySY8To1N8/Bv7uGF9\nB5FGoA4PHDqNRpFRbq+aapJuXOMnVCXpuQxOlNnYk5tzdNhMGmRtSgGgVA3Y2z/B6akqP7Gtm739\nY0uaf4y0VptdZqxbLM0y70hTRASTNnkL8M/A/wJuVdUvrUDfLHUsVkj39uv6ODI8yQMHRyhXjFhG\nGEFfW4pffPEmfuf1O/Ach2PDRSbLxmCqCuVKRLkakU16nJwo055OgJhH+WIQkU141KQya6PNMDLC\nxsUgYltvds7RYS0PvD3tT48sZzt4am0qQcgPnxkB4EXbuvA9d0niwXc/NcDG7iy3XbWKV1+7mtuu\nWsXG7qzV1bQsiHlHmqqqIvJvqvoc4K4V6pNlDhbr+b1mjdGkHJmsMFGqUgoikp7DUKHMF+47wu+9\n8bnsWJ3nhwdHCAIl4QqOK5QDpVSs4jlCJQi5dWs3Dx8ZxXeEoik8ie8KIhDGVSUdRxAR1renSHhu\nwwyfZtIgr1nTTm8+xcuvXjVD0q2Z99yIRvXdrSPIshCaeTx/REReoKoPLXtvLA1Zyg++HCo71uZ5\n9OgY+ZRP0jPZPd87MMzXnzjO4ZEinRmPIIrQeD4ym3ApVkOK1cjU+AFu3NjB9/cPcXqqwnipSsp3\niELF9126cwk2dqUZGK+wtjNNe9qfV0momamGVhs5q6tpaQXNOIJuBu4XkWdE5AkReVJEnljujllm\nshTP77qONHtOTJD0HFK+i4gZEaY8hz+79xnCKGKyEiJAsRrhieA5Qtp3iVTJJ11+PFBgomhUiV60\nrYcrV+VwHQccoT3tAcpgocqNGzv49Vdt5/2v2n5O1aRzTTW02tt9+3V9M0KoasvWEWRZCM2MNF+z\n7L2wnJOlCOnefl0fX3n0OJ1pj0KpymChTDnQuPa4MY6VUEknXArlkIlKSDGI6Mz4XLM6TzVSBibK\nIMLzN3SwqSc3fe7DQyZ285at3dP9OlcgerNTDa0WD7a6mpZW0NBoikgKeDewDXgS+BtVDRq1tywv\nS/nBX7OmnRdv62b34dMMTJRIeS5XdCR5drRIoRzQnk5wRYfPwHjR1BlX6Eh7rO1IMzxVZfuqHLdu\n7Znzcbl/zFS5XMhca7OP3cth5KysnGWpzDfS/BxQBb4HvBbYgckMspwn5vvBn2uO8O23buSRI6Os\naU/TljK54p7jEDoRhXLA+myGhOfSllLKQUSoxhNeDiL2DxR4z8u2cvdTA2fNCQ5PVujOJmb05Vzz\njguZW7RGznKhMZ/R3BF7zRGRvwEeXJkuWRbC3v4xPn/fEb5/YJjOjM+OtfmGudrlMGSiFDBUKLO6\nLcXNmzt54NBpCuUAVWWqEuKIsCqfpBpprJfp0Z7xp88z+3HZdx1Wt80cNdYbwLmMua3ZY7mYmc9o\nVmsLqhqIyDxNLc3QKlmyaQGOWB8z1IjOtPkoHz06xo0bO6azcepLXLSlfPJJDxGhHER0ZBI8Z22e\nJ09MMDJZJe27JD2HpO9ybU+G4ckqQwUzl1kLKp/9uPy+l2/lW3sHGStWzzKA86kK2blFy8VKwxpB\nIhICNRUFAdLAVLzclI6miNwO/CngAp9W1Y/P2v/rwC8BATAI/IKqHpnvnBdrjaDF1OY513n2nBhj\nvBTQP1biio4UuZRPqRqS8l1u2txF/1iJP/rZ67nznv2MFatUw5CHj4yS9BymKgHjxYCk73Ld2jwd\nmQSnJso8e7rI6jaT0ihi8iS39+VwXadhXxv9M6hdt/4xfL6qmFbr0nI+WXKNIFV1l9gBF/hz4FXA\ns8BDInKXqu6pa/YosFNVp0TkPcAngEuylEarZMnqz1N7fB6ZFIYKFXJxDOZ4qTrjEbm+DvqNGzt4\n/NgYJ8dKeK7Dy6/pJekZ8d//9mpjyD501x4jMpzz2dabnS4B3KivjeYdFxJnabUuLRcLzQp2LIab\ngAOqelBVK5gCbW+ob6Cq31HVqXj1fuCSndRqlSxZ/XnaUqZwWW8uSSkIKVXNK+E6M+IP6+Mde3Ip\nskmPdZ0ZruzLsyqfniG9ds2adjZ0ZXjtc1Zz65bu6Rrki+nrQuIsZ0vGtUoOzmJpNctpNNcBx+rW\nn423NeIXgW/MtUNE3iUiu0Vk9+DgYAu7uHK0KlC7/jzbVmWnPd1XdJrzjBYDrlvbNmOENjuo+/jo\nFIMTZQbHS9x3cJjBidIMo9iqvi4kmNxqXVouFpbTaDaNiNwB7AT+cK79qvopVd2pqjt7e3tXtnMt\nolXZKPXn6comuarPBJqvaU9z21Wr+Ku3P5/fe+Nz5xTAaE/7/OjkOMVKRD7p0ptPUq6GPHJ0lCND\nk9NGsVV9bUaco4bVurRcLDR0BC35xCK3Ah9W1dfE678JoKofm9XulcCfAS9V1VPnOu/F6giC1nvP\nF3OeO+/Zb7J4BgrGU+45jJcCULjzzdfPkGdbSadMqxxlFstiadYRtJxG0wP2A68AjgMPAW9V1afr\n2twA/BNwu6r+uJnzXsxGs8b59BJ/4MuPs6Y9xchkmQOnJhkvVcknTSzmX7/jBSvSh0ZY77nlfHLe\njWbciZ8E/gQTcvQZVf09EfkIsFtV7xKRbwHPAfrjQ46q6uvnO+fFbjTP94hqrjCgw0MF+sfLbOjK\nWGNluWy5IIzmcnCxG82Fxi5Ca0dgs4320eFJHjk6yg3rO9jYk7WPxZbLlmaN5gXhCLqcWKiXuF5G\nzXdh175T/MrnH+H//ucnFqVgPts50z9e5ob1HWzuzdlQH4ulCazRXGEW6iWuxS9Ww5BHjxojOV+1\nx2a4Zk0773/Vdv7oZ69nQ1eGjT3ZGfttqI/F0phm9DQtLeRcYhWzH8X39I9x9eo2Hjw0Pi0iIqZE\nqgAADo9JREFUrKoUyuGMgPR6FvI4b9XMLZaFYec0zwOzjdr2viz7ByanBTi29+Wm5xd37Rsg6bqc\nnCiRTXh05xI4cYGztO8yMFHmp5+3btowLtTRdL4dUxbLhcKSc88ty0d9rna90RqbqoLA/lMFcvG8\nZxAoo8XydF2fYyNFsgkH33WohkpfPjkjT/vupwYIw4g9/eMUSgG5lMfqfHLevHGrOGSxNI81mueZ\nuQQ4ykHEgUEjMNWdS+CVquSSPsdOT+E5DqUgJJ82wr/bVuVmiH88fWKMZ0eKJH2HXNKlXA3ZP1Bg\nqho27IMV+rVYmscazfNMvRJQWyztlvQcCrGzyHOgry3NLVu6GZwoceBUgaf7x2lLeWxblTtLUGO8\nFIBAyjciVSnfpRxEZvsSsIHnFovBes/PM3MJcIyXAnJJF98VCuWQbauMd7s3n2LH2nauXdvOjrXt\n0wYTzjhv2tMeqkqpGs74255e/P/HZqtHWiyXA9ZonmcaCXC0pX2uXdvG5p4svuvOEM54560bGgpq\n7FjTztWr86R8l4lyQMp3uXp1nh1LGBVa2TaL5Qz28fw8M9sRs6knx7tv29pQOKPmpNnSm5ux/QWb\nOmaUwNjel+OmzV3T3vCl1PZeiJiwxXKpY43mBcB8jphG+xp54K9e3UbGd9kXO3+uXdu+ZG+4jeW0\nWM5gjeYlwOxSGpt6cnRmk/Pmsy8EWz3SYjmDndO8BFhu1fOFiAlbLJc6dqR5CbASj882ltNiMVy2\nRvNSiju0j88Wy8pxWT6eX2pxh/bx2WJZOS7LkWarapBfSNjHZ4tlZbgsR5q2XKzFYlksl6XRtOVi\nLRbLYrksjWar6npbLJbLj8vSaFrHicViWSyXpSMIrOPEYrEsjstypGmxWCyLxRpNi8ViWQDWaFos\nFssCsEbTYrFYFoA1mhaLxbIArNG0WCyWBWCNpsVisSwAazQtFotlAVijabFYLAvAGk2LxWJZANZo\nWiwWywKwRtNisVgWgKjq+e7DghCRQeDIMp2+BxhapnPbPtg+2D5c2H3YqKq952p00RnN5UREdqvq\nTtsH2wfbB9uHRtjHc4vFYlkA1mhaLBbLArBGcyafOt8dwPahhu2DwfbBcCH0AbBzmhaLxbIg7EjT\nYrFYFoA1mhaLxbIALhujKSK3i8g+ETkgIv9jjv1JEflSvP8BEdkUb98kIkUReSx+fXIZ+/ASEXlE\nRAIRedOsfe8UkR/Hr3eepz6EdffhrmXsw6+LyB4ReUJEvi0iG+v2Lfk+LPH6LbkHTfbj3SLyZHyt\n74vIjrp9vxkft09EXrOS11/J30Rdu58RERWRnXXblnwPFoWqXvIvwAWeAbYACeBxYMesNr8KfDJe\nfjPwpXh5E/DUCvVhE/Bc4O+AN9Vt7wIOxn874+XOlexDvK+wQvfhZUAmXn5P3Wex5PuwlOu36h4s\noB9tdcuvB+6Ol3fE7ZPA5vg87gpef8V+E3G7PPBd4H5gZ6vuwWJfl8tI8ybggKoeVNUK8EXgDbPa\nvAH4XLz8T8ArRERWsg+qelhVnwCiWce+BrhHVUdU9TRwD3D7CvehVTTTh++o6lS8ej9wRbzcivuw\nlOu3kmb6MV63mgVqXts3AF9U1bKqHgIOxOdbqeu3imZ+lwAfBf4AKNVta8U9WBSXi9FcBxyrW382\n3jZnG1UNgDGgO963WUQeFZH/EJGfWMY+LMexrTxPSkR2i8j9IvLTi7j+Yvrwi8A3Fnlsq68PrbkH\nTfdDRN4rIs8AnwD+z4Ucu4zXhxX6TYjI84H1qvr1xfR/OfBW4iIXOf3ABlUdFpEbgX8RkWtn/Re+\nXNioqsdFZAtwr4g8qarPLNfFROQOYCfw0uW6xiKuv6L3QFX/HPhzEXkr8NvAouezW3j9FflNiIgD\n/DHw860871K5XEaax4H1detXxNvmbCMiHtAODMfD/2EAVX0YM3eyfZn6sBzHtuw8qno8/nsQ2AXc\nsFx9EJFXAr8FvF5Vyws5dhmv36p70HQ/6vgiUBvZrth9mOv6K/ibyAPXAbtE5DBwC3BX7Axq1W9i\n4azExOn5fmFG1AcxE8a1CedrZ7V5LzMdQf8YL/cSTzBjJqyPA13L0Ye6tp/lbEfQIYzzozNeXuk+\ndALJeLkH+DFzTNq36LO4AfNDvHLW9iXfhyVevyX3YAH9uLJu+aeA3fHytcx0ghxk4Y6gpVx/xX8T\ncftdnHEELfkeLPa17Be4UF7ATwL74x/Db8XbPoIZSQCkgC9jJpQfBLbE238GeBp4DHgE+Kll7MML\nMHMzk8Aw8HTdsb8Q9+0A8F9Wug/AC4En4y/qk8AvLmMfvgUMxPf8MeCuVt6HxV6/lfegyX78ad13\n7zv1BgUzCn4G2Ae8diWvv5K/iVltdxEbzVbdg8W8bBqlxWKxLIDLZU7TYrFYWoI1mhaLxbIArNG0\nWCyWBWCNpsVisSwAazQtFotlAVij2UJE5P+oU36pvSIRee357ttsROTDIvKBBbTfJCJPNdj36ToF\nnMMi0hMv/7Du2Lcuoo//ECsNvX+hxy7wOrtEZHfd+k4R2bUM1/l5EVlbtz5935Z43l2x0s9jIrJX\nRN5Vt+/fRKRjqdc4x/V/2ESbwnL2YSWxaZQtRFW/Anylth5/ed8GfLOZ42OBEFHVlohliIinJo9+\nWVHVX2qw/YXx4ibgrcD/3+w5RWQ18AJV3TbHvuV4X6tE5LWq+o1zN100Pw88BZyAxvdtkbxNVXeL\nSBfwjIh8VlUrqvqTLbzGnNR9zpcFdqS5TIjIduB/Am+vGUER+aCIPBSPnn4n3rYpHiX8HeYHtV5E\n3hLrGD4lIn/Q4PyHReQTcbsHRWRbvP2zIvJJEXkA+ISIdInIv8TXvF9Enlt3mutF5D4x2pS/HB+f\nE6Mh+Uh87nrVGU9E/j4ezfyTiGTiY3ZJnc5hXR9ro4uPAz8Rj4TeLyLfFZHn1bX7vohcP+vwfwfW\nxcf8RHyNP4lHhL8W37d75Yzm5Ya69/+X8Xs9KCK3ichn4j5/dp6P7A8xwdKz34MrIn9Y97n9Srzd\nEZG/EJEficg98YjuTfG+/xm3f0pEPiWGN2Hy2P8+fk/p2n0To1v5h3XX/HkR+f/i5Tviz/cxEfkr\nEXHneQ8AOUxiQhgff1hEeuL7tVdE/lpEnhaRfxeRdNzmefH9ekJEviIinfH2XSJypxiBkr0i8gIR\n+ef4+/K7sz/nc3x3Lh1WKor+cnoBPrAb+Lm6ba/GFIcSzD+rrwEvwYzCIuCWuN1a4CgmVc0D7gV+\neo5rHOZMBsU7gK/Fy5+Nz11Lc/sz4EPx8suBx+LlD2MyW9KYlMBj8bU9Yh3FePuBuM+bMNJgL4r3\nfQb4QLy8izPpbYeBnni5EP+9rda/eP2dwJ/Ey9uJ0/Nmvb9N1Gk2xtf4i7r1rwLvjJd/AfiXuvf/\nxbjPbwDGgefE9/xh4HlzXGsXxqDdi9HS3Ansive9C/jteDkZf66bgTcB/xafdzVwmjjtlLqUQuDz\nxBkznJ3RUrtuL0Yirbb9G8CLgWvi9+nH2/8CeEeD/u8DngCKwK/M+p70xPczqL1/4B+BO+LlJ4CX\nxssfqftsdgF/EC//GmaEvCa+D88C3bM+5zm/O/VtLoWXHWkuDx/FpB9+qW7bq+PXo5jUs6uBK+N9\nR1T1/nj5BZgf7KCaR9C/xxjXufiHur+31m3/sqqG8fKLMT9cVPVeoFtE2uJ9/6qqRVUdwqTJ3YQx\nNr8vIk9g0gnXAX1x+2Oq+oN4+QvxuRfDl4H/JCI+xuB9tsnj6u/nrZx53P/8rL58Vc0v9UlgQFWf\nVDPafxpjPBrxuxgln3peDbxDRB4DHsDIBV4ZX+/Lqhqp6knM/avxMjHq/09i/lFdO9+bUtVB4KCI\n3CIi3Zjvxg+AVwA3Ag/F138FJtd7Lt6mqs8FNgAfkDq1+ToOqepj8fLDwCYRaQc6VPU/4u2fY+b3\nraZO/yTmO92vRsDkIDMFM2D+784lg53TbDEichsmN/f5s3cBH1PVv5rVfhPmcWoxaIPlZs83O4dW\nMXOwvcCNqloVoy6Tmqf9glHVKRG5BzMS/M8Yw9AMzb6vmipRVLdcW2/4nVfVe+PHzlvqNgvwPlWd\nMS8tInPOFYpICjMi3Kmqx0Tkw5y5f/PxRcy9+BHwFVVVERHgc6r6m00cX3sPgyLyCHAzcGTW7vp7\nEWKeMs7FQu7lfN+dSwY70mwh8VzQ32IeoSZm7f4m8AsikovbrhORVXOc5kHgpfE8lAu8BfiPOdoB\n/Fzd3/satPke5stcM+hDekb38A0ikopHN7cBD2Ek8U7FX/qXAfUjlg0iUhvRvhX4foNrzmYCI/NV\nz6eB/xd4SI0K+0L5IUaNCsz7+94izjEXvwv8Rt36N4H3xKNiRGS7iGQxI8Gfiec2+zD3D84YiaH4\ns66vszTXfajxFcw/kbdgDCjAt4E31b4nYuan5xpBTiNmnrmm0nROVHUMOC1nhITfTuPv27mY77tz\nyWBHmq3l3cAq4C9lZqWMj6nql0TkGuC+eF8BuIN4wr6GqvaLKTD1Hcwo5+uq+q8NrtcZPwqVMT+2\nufgw8Jm43RQzRWyfiK/TA3xUVU+IyN8DX40fLXdjRj419gHvFZHPAHuAv2x4J2byBBCKyOPAZ1X1\nTlV9WETGMf9kFsP7gL8VkQ8Cg8B/WeR5ZqCq/yYig3WbPo15pH8kHvkNYnQl/xfmcXkPZj74EWBM\nVUdF5K8xTr2TmH9ENT4LfFJEisycTkFVT4vIXozU3IPxtj0i8tvAv4sR5K1iJAxnjyDBOJiKmPnG\nz6rRuWyWd8b9ymAeuxd7L+f77lwyWJWji5T40WdnPB950SEmXnEXcLW2KMRqpRGRnKoW4pH6gxgn\n2cnz3S/L8mJHmpYVR0TeAfwe8OsXq8GM+ZqYwPEEZqRuDeZlgB1pWiwWywKwjiCLxWJZANZoWiwW\nywKwRtNisVgWgDWaFovFsgCs0bRYLJYF8L8BSKgTRgPeXmkAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f24f1d7d910>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(5, 5))\n",
    "plt.scatter(np.mean(process_proba(results[\"dropout\"], results[\"mean\"], results[\"dispersion\"]), axis=1),\\\n",
    "        np.mean(data == 0, axis=1), alpha=0.5)\n",
    "#plt.ylim((0.045, 0.27))\n",
    "plt.xlabel(\"Zero probability from Negative Binomial\")\n",
    "plt.ylabel(\"Proportion of zero entries in each cell\")\n",
    "plt.savefig(\"unnormalized_zero_NB_genes.pdf\", dpi=300)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAU0AAAFACAYAAAA4bi4aAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nkP8AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f24f37ff0d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(5, 5))\n",
    "plt.scatter(np.mean(process_proba(results[\"dropout\"]), axis=1),\\\n",
    "        qc[:, 5], alpha=0.5)\n",
    "#plt.ylim((0.045, 0.27))\n",
    "plt.xlabel(\"Zero probability from Bernouilli\")\n",
    "plt.ylabel(\"Genome not gene QC in each cell\")\n",
    "plt.savefig(\"unnormalized_zero_B_genome.pdf\", dpi=300)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAU0AAAFACAYAAAA4bi4aAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzsvXmYXOd1n/meu1TdWrqqFzQa+8IFJEiGNC1IFGVbm0WL\nsfzQsRKN5UWWPY7pJLYcc7yMPZNYtmacURJlGEdjT4ZeFckzsjyxHVoLJUqyJMuGKIKkSJEACWIh\n1kajF3Tty11O/rhVhepGL9WNLmz9vc+Dp7tu3b71daPq3POd5XdEVTEYDAZDb1hXewEGg8FwPWGM\npsFgMKwAYzQNBoNhBRijaTAYDCvAGE2DwWBYAcZoGgwGwwowRtNgMBhWQF+Npog8KCKviMgREfm1\nBZ7fKSJfEpEXROQrIrKtn+sxGAyGy0X6VdwuIjZwGHgAOA08DfyIqh7sOufPgU+r6sdE5O3AT6nq\n+/qyIIPBYFgDnD5e+w3AEVU9BiAinwR+EDjYdc4dwP/U+v5vgL9a7qIbNmzQXbt2re1KDQbDuueZ\nZ56ZUtXR5c7rp9HcCpzqenwauG/eOc8D7wZ+B/ghYEBERlR1uvskEXkYeBhgx44dHDhwoG+LNhgM\n6xMROdHLeVc7EfTLwFtE5DngLcAZIJx/kqo+pqr7VHXf6OiyNwKDwWDoG/30NM8A27seb2sd66Cq\nZ4k9TUQkC/xjVZ3t45oMBoPhsuinp/k0cKuI7BaRBPBe4PHuE0Rkg4i01/DrwB/1cT0Gg8Fw2fTN\naKpqAPw88HngEPApVX1JRD4kIg+1Tnsr8IqIHAbGgN/u13oMBoNhLehbyVG/2Ldvn5pEkMFgWGtE\n5BlV3bfceVc7EWQwGAzXFcZoGgwGwwowRtNgMBhWgDGaBoPBsAKM0TQYDIYVYIymwWAwrABjNA0G\ng2EF9LON0mC44Tk0XuCJFyc4M1tj62CKB+8aY+/m/NVelqGPGE/TYFglh8YLPPa14xRqPpvzHoWa\nz2NfO86h8cLVXpqhjxijaTCskidenCCfcsmnXCyRzvdPvDhxtZdm6CNme24wsLpt9pnZGpvz3pxj\nA57DmdlaP5dquMoYT9Ow7lntNnvrYIpSPZhzrFQP2DqY6udyDVcZYzQN657VbrMfvGuMQs2nUPOJ\nVDvfP3jX2BVaueFqYIymYd1zZrbGgDc3UtXLNnvv5jwPv3k3+ZTLeKFOPuXy8Jt3m+z5DY6JaRrW\nPVsHUxRqPvmU2znW6zZ77+a8MZLrDONpGtY9ZpttWAnGaBrWPWabbVgJZntuMGC22YbeMZ6mwWAw\nrABjNA0Gg2EFGKNpMBgMK8AYTYPBYFgBxmgaDAbDCjBG02AwGFaAMZoGg8GwAozRNBgMhhVgjKbB\nYDCsAGM0DQaDYQUYo2kwGAwrwBhNg8FgWAHGaBoMBsMKMEbTYDAYVoAxmgaDwbACjNE0GAyGFdBX\noykiD4rIKyJyRER+bYHnd4jI34jIcyLygoh8fz/XYzAYDJdL34ymiNjA7wL/ELgD+BERuWPeaf8K\n+JSq3gu8F/i9fq3HYDAY1oJ+eppvAI6o6jFVbQKfBH5w3jkK5Frf54GzfVyPwWAwXDb9nBG0FTjV\n9fg0cN+8c34T+IKIfADIAO/o43oMBoPhsrnaiaAfAf5EVbcB3w98XEQuWZOIPCwiB0TkwOTk5BVf\npMFgMLTpp9E8A2zverytdaybnwY+BaCq+wEP2DD/Qqr6mKruU9V9o6OjfVquwWAwLE8/jebTwK0i\nsltEEsSJnsfnnXMS+F4AEdlLbDSNK2kwGK5Z+mY0VTUAfh74PHCIOEv+koh8SEQeap32S8DPiMjz\nwP8H/KSqar/WZDAYDJdLPxNBqOpngc/OO/YbXd8fBL6rn2swGAyGteRqJ4IMBoPhusIYTYPBYFgB\nxmgaDAbDCjBG02AwGFaAMZoGg8GwAozRNBgMhhVgjKbBYDCsAGM0DQaDYQUYo2kwGAwrwBhNg8Fg\nWAHGaBoMBsMKMEbTYDAYVoAxmgaDwbACjNE0GAyGFbCoNJyIfJt48NklTwGqqnf3bVUGg8FwjbKU\nnuYPXLFVGAwGw3XCokZTVU9cyYUYDAbD9cBS2/MSF7fn0vqqXNye5xb8QYPBYLiBWcrTHLiSCzEY\nDIbrgZ6y5yLy3SLyU63vN4jI7v4uy2AwGK5NljWaIvJB4H8Gfr11KAF8op+LMhgMhmuVXjzNHwIe\nAioAqnoWMFt3g8GwLullhG9TVVVEFEBEMn1ek+Ea5dB4gSdenODMbI2tgykevGuMvZvzV3tZBsMV\npRdP81Mi8v8AgyLyM8AXgd/v77IM1xqHxgs89rXjFGo+m/MehZrPY187zqHxwtVemsFwRVnW01TV\nj4jIA0ARuA34DVV9su8rM1xTPPHiBPmUSz7lAnS+PvHihPE2DeuKZY1mK1P+t21DKSIpEdmlqq/1\ne3GGteVyttdnZmtszntzjg14Dmdma/1YqsFwzdLL9vzPgajrcdg6ZriOuNzt9dbBFKV6MOdYqR6w\ndTDVj+UaDNcsvSSCHFVtth+oalNEEn1ck6EPLLS9vlBp8MHHD7JjOL2s5/ngXWM89rXjQOxhluoB\nhZrPD79+26Kv2a/EkUlIGa4mvXiakyLyUPuBiPwgMNW/JRnWkkPjBR598jB/9a0zHDxbYLJUB2Cq\nXOflcyVmys2ePM+9m/M8/Obd5FMu44U6+ZTLw2/evaixmu/ZvjZV5pE/e55/+rGnefTJw6tOIJmE\nlOFq04un+c+APxWR/6v1+DTwvv4tybBWtA1MPuUyNpCkWA949uQs37ljkKOTFUSE4ayLJdJTYmfv\n5vyc59oGeSGPr9uznSrXeWWiDECx5ncM3VJGdzFMQspwtVnW01TVo6r6RuAO4A5VfZOqHu3/0gyX\nS7eBuXUs2zl+5HyZqXIDFDZkXL5xbJovHDzHwbMFDvbosS3n8Z2ZrTHgOa3Xq5B0LHKeQ7kRdtb0\nxIsTK/6duq/bxiSkDFeSXjxNAFS13M+FGNae7oz3hqzH63YO8upEmYlSg7GcR9oRjk1VSToWA0mH\nYj2gWAs4NF5Y1mtbzuPbOpiiUPPJp1yKdZ+BpEMjiMi2DN5qDV33ddusJiFl4qKG1WLGXdzAzM94\nb8h63LElzz/6jq381kN3cK4U5/eSjkWh5jNZalBu+Hzw8YPLxgiX8/gevGuMQmsr3jbIjSDiltG4\noWy1mffu60aqne8fvGus52uYuKjhcjBG8wam28CcL9X4yivn+fLL5zvJoG1DKXKew1S5wVS5yXDG\nZXPeY6bcXNSItOOYL50t8LXDk0yV653nug1hd+Ion3ZBYc/GLCPZ5KoMXZuVJqQWottLbsdzVxsu\nMKw/etqei8ibgF3d56vqf+nTmgxrRNvAfGL/Cf7uyAxDaZc33TxMwrF57GvH2TiQJO3anJmtEalS\nbYYADGeTHSMyP/HTTizdsy3P1w9P8ZkXzjGUccl5LsOZBL/64G2dc594cYKXzhYo1gMGMy7nSg2q\nfsidW/L88Ou3rXo7PD8htVJMob7hcuilI+jjwM3At4gL2yFWcF/WaIrIg8DvADbwB6r64XnPPwq8\nrfUwDWxU1cGeV29Ykrbheu5UgaG0y51bcojAofEiU+UGApQbAdVmSCZh0fAjSvWAW0azCxqRbg+t\nGYQ4joUTRBRrPjnPpdII+OiXXuX5U7OcLzVpC/+nEw5DaZd7tg9iWdaC8cMrGWNcq7ioYX3Si6e5\njzhrvtBkykURERv4XeAB4jKlp0XkcVU92D5HVR/pOv8DwL0reQ3D4nR7hZFGoBZffeU8pUaIbQme\nY1Fphgx4Dp5jUQ+UVMJmOJNguuoztoAR6fbQjkxWyKdcNg4kKTUCtg56/O2rU1SbIapzW8hqfkhY\nVo5NVnjdruElPdjuGONqSpJ6YTWF+gZDm16M5ovAJmB8hdd+A3BEVY8BiMgngR8EDi5y/o8AH1zh\naxgWodsrzKcSzFabTJYaRIBrW9T9kCBUNuWSRK7NVLnBTLnJDA3OFesMJB1++J175lyz7aE1g5Bj\nk2VUFceyGPAcnjp+AT+KTWU0by3NUAmikCOTZbYNeXzzeJ0zszUStiDAc6cKuLZw55YclrgL1l4u\n5Im2z1mpd9oOW3T/7OWECwzri6UGq/018f5qADgoIt8EGu3nVfWhxX62xVbgVNfj08B9i7zWTmA3\n8OVFnn8YeBhgx44dy7ysAS56hVPlOtPlBscmKx1jZlsRICjKeKFORJwRtASCCBpBRKHWvOSaD941\nxr974hVOTlexBQKNz22UG0QKGoGIwAKbkkih5kd87qXzJG0h79mcLzVRINKIvOd2Cu8hriWdKMVv\ntz1jGb54aHKOJ/qRzx8mUmXnSGZV3unlxkUN65elPM2PXLFVwHuB/19Vw4WeVNXHgMcA9u3bt6Iw\nwXpl62CK16bKPH+6QKHqz3nOD8GJNaUpNcLYiKVcpGU0hzMufqh8fP8JRge8OZ7clrzHTKVJqBGl\neshYLsFEsU4ziAgipZf/nEaofOP4BcZyCcZyKSbLDRAh6QgvnC4QRPFVxgbiTPtHv3SU2zZl8cOQ\nbx4vUqz7TJcbpFybajOkWI9jqptySdMZZOg7S02j/Cp0pOHGVbXeepwCeqkVOQNs73q8rXVsId4L\n/FwvCzb0xoN3jfHIn52jXA9wbUEk9iYjjf/5enEusx9pHHdUJYyUajPAEpgo1Xnwzs1zPLlS3efN\ne0axRJgs1TkyWeHsbI16EGEt7GQuiALnS002ZJMIcPpCjTCKaATK2EACL+Fw61iWfMoliJQj58uE\nSqcQ//RMlZmKj+fa5FMudT/k5XMlqv6C912DYc3oJab558Cbuh63peFev8zPPQ3c2jK6Z4gN44/O\nP0lEbgeGgP29LNjQG3s359k2lGKqFcdMuhbNIIJWkqY9wN4RCBXKjRARcC1QgZqvCAF+GM6JM56Z\nrXFiqsK5UoNyPaAZBNSaIZZAwhbCSPHnBzUXIVJ48WwJWuuwrNgLPluosSGb5Mj5ChcqTarNgJMz\ndYbSSTbmkogIoSq2JZQbAYPpBJ5rt8IKwTKvajBcHn2ThlPVQER+Hvg8ccnRH6nqSyLyIeCAqj7e\nOvW9wCdXmp1fj6y0LOfOLXlOTFco1gIStk0YaWfr61iQdGxEoNIIYyOq8fZcVLGAhG1x5HyFDdk4\nYz7gOVgCz52aJZN0sFBOzdQIFLKuoCKEEdhoxzD3SqBAGBtPVaEZRMxUGjx/6gJBqAQKU+W4Y2nb\nUBpBSDq0svVKo3VDyHk9dwYbDKuil3fYpIg81DZyK5GGU9XPAp+dd+w35j3+zd6Wur5ZTVnOnrEM\nn3w69gQ9V1C1aQQBNjCUdgHBEmgGEc0wNnGh0rF2hbrP0ckyt2zMsCHrcWKqwrHzZepBxPlSnWZw\nMYZZC5ThjEsY+YQaG2BLYm9yxcZTlamKz1TlYizWkdhDrvoR54sNdm3IUKn7+JFSboRkPYedw2l2\nj2YXvfZSmF50Q6+sVBpOiDPiP9HXVRkuYTWSaIcnKrxx9zBHJyucK9ZxHSHnOagqg+kEpy/UiaKI\nMFLsloHrJlSYKjf5f586hd0KgEaAY8mcpI/VOne25uOHXRfReIuxFlHGSMGx42vO1nyCKKIeRGzJ\ne9y9LYfnOqtuzbzSdaKG65teBqsdBd4oItnWY6N2dBVYTevfmdkaO0Yy7Npw0fs6X6rxuW+f4+RM\njWYQoVz0CBdDWuVFbZrh3JPbIUw/vPQia5WWiYBm18VSCZuhtEuhFvCVw5O8/baNqzZyRqPTsBJ6\n7T1/F3An4InELoeqfqiP61q3LLZN7LX1r/vnT85U8YNwjtGsNmLL4zkWYRjFnqNAcwmjuZxRvRrM\nlJsMZRLkU3FHzxcOTvCtUwXu3Z7nx+/fuSJjN/+G9OpEkW+dKlCo++w/Ns3779/Bu+7e2o9fw3Ad\n0kvv+X8m7gt/G/AHwD8Bvtnnda1Lltom9tL6d2i8wEc+f5ipcoNGEBJGysnpCgA7RjKU6gGvTJTJ\nJh0iBc+1aAQR5cbcjPP8LXXQYzb8ShIoTJebTBKHB7JJm6lynf/2fJW/PzrNL79zDzeNZufcgPaM\nZTg8UVnyhvTqRJGvH5nGEiGXdCjWfD78uVcAjOE0ACDLJa1F5AVVvbvraxb4nKp+z5VZ4lz27dun\nBw4cuBov3XceffLwJd5k+/EjD+xZNlnxv/7FC+w/NsOA55B0YoM4VWownEmwadDj3GydY1MVwigi\nUkjYUPX1mvMiV0PCFobSbtz3rrERvXnjANuH0wx4DienKzx7cpZ7tw+yc0Omc9N5+M27ATo3qy+8\ndI5qM+7P35z3yCTjWGku5fKpn73/Kv+W/cEkwWJE5BlV3bfceb1sz9tBs6qIbAGmgc2XszjDwiwX\nt1yu9e+5UwWySRvPtQHimskw4vh0hXIjoBlGZJI2DR8K9XBOjPB6xxLFsS1UoRGEzNYCpsoN7toa\n/73OFRtkkg7nSg12j2bnxC0feWBPpxe9UPfJJR1GskkyyZbKfNJmolhf9LXXkittwEwSbOX0YjQ/\nLSKDwL8HniWuIPmDvq5qnXK5kmWz1QYzFR8/vJjgySRs/DBivFgjjOLHIhauFfZchH49oK3+pkYQ\nxt1NkXL6QpWpcp0NWa81csOm3KVkv9ANaf+xaYo1v2MwIW41HcvNvZn1g6thwEwSbOX0Mljtf1PV\nWVX9r8BO4HZV/df9X9r643JGOXzmhTNcqDZptDLiUatQvVgP8QNFI3BtwQ+VSBXLurFE+xuBUqr7\nFGsBtlhkEg6C8MyJWabKdXKeS6lVz9lmoRvS++/fQaURb92jKKJQ86k0At5/f/+FYq6GorwZVLdy\nekkEpYFfAnao6s+IyA4R+R5V/XT/l7f+SLkWTx2fRhDu3Z7v2cv42P6TJGybIAwBRVtF5Urcr21Z\nELU6ghxLiKIIIS68bYc0r/fQZrEeYkt8c/Aci4Rr0/BDvvzyeWxLKNYCNg0kiFQX1dBsJ3s+tv8k\nE8U6YzmPD7z95iuSBLoaivJGkHnl9LI9/2PgGaAdBT9D3HtujOYirCYu1b01e8fesc6HulcminUc\nSxhMu9T8EPXDTlLEc61YiKOVBg+7CtOvd0M5HyHWC004FtsGPb51apaaH+HYQtK2ePFMgRMXaqRc\nh3u3L64ef/9NI1c8IXI1DNiesQwf/fJR/DBiJJNgc97DsiwjyLwEvezRblbVfwf4AKpa5aJAjmEe\nq510eLlbs7GcFwtZRHF0z7bipIgIJByr41XCjWcou2n3qJ+ZrfH3R2coN9vxXaEehMzUAqJI+d69\nG3Fbs5IOjRf4zAtneOSTz/OZF8Y5OV3h+GT5ik+oXItJmyvh0HiBLx6ajAfeZRLMVHxeOVfmHXtH\nTTxzCXrxNJstOTgFEJGb6RIjNsxltYH1y92avf/+Hfzr//YSxbqPY8WK6CJgWxCEcW/5YMqhGUbU\n/XhrHtyg1jPUVg99i0jj/vr2sZMzNZ46PsMtoxnyKZdP7D/BMydnQWIt0UYQcfh8mT0bs3M0RZO2\noMQdUavJbC+3A7nSivLd79V2z36h5nN4osK7+vKKNwa9GM0PAk8A20XkT4HvAn6yn4u6nlmt8bvc\nrdm77t7KZ54f5xvHZ6i0pNp2j6S4c2uebNLh8wcnCMKIIJprUNYL81riKVSbPHvS5zu253nhdLGz\nPRWRTsnW0ckypUbI22/fiGvD/mMzCPD63UNLZrYXG83RS2b8SirKm6mcq6OX3vMnReRZ4I3EO7x/\nqao9qRytR1Zr/Bbr+Hn9rkEeffJwT/HRtOfyY2/ciSUXoyeRxiMt3nnHGE8ePMds7QaqM7oMyo2A\nDdkkh8ZLKMpIJkEjiDoGM+lYnJyusmMkTT7l8o1jRWwLirWALx46z02jWTYNXKoUv1jZUMq1rkhp\nz0ri6SYJtDp6qjtR1WlV/YyqftoYzKVZbVyqvTXLp1zGC/VWQmiULx6a7Dk+unUwRak+tyWy/SF4\n0y0jXKj2nli6kUnYQt2PdTjPlxpYIowXahyfqvDqRImXzxX59pkCpUZAGEZMletMFGvMVHwUjfU7\n/ZDDE2VeOjv3/2Kx2PRzp2b7Xtqz0nj6lY6h3ijcWMV61wALGb+VDPt65IE9fOQ99/DIA3s4PFFZ\nUXJoqQ/Bp58fJ1KTwfMcIZt0WsXvNYq1Js0gnswZqlKs+ZQbAWGkeI5Q8UOeOTFLpRG2kmlC0m11\nXQkU592kuuseJ0t19h+b5qlj00yWGpyYqsw5d629upUmEy/nvbqeMTLXfWCt4lIrjTktlEh4/a5B\nnnhxgi+/MgkCVlf95joMbeIHSuhEJFwbP1S25D3SCYfzxUZL+UmxLLh5NEvdD5mp+IwOxF1VqhBG\nwpYBr+Op5lNzP0LdY46fPTlL0rFwbWEkneC5U7MAc3rf17K0ZzUxSjOVc+X0Kg1nEw9T65yvqif7\ntaj1zHxpt6YfzlEjX8476f4QdMfXVGOhYZ/1aSzbRMS6nyJK2rUYysQzhyxLEBSRuI71XKGGHyqV\nZki1GeCHEVsHUwx4LqEqnmuzayQ9R3YPLsamj02WSbaUm5uhsm/3EOV6wHixQcK1+5IZNzHKK0Mv\nHUEfIM6gT3BRb1aBu/u4rnXJ/CSC3/JWYHXeSXu7dr5YI1JdVOLN4uJ/7I2OY4HjWIxmXSp+1En+\n2NLqKGqN1CzUAvxQSdiQSTg0gjgO+h3bBzsyewvF/9re/i996oV4nnsqwV1bc2zIegxnFNex+ch7\n7unL79aLfKDh8unF0/yXwG2qOt3vxax35td4tr2Y1XonZ2ZrVOpN/u7oDG5rRMV8LK5NkeF+4Ufg\n1wJKtQARmCo1GPAc6n4c14yi1lA4jbum/BAqzZCNA0mSjs14sYHrLP1/sXdzngfuGFuV13c5KkdX\nus5zvdKL0TwFXLm2iHXMQjGpHSOZVXknh8YLnJyp8q2TM4TR4kYxYv0YzG7a0zejSJmtxsPgkjZY\nlkXdjzohjAhoBBHjszUG0wkc+2Lu9NhkeVEDtxqvb76I9KsTJV48U+CX37lnRYbTGMn+0ovRPAZ8\nRUQ+Q1cnkKr+n31b1TplNTGppQqp047cUJqZa838G0YzBBe9JOZrSeyhTpab5NNx6OS1qTJ/8ezp\njqjx/GL11Xh9n9h/guNTFQY8h5wXdyYdn6rwif0n+O13m2jYtUIvRvNk61+i9c/QJ3odadH+ICZs\nYaLYYPtwek5dXrpVSH12tobF+s2Ur5R2i+Qlx7vGECcdG0ukI2p8dKrCuVKDcj3AtYWP7z/Bv2kZ\nuJV6ffNFpD3XRlV57pTZ6F1L9NIR9FtXYiGG5WNS8xNFn3/xHOdLDY5PlRkd8LhlY9xL/c3jM3zv\n3o0U6z6phEXNj8tlDKujbUcFOFesM1WuU6z7WCinL9TZNZImm4xl6L5+ZJpD44VVbZEVReZV0krr\nuOHaYVGjKSL/UVV/UUT+mgUcFVV9qK8rW6cs5p0cGi/wwccPMlNuMpxNMJJ2mSjVsUWo+xH1VhH2\nvTvyKLFeZM5zSTo2zSBal3HLtcaxwBbh749MU24JFTt2S1lKBCSeU7Ta1sh7tw/y1LEZEOnMeCo3\nQu67abgPv41htSzlaX689fUjV2IhhoU5NF7gE/tP8LdH4jEMW/JJGn7IN18rY4vg2EIzvNgzfeC1\nCyQdmy+/fB4h7nC5UdWMriSOAAieazFTaRJGUSfbfnSyzMaBJKmEw7078qtujXzf/TsZL9SZqTQp\n1QMSjsWOkTTvu3/nmv4uhstjUaOpqs+0vn71yi3H0E17O35sssxgyqHhB5wtNtg2mIpHVojSDIkl\ny1SpNnzOFBo8sHcj6aTN51+cWJeKRmuNJTCUSZCw47KtSBU/1LhUKQhpBBGnL9TYvSFDtRFeUvDe\nK3s35/nVB28zkyGvcUwb5TVMu26zGUYMJB025jxOzdSYLDVIuRblRshw2mEg5VJuhBTrIdsGU50P\nrWXFnlEQthTbdX2WF10OtsCGbJJNeY+ka1Oq+9SaIaEqYRThh4pjxT3ps1WfZ0/O8n13rl7w4kYp\nGbqRxwIbwY5rmLb4Q7v8JJ1wGMm4lBoBs7W4tW8kmyDl2jSCkHoQcvNouiMUUaoHNIJYQ9OxBUti\nrwnMf/xKqDYDpksNbtqQJgyVYj0gaVvxDUjjUqUgUoIo4t7tgxyeqCx7zRuZ1U4vuF5YKhHkAQOq\nOjnv+ChQUtUrMwh6HdOu2xzJuDx1vEwzCGmGESnXZiznMeDZHD5XIZdy2TLoYYvwwukiCccil3LJ\nJW2mKrGXGani2nHhtgU4NoQhmDLOpUklLIYzCWZrAV95ZZIwUhItBfd6EGFZcZG8bQm2ZZHx7HUv\n4nujjwVeyuH4T8D3LHD8u4FH+7McQzcP3jXGqZkqL58rM5yOhSKiVnfPzaMZbMtmy1CKLUMp7r95\nA/t2DVFphpTqAUnHYjibJOkIFuAHEEYX56EHESSc9S4UtzSuJWwa8EglHDbnPXYMp8mnEuRSDrVm\n2CmAdey4UGg443LwbGndC2Tc6GOBl4ppvk5VH55/UFX/UkT+9z6uydBi7+Y8Y7kkU+UGfqikEw5j\nI0kc22K64lOs+wwkbcotTcfRAY+hjEux5lNqBAymE9w8muHEdJVjUxWiKBagyCZsivVgwUJuw0X8\nSHltpsrWwRRjOS+OIWcTOLZQ8yNytkWxHqAaYVsWE8UGjaCG51g8CjdUHG8l3OhqS0sZzfQSz5mQ\n2BWiGSpv3jOKJcL+Y9M0/JCkY1Gs++Q8l0LNJ9f15sx5LjnP5a23bewc25hLkfFc7tic6+g8Vv0K\n9S7Zo+75523Wk5DHYkQRjBfiSFTWc6k2Ak7PVKn5IemETdIWan4cL677ISnXplALWtMsq+tS1PdG\nV1taymieF5E3qOo3uw+KyOuByUV+xrBClssydt+1bxnN8OzJWRpBRM5z2JRLcna2xp6NWSKNC9qH\nMwksEQo1v/OGPTVT5fRMhWOTZQA8O9aO7GYh27jeDSbE/elhpIwXamwRIdES7HAsoebHA+w8145n\nrduwdShoeYP3AAAgAElEQVSFbQnnSg3u2Jy7YeJ4K+FGV1taymj+CvApEfkT4JnWsX3ATwDv7eXi\nIvIg8DuADfyBqn54gXP+B+A3iT+3z6vqj/a6+OudxYZwdXsn3XftkWySPRuzHJ4ok0+77NqQ5fvu\nHOPwRKXz5vzVB28D6Lxhk7ZQqvtUmxGWxPG3yYqPBbhWLERhWJpQ46TZqZlqPFPettg+nMa2hJMz\nVeyWAb1pNEMm6aCqlOvBDRXHWyk3SunUQixV3P5NEbkP+BdcHNn7EnCfqp5f7sIttfffBR4ATgNP\ni8jjqnqw65xbgV8HvktVL4jIxoWvdmPSS5Zx/l1792iWf/62m+e8IRebUf3EixM8eXCC6XKDnOdQ\naYat3mbFjxTbEm4ZSVHzo1aXi5J0hGojMln1BWiHgMOW+lDCEcIIEk7b+4y/NoKIbMvLv1HieIaL\nLFncrqoTIvJvgVtah15R1cZSP9PFG4AjqnoMQEQ+CfwgcLDrnJ8BfldVL7Reb1ljfCPR60yXld61\nuz3YSCMaQUgQKYMph1oQxWZT4zG1I9n49avNEM+xKDWC1gAxo4y0GEpsQGu+YhEnORxLmK02SSUc\nUNg5nF5SoepGK/heTyya0BERV0T+I7EI8R8DfwIcE5Ffaz3/Hctce2vrZ9ucbh3rZg+wR0T+TkS+\n0drOL7SWh0XkgIgcmJy8ccKpS43cvRy6Pdh8KkHSsRHiusLtQ2l2jmRwbIuEHUuPXag0qTR8psoN\nKo2QAGMwe8WywA8ihjMJbEvIpVy2DafYPZqdE2a50Qu+1xNLZcH/A5AFdqnq61T1O4G9wE0i8n8D\nf7kGr+8AtwJvBX4E+H0RGZx/kqo+pqr7VHXf6OjoGrzstUG/5k5318ndsjFDOhFPU6w2Q+rNgCBU\nbhnNkPUczhZqnCvUsC2LKDLGcqV4rk066TCW8/jOncP81kN3cOeWWLTjiRcnOkZxpeN1DdcuS23P\nvx+4VfWiEqOqFkXknwNTwD9c5tpngO1dj7e1jnVzGnhKVX3guIgcJjaiT/e4/uuafmUZuzPuG7Ie\nb7plhAPHLzBdbdKMlPtuGu4o53zw8YNU6iVsS3EsodQw0cxecax46JptCdOVJrdszC6a2FvNeN21\n5DMvnOFj+08yUawzlvN4//07eNfd8zd+hl5YymhG3QazjaqGIjKpqt9Y5tpPA7eKyG5iY/leYH5m\n/K+IPcw/FpENxNv1Yz2v/gagH1nG+XVyrm1zx9b8JTWDh8YLlOo+EbH4RMK2caxw0amVhrkEERRq\nPsNpF9e2EFg0sXc1C74/88IZPvy5V8gkHTZmExRrPh/+3CsAxnCugqWM5kER+QlV/S/dB0Xkx4FD\ny11YVQMR+Xng88QlR3+kqi+JyIeAA6r6eOu57xORg8Rt0L9ipl5ePt0e7MHxAoVaQM5zeOLFCY5N\nljk8UeGp49O8OlGm2vDxIyWMoEzEQo2VJim0OKpKuRHy/u/azIETsxSqcTdWznO5ZWOG4UySM7M1\nfvq7d121gu+P7T9JJul0GXOrc9wYzZUjCziT8RMiW4G/AGrMrdNMAT+kqvO32leEffv26YEDB67G\nS1/TLDVgLZ9yGfAcXjh1gedOFUgnbGarPkI8Cz2IFh6o4Fq0etUFPzRDFxbCtWFLPoVlCcXWzWl0\nIEkjiGeq3zaWZdeGLI88sGfJ7Hk/M+tv+fd/w8ZsAsu6mMKIoojz5SZf/ZW3rclr3AiIyDOqum+5\n85aq0zwD3CcibwfubB3+rKp+aY3WaFgjFiuST7UGrOVTLpOlOs+dKlBvCXq0DaAjYFtcsiV3BMII\nROIebMNcLOK/TRBCseZj2xYDSZvpShPXthhMx3J+r0yU+WdvvRlYepTJck0Ol8NYzqNY8zseJkCp\nETKW85b4KcNi9DJY7cvAl6/AWgw9sJBHsliR/FPHp3nH3tjj/Obx6TnGsk2gsQHIJm3qfojn2tQa\nIRGAgGtb8YyhK/YbXh9YVnxTsQSqfsiAbbEx5+E6FuVmgGNb5DyHXMpd1vD10uRwOZ7o++/f0Ylh\nDiRtSo2QSiPgA2+/ebW//rrGKLdfRyzmkZQbPrdvys05d8BzEIRSPcAPQ05dqKEsHJ9UhVozRInl\n0JyUQ6TQCEJsywh3LETbM7dbM5I35TwaQcRQOoFrhzxwx9gliZ/FWC6zfrmeaDtu2Z09/8Dbbzbx\nzFVijOZ1xGIeyZnZGqV6cElmdtdwim8cnWaiVCdo9QC29TS7Z3krsVFMu4IiNIOIXMrGcyxma77J\nBC2C3Rq2tnUoxT3b8zxz4qKYSrvmtpdEz3KZ9bUQ9X3X3VuNkVwjlpV4E5GMiFit7/eIyEMisvzt\n07DmLCbu2v0hbRfJn5iu0AiV2zZlsaT9AY/LGOx42iwQvwEyCYutgx6DmSSDaYcBzyHvJXjgzjEG\nvLicxrAACn6onJyp8YWXJsgk4r9TrmXgHn7zbgAeffIwv/znz/Pok4cX7ABarsnhRhf1vd7oxdP8\nGvA9IjIEfIG4/vKHgR/r58IMl7KYR3LnlnwnttmOeW3Je7iOTT7lcq7Y4EKlyenZajzuQgTbBtey\nSCcd3nbbKGO5i/WCE8Uaf390Bte2ecOuQb5+xFSBLUQIJGwAZarc5ELV5403DfFL37eHvZvzS26r\ngTklYRbKmVkh5zncuSU/p8nhRhf1vd7oxYUQVa0C7wZ+T1Xfw8VsuuEKspRHsndznkce2MNH3nMP\njzywh0ZrxCzErZSWJWzNp8h5DumkQ8K2uWdrntFskgOvzbL/2DSTpVhs13MdvvuWEfIpl4yX4KF7\ntpAxe4sFaYatKgPims2DZ0udnvLFWic/vv8Ej33tOK9NlTk5XaVY85mtBmwaSDLguZckefrVbmtY\nHT0ZTRG5n9iz/EzrmN2/JRkWo120nk+5jBfqnS3gQnGtbjGQDVmP1+0cJJ102DyY4ifu38X/8e67\nGM2nWrOG4rKZZ0/McnyyTKHm8777d3aM8JtuGcFLJOJt/QrWu5JzXQsS12kUINQ4m67EHmAURR2v\nf6Ft9XOnZjs7AM+NdwNJ1+JcqbFgP/pK/t8N/aeX7fkvEmte/mWro+cm4G/6uyzDYvTadrlQK+VN\nXco7jz55uOP5ZD2HI+crTJUbnCs1+K2H7uhsLz++/wSf/vY4loLnWvihLlno3p0zWknuyI/iO7iz\nQM3o9UB7zZYoR8+XcR170W21IAx4TmvGU/wRTDrWksLFN7Ko7/XGsvd2Vf2qqj6kqv+29fiYqv5C\n/5dmuByW8066vaANWY833jTC9/+DzewYTs+Jx710toiNkHAtXNvCc60lDabdlXRaiuQCt2slnsnj\nXqceJ0DatXltukrSlkW31fduz1OqB5159mCEi68nlpp7/tcs4Syo6kN9WZFhzVjKO0nawtcOT9IM\no06ftGvbl5S5+KHiJWLZuKRjU6z7nTEZQly+1FY0V+Ji+aGUSyZhcbqwuF51znO5UPEJut5hrYm4\n0PI6rzeH0xFIODaBxp74YipWELe3bsoleflcKTaciwgXG649ltqef6T19d3AJuATrcc/AhgRwOuY\nQ+MFzhbqlOsB2aRNrRmw/+gMuzdk+OF37gEuFlxnPYcgjJiuNHGs2CNybcFSxe4ymNDatgiUGz51\n/+JxWy7WgrZOoVgPiFh4TlEEZF0h0LhmtG1Mr7Vy0YR98XdyrFjvXoH7dg1dMh55ttrk5EyVp45P\nEymIKirCUCZBpJDz4lIvBf7w668ZZfdrmKV6z78KICL/YV4T+1+LiFHMuI554sUJdo5k2Jz3ODJZ\n6cTSxnLJS8pcbhnN8GzNZySTYLba7LiDW3JJaoFSaTRpBrGhs614ey0tQ5h0hEbLlRQueo9Wy/ol\nHYv6IpPdmhGkXIuhtMNUxSfpCLVraHaRLbBzJMOFSpNSI+6munVjlru35Um0Sr3aIY4wjDg9U6MR\nhEyVG2wYSJK0bfaMZbFtq1OC1C5PGsk6a95/blg7ekkEZUTkpq5ZP7uBTH+XZegnbS/SEpfRgbh9\nL1LtzPeGi4mkKIqwBc6VGijK1iGPKIKRgSRBGHF0MsCxlUzCQVHqfoQIaKStuezNlvCHIKI4ArtG\nstw8muErhycX9RyboeI5SrUZkbQFP7x2tuu2wO4NsexbGCmRQibp8LbbN86RfGuHOA6OF0m6FsWG\njyXCVKmBY8eP9+0Y6mTLL7frx3Bl6MVoPgJ8RUSOETsMO4Gf7euqDH2ll2LpvZvzvGPvKB/90lGC\nSNkzlmVTzqPUCJgs1Rkv1Kj7IZYF+aTLLaNZDpycJYqizlb8QrWJLUKoitfaygYRHJsqc77UoLnM\n/OBGEBIpOLZFozUQzoZLvM0rtW1v57eSjsUbdg93bjivTZUZLzYYL9TnxC2/cPAcKEyWG4wNJKk2\nApphhCpkkxa1ZsjhiTJVP2Qwnbiqyu6G3ulF5eiJ1qjd21uHXl7BRErDNcj8cqTFBHEPT1R4480j\nNIOQI5MVjkxWCMKIUr3J5nwqVj9SZfxCladPXOjE8dpZdEFohErCElRjGbVswiIkVjxfztD5YWwM\ncymbuh8b44W251cyzjngOQynXc5cqPKVVyYp1QMSjsXDb97Fz77lViBWSv/ol48yUazjWoKqcma2\n3vp7QcIWIlXSCRskju/euSVvun6uE3oV7LgVuA3wgHtEhPmK7obrh15nE52ZreFY8I2j09T8iFCV\najNAI+Vtt48xOuAxWarz17N1gjCYo4akQNh6kHQtcimXdMJmOJPkmRMzwPLqSRGxiIjn2oj4LKKX\nfcWwgJFMgru3DfCFg5M0WobcDYTHvvYa24bS3DSa5aNfOgoCWwY9zl6o44dxyMIPY8/adh2CKL4Z\nqCr5lNPzjcxw9VnWaIrIB4mnRd4BfJZ4oNrXAWM0r2N6KZbeOpjis98+y2w9IGkLCVsohooIPH+6\nwDv2xokkxxYSTit7rFHcWqixt+m2ssquLQylE1SbQacQfKlyTovYqAZRbGxsib2zq8lAymE47fC5\nb0/QjOJC/LaYSaHa5NEnX+Vdd28hiJThjIuIMJKNGC/UqTdDkDhW6bk2ri0MZ5JsyiXZtSHbtyF7\nhrWnF0/znwD3AM+p6k+JyBgXy48MNzAP3jXGH//dcVxbsC2LMFJsS0jYFhPFOGlUrgdxCVLCoeqH\nBGEri66QcCw2DiQp1AJqzYjZapPzpSbQ7tVe+HUtIJ2wSNgWpUbIdLlJeIXt5fxyqsFUPKb31ckq\nzY7RF/wwImFb2JZw6kKNM7M1hjNx0XoYKReqPtlkbCRHB5LU/Yjv3DHIjpFMx5ts95Cbrp/rg16M\nZk1VIxEJRCQHnGfuaF7DDcrezXk2ZBOU63ECI+lYbB1MMVlqtLLGimsLrmWRSVhkPScWO24pmm8c\nSJJKOGwb9Dg4Xub4dBXHEkbTDjO1lop8l66nIzCYjsUtqs2QdNKhUAtQmVu+dCUIdW6CqdoMOF+s\nk+iSyQtVUYW6RjhWfP7WwRRNP+Tw+TIzlUZsfCOwLYvX7xqmXA8YLzY6bZZX25vs52yiG5VejOYB\nERkEfp94wFoZ2N/XVRkW5Gq8we+/aYT9x2YY8BySTpzFrreyveOFOnduyTFRbJBJ2Jwr1pkpx7PV\nR7MJxvIpNmRcXj5XZnQgSVRSGn5EsRkxnHZALIr1AD+M2DaYIlRtiSUrmaRDBAymHUr1kGaone38\nQqbTFrAtuaSo/HLovlIQwoVagNMVU+jEbzVWOxpICgfHC5yarjGWSzBVahCqYolw3+4hRgc8RrJK\nwrX5yHvuWbN1rpZ+zya6Uekle/4vWt/+ZxF5Asip6gv9XZZhPlfrDf7j9+/kXLHBVLlBse6TdGxu\n25Tjl9+555L5NbUgZMeGDNPlBiPZJDePpjl4thRnjS3wHIukbdEMI8rNiMGURSZhU6wrjSAkm7SZ\nbsSxvy25BEfOl1HiGsggjKj6cblOu7c9bM03aouO5DyHwxOlSzqMVkPKgUaraF+4WCM639lte6MC\nDGeS3L4pR9q1eWWiTNazyacS3LU1x4ZsXE50LWXE10IRfj3SSyJIiGXhblLVD4nIDhF5g6p+s//L\nM7S5Wm/wvZvz/PI79yzp4ba/f/FMAccSPMfiXKHO+VIDz7XIpxwc22JAlVMzVWrNkEiVmWpA0rZI\n2nF7Zj1QRrMunutQ8+MaTcuKt8ZBV3F7qHGM0bYshtMupUaAYwlVP8RzLPzm6q1mW9U+1Ngczu+v\nb+NKbEDtVjJo40CCdNLBEmHXhixDmSR+EFL1I1zbJlJd84z45e48lptNZFiYXrbnv0f8fn078CGg\nBPxX4PV9XJdhHlfzDd5LguIT+0/w8niRIFJCVWyReMplwqHmR2SSFiKCa1s0rIggUGxgx0iaZhD3\nto8OJMl5DndsyXNiusLpCzX8RSZhFmoB+ZTTKRYv1ZsU6yGqSrJVSL8aj9MS8FybIIol8OYnhOJ+\nc8ESSIpwy8ZsXIBvQSoRf5ymynVenSgzUWrw3a061/FCsKYxzLXYeRhF+NXRiwjXfar6c0AdQFUv\nAIm+rspwCd2iwm2upTf43x+dptIMASVhx15ase63un/qHJusUGn4BJGSTjh4roVjw7linXIjwHPi\nkbcTLSHeLXmPjGstWpdkCdT8kLOzdRJ2bCAdS/Cj+HslTiylHGtFYshBK4mVSVgd1aa2zRTAsy2G\n0i6phEM+5RBEsYBJuRFyy8YMU+U6z5yYpVgPGBtI4jo2NT/ip797F488sGfNdgWLqcLPFzBeCqMI\nvzp6MZq+iNi03jsiMsq10wa8bujnG/zQeGHZ4V/LEW+R45ZHEWmNAI4II7j/pmECVV6bqmK3MuGx\nFxeXFcVb84h0wmYs53Fmtsazp2ZJJ51OWdL8N2rUGmqWTthUGiGVekDQysy02zUDhVpLJalXBKgH\niohFwhZsoXUTiD8A9VAZTLsMpl1yKRdLLO7ckmP3hlha79WJcudat2zM0gxCjk2W+aVPvbDqv+1C\nrMWwNaMIvzp62Z7/J+AvgY0i8tvEdZv/qq+rMlxCv4qf1yrBlPMcZspNbCuu5aw2462yl3C4fXOe\nkWySl84WKdV9CrWAkYxLMySWflPFcy2ePTnLW/eMsjnv8epEiUItIJuwqDQv3aJ3pmmKUmkNi/MX\nyZz30psurX+OHW+9Z2s+URQX8iccC8dW6s24DXIonWBTzusoFB2bLPN7XznGgRMz1P2IbUMe9+0e\nAeDZk7MkbSHSaE2Td2u1tTa1oSunl+z5n4rIM8D3Er+v/pGqHur7ygyX0I83+FolmO6/aYSvHJ4k\nCJVmq7Dbc2y2D8cf4tEBjzfvSfLyuSKHzhaZKNY7W+F0wo5jiY5wrtTgyGSFuBxSsSyLdDLe/rZp\nj8WINPYKU65DzQ/QRfY/S5UqMe+5IFS8VmmVtrLzSccChNFsbMB3jGQ6iZdjk2U+/LlXyCQd9mzM\ncnyqykSxwWy12ZK0s1p/18SaJu9M2+XVY0mj2dqWv6SqtwMvX5klGa4ka5Vg6i5NagSx+lHCtvkH\nWy8ah5PTFV48PctksUk7OhtqXCReagQkHZuGH5ceNYKIbNJhtharGc83fKOZBLVAKdQDUi6ESxS+\n9xpLal/BsqQzq8i2hIYf4rkOhXrI99+1iX/z7rs7P/PBxw+SSTodg7gp73FyusozJ2ZJJ21mK00q\nfkjatak2Au7aluPMbLDAq68M03Z59VjSaKpqKCKviMgOVT15pRZluHKs5TavuzQpaQtnC3USzsVy\nm+dPzVKoBx2lova2udaMZd9cO2oZ3bj7yEvYZKKIrYNpzszWsEUY8GzOFxuMl5o4ljCQtBnwHC5U\n/UXXthLpOAVmawEOsZcZREoYKQlHSToW77t/J3Cx3OelswXynoNjCZmkQybpsG3Y4+RMjXLDxw+V\nbNLGsS1OXahxodrkLXtGV/S3XQyztb469BLTHAJeEpFvApX2QTMj6MZgLbd58z/E8+sIxRLCSLEs\nLrZP6sXi8LqvJJ0IIWKmEuCHcZvm3s0D3LU1xzMnZlFVmmEUG0zPZSSTwLKE4YzLdGVhw7maHqGg\ntSZHYgHlSjNk2BGeeHGCY5NlvnhoknzKJec5VJsRjaDO5rwXdzKpkEu5ZBI2k+Vmq3dfiCwoN8Nr\namSHYeX0YjT/dd9XYbhq9HOb121ED40X+NOnTtBsbaMtiUuEolb/NkA2aSOilGohti1kEnFs8YmX\nJhhMOWSTLueLcd970rU6RqruhyQdGwsfkbmziHrdmi8U9xTiLbrnxAXsw9kkhZrPR790lNs2Zcmn\nXO7dPsjXj0xDBDOVBkGkVBoB2wdTFOsBW/MeszWfRhCRcCyy9tq2ehquPL0kgr56JRZiuHr0e5vX\nztCPZBIUqk38KI5lamt+ertUaTDlUvMj8mkXQagH8UY+CpVaMyTnuR0hj40DSTJdM8ODMMRuDR9K\n2ELYKkmC3iZbWq3hb93KS9r5J2wfTqEa10MGUTwaZNeGLLeO5QB4rhV62Dac4QNvv5nDExW+8sp5\nALYNpQGo+/Hvc63U1hpWR68ixAbDgvTSytfO0L9u5xDlus9kuUGzpcruWLFk2ndsy+PYFn9/bDru\n6nEt7FDIpxJkkzZniw3qMzWUOLttW1bLw4xFP0IV7tme48j5CtVmiN8yyAJ4CYtac/F6zU7rZJdl\ntYhbOMNWobwfKoOZOO4bhwKanXNvHcuxMZcin3J55IF4mudN4wVeOD3LyekqtEb6lhshuzdkOrW1\nRmHo+qSX4vZVIyIPthJJR0Tk1xZ4/idFZFJEvtX690/7uR7D2tL2IAs1f06N5/wC7nYh9uiAx1tu\n28htm3JsyCZIJyxu2zTAnVtzDGUSlBpxAmrbYIrRbJJGGNEIQk5eqGKhbB/y2DKYotSIGBtIkHQs\nZip+PB1z0MMSi0zCwbbiGKIFuLbgWFZLDPki8XEYSMaTIy/R9uxWoY+U6UqTW0bjeYKbch6ubS3Z\naLB3c55fffA27rtpmGakBK0i/7bQSa9/O8O1R988zVa50u8CDwCngadF5HFVPTjv1D9T1Z/v1zoM\n/aPXGs/uDP3ogMfbb/c4Plnm8Pky+3YNdxJQlghb8h4vjZdIORY2caukJULSthgvNNiQTbBtMEUt\nUHaMZLj/5hR7xjJ89EtH8cNYozOdcKg2A0KJjWPKtQmiCMK4hGgonWB0IEmpHnD7WJahTIK/PTLN\nbKVBqRF2BJJdO55tpCgp12akFdO0bauzBV8qDrx3c35OedJq/naGa49FjaaIfJslEo+quvC74SJv\nAI50jf79JPCDwHyjabhO6bXGc6EM/eGJcieZAi2jMZwmn3L5hXfcygcfP0jVDzlfqpN2bbyEQyOI\nmCo3+b47N+KHdDQp/5e/eIFQI84Xm9T9gIGkQ8K2qIYhQtyNE4QRScdGREi6Fp5rszmf5PhMlddm\nagylXWyBSOtUW1v5uFzIIlQhQnn6+DQR0pKgq1zWdtooDF2/LOVp/kDr68+1vn689fXHerz2VuBU\n1+PTwH0LnPePReTNwGHgEVU9tcA5hmuQXms8F8rQbx9JsWMkM+e8ttHYuznPjuE09+0e5m9ePs/5\nUr2jHA/KwbMl/FB59MnD7BnL8PUj0wylYoX449MVivWAfMolnbDYMOBx+kKtJSOXYCDlEmqc4T5z\nocZINkEjCKk2lTMXajQj7bRUKlBqxDWkok2+VW7EY3vzsaf82Neqi7ZELhevXKv6WBMXvfIsGtNU\n1ROqegJ4QFV/VVW/3fr3a8D3rdHr/zWwq+W1Pgl8bKGTRORhETkgIgcmJyfX6KUNl8tKRET2bs7z\nyAN7+Mh77uGRB/Zwx+b8kqpNbVWnu7flGc4k2TqYIpOwqPoRpXrAPdtznfIfz7FAhKznctOGLOmE\nQ6SwfTjD63YO89bbNvKL77iFiLhYPZuwGS/Uma022Tro4VgWU6VmR/BjfjG8Egt/NAKlVGvSCCIO\nny8ThtGCqkK9xCvXQoDFxEWvDr0kgkREvqvrwZt6/LkzzJ0ltK11rIOqTnfNUP8D4HULXUhVH1PV\nfaq6b3R0bbopDJdPryo5C6koLWc02s8nHJvv2B5fb6LUZCyb5P6bh9k4kOqU/yRs6YzhSCdsRrMJ\ngjAuXWqvqdpU7t0+SC7lUmnGI3U35z2mK37HSCoXx2a0ac/+aYsRl5shnmuTdCzOFesLbqd7kW1b\nC4WhtZCHM6ycXhJBPw38kYi0/zdngf+xh597GrhVRHYTG8v3Aj/afYKIbFbV8dbDhwAjBHKDsZSK\n0lJF9d1b+nIj4K23bWRjrsDtm3JYctGotct/3njTMEfOVyjWfdJJhx/YPcxvdyVh/vDrr7FzQ4bd\no1kAvnhogqlSnfPlBq5tMZJxKDd8Io3ntPtdAiEiF2e0t7Ps7cz9/Tdfup3uNV55ufWxJi56deil\nuP0Z4J620VTVnnx/VQ1E5OeBzwM28Eeq+pKIfAg4oKqPA78gIg8Rd63NAD+5ul/DcDXoRVZuqSzx\ncqK8843Ko08eviQOuCnnUawHuLbNG3YPd9pAf7zVI96mHUNsBiHPny7w2lQFP4zItBSWJss+QymH\nSjNOGFWb8biNKIKEI7EgcRD3n6sqxXo8YmOh7XR3vHKyVOfIZIWZcpPhbIJD44U1izka5fWrw7Lb\nbBEZE5E/BD6pqgURuUNEfrqXi6vqZ1V1j6rerKq/3Tr2Gy2Diar+uqreqar3qOrbVNUoKV1H9LI9\nXAux3DYLbenb5T/LbXMfvGuME9MVnjo2w1SpjudasQEkbpG0RBjKeLxlzwZsK96W2wKZhB2P7dXY\neP739s49Oq76PvCfr2ZGM6O3LMmyLPzEGGxcQsE83AAhlBA23UOybVLyakh6dtmeTZpsu6SPTQ+Q\nZLfZZNuQkyZNltMCIaRA4SQNIRxYA2EJYAzGvPyIjd82FrJsmZFGmhnN47t/3DvjK3lGuiNpNJL8\n/Y1jtgIAABxfSURBVJwzZ+69cx/f30j66vv7fV/dLZFCYPuf/u7ZRRXg6s56Xtx7gp9uPcIT23s5\nFksQqIGupvC0rjla5fXq4Gd6fg9wN/AVd3838CDwzxWSyZgj+JkeTqfVVcwLf8nyllHxkqW8x2u6\nmlncHKF/aITjQ07L4cUtUWpEiIQCrOyo4/XDA9RHarl2TScC7O2Lc+hkAlVYUl/LsgVR6sKhcZ+z\nsyfGkzv7OHdRAy8d6CedzTGYgrWLm1je3kAskea+TQdpb4xM2eNt5eGqgx+l2a6q/yoifw2FaXd2\noouM+Y+f6WE+RrM/nmJ3bxyEUVZXuc6PsUVAyqk6n8oqV63uILw/4NbIDLhT7TThYJAPrO0syNsc\nDXHJijbO63Km+37l9Frfu4/FWdLqxpcOpTkHSGUyPL+nn2vOWzgtrZitPNzM40dpDolIG6d6BF0O\nWEyD4ausXN4auu2RHaRzOdobwqxaWE97g6Mw8lP5cmIN87GJG3f0EgoI5y9uokZCo9ZLi12fV/Kr\nOup5Ye8JemIJRtzqQ4f7h7nRrQc6lUwdr/XdFAkV8uPjbnjVjqODtNZZJtBcxo/S/HPgEeBsEXke\n6MDpE2Sc4fidHnqD1b2e78ZIkB09MQ71D/u2Fr3WZU5zoDVs2nuC+nCQrCqN4SDNdaHTroNTSj6b\nzYEq2RyFnj851y0+GY+0N8B8V88Abxx+l0BACIgwlMoQDgVoigSJJdKcHE7zO2cvKOv+xuzCj/d8\nq4i8DzgXJ1xtl6qWLpNtnFH4nR6WmsrHEhm6W+p8W15eS7A5Wsu7wyOcHE4zPJJlWVsdA8kMA4nM\naeulecUWT6XZfnSAGpxukWOt3ryc6Wy2EMJUG6hh3eKmouPyKvFQAPqHRoglnWDzUCjAiFs6qanV\niSu9YlUbtcHAqHuYx3tu4cd7HgG+CHwd+CrwefeYYfimlKe3KRIs6V0vFhTv9cY7fcZHqBGnz1Aq\n4yio1Z0Nozz43syZ8xY1EQ0FaAiHOLvDUZjeZ16/rpPD/cNs2ttPYiRDbY0QT2Y4GksW9Xp7lfi+\nvmHaG8N0NUcYGsmSyUF7Q5jzFjVy/uLmgjV5uH/YPN5zGD+ZPfcC5wP/AHzP3f7xuFcYxhhKZcCc\nv7h4OmVtQIqmCNYGpHB+e0OEpmiQoNuXPBwKcNHSFpa114+a7o4NjWpvCIPAnr6hUc/sbomypquZ\nzqYwjZEgmRyEa4NctnIBy9rqi2baeJW441ByermPuAo8k805BYrdceR7JqUzWes1Pkfxs6a5TlXX\nevZ/JSJWqcgom1JT+WLOpLpQTVGHTDqT5XD/MK+6XS+HUhlqAwGuWt3uFNMADhyP0zOQ4paHXqe7\nJcqOHieTKM+qhfVsOXCS/vhIoemb14E14nrZveuvOdWi647eZYemSIiTQyMcG0wRDtbQEA6wt2+I\nbE5JZ7MFZ9WytvpRBYuNuYUfS3Or6zEHQEQuA7ZUTiRjvuOddj++rZdr13TQHA3xm3cG2NEzwGAy\nzauHY6Qyoy3QxkiQY/FUwWkjOLUx09ks8WSGnCoHjsfZeuhdFrnT5FgizeETCQ6dOGVVtjdEOG9R\nIwsaaotae/liIV5KrTt6lx1WdtRxbNCRr72hllQmRyaXY2FjLXuOnXq+OX7mNn4szYuBF0Qk38J3\nKbArX2/TR11NwyhQLLbyyZ19XLumg0P9w3S31NEYCdI3mGLzvpNsOFsK6455x9Haria6miPs6Rsi\nnszQUlfL3uNDJDLZgpOndzBFYzRIe0OE1Z0N7OqN01ofLliz8VSWVR31o5qc7eyJcd+mg2zad4Lj\n8RE6m8KsX95KOBgs2qEz71zqiSXYvP8EuZxT9qO9PkQOiIQCLF1QRw3O1D2POX7mNn6U5vUVl8I4\nYygVB/mjTYdY29VU2D9/cROb9/Wz7e0BrlodLkyhmyJBkukMrx2OFabAqTT0xlMsbAxTFwrSWud0\nqHzl4LtcvMxZ4xxOZ2mOhgo92XNuTc2eWJJXD53k568eocFtx9sQdlpm9MVHePo3fVx73sLT1h3z\nyj+Xy3FyaISmSAgUhgJp+oczLBBBIyHOaomyuzdOQyRYdCnAmHv4CTk6KCIXAVfgBLg/r6pbKy6Z\nMS8pFQfZO5DkshWn4hc7GiNcsqKV14/E6IklCzGgj2/r5ZldxwgHnerrAIiQyyn9Q056Zj7bB2DP\nsSHWdAU4f3FzYQ3xjo27iacy7OqNEw7W0FZfy96+IXoGkixbUE+0Nki0NkhdrfPn0d4YGZWF5A2s\nz+aUSChAJBSgfyjFQDKL5pR4MkNjOMPuwRSt9SHO7WwcNY5Sjp9SRYWt2PDsYUKlKSK3Ah8Dfuoe\nultEHlLV/1FRyYx5Sal4zc6mCINuxfU8kVCQ69YuOs1h8rNX36YlGkTdMKNUJkddbYCRTI7FzRFe\nOhAnp0o05BTiWOwqqjxvv5ugJ5YcpXhFIJdT4qkMrfW1gFP+bTCZYftRZw12+9EYR04mOLezoRBY\nf+jkMEtaohAKMJTKUiOwuC1K70CKkZzSEAlybmfjqDJ1pSiVFnrtmg6e3NnnOwHAqCx+HEGfAi5R\n1dtU9TbgcuCPKiuWMV8pFa9504alvir2rOlq5spVbYgIg6kMkVCAi5e10BytJZPLsf/EMG31tURC\nNcRTWRLpHNeu6Tit1cSJoRG3fYZDQIRQsIZE+lRZhVQmRzanHDmZIJZIM5BIk8pkeW7PCd6JJXln\nIEVQhONxp+pRIp0lEqohWFPD2R0NXLd2EWsXN/Lq4dioWNNSlKoa9aNNh6zY8CzCj9I8CnjnU2HG\nVGA3DL+Uitf8vQu6fVcy//SGZazsaOCyFW1cumIBoUCA9oYwaTf7pqUuxMJGp93vFava2N07NOr6\n69d1Ego4/dJVlWQ6SzRUQ2M4iCqcHEqx/3icvX1x3hlIsqgpTHM0xPF4itiw49AJubGYI5ks8VSG\n3liCZDrLYCLDkZMJ2upDHI8n2bzvJKGA+GpHUaqMXu9ActrK6xlTx48jKAZsF5GNOGuaHwBeEpHv\nAqjqFysonzEPKRWv6Tcls2jO+wdX8+2Nu4kNpxlMZWiKhFjX3cSC+nBBuXjXBVe217Hj6ACZrLKg\nPsQFZ7UwmMqQzmZ588gANSIsW+CkZfbEknS3JkllctSIUBusYSSbo7s1yjuxJKlMhsFUlsZIgGxO\naIoG2ds3hB4bQsAtKCK+C4r4WbowD3z18KM0f+a+8jxTGVEMwz/FFOzarubTlE4skaa7JXraeuFg\nMkgqo3Q2hRnJasG58vi2XlYtPOXF37TvBAOJNHuODREKCMk0jGRyhIJCoMbJLkpmAlx9bmehZuib\nbztrn8MjGVZ1NOCJkR/XQixVNeqmDUt5cmffacfNA18d/HjPi3aINIzZxnil6oqGOrl91r2Opn9+\n7sAo7/6qjnq2HnyX4/EUCxsjoEkGU1mioQDhUIBlC+rY0zdUmD6LOP3Sl7RGORpLklUKoU/tDZFx\nLcTxqkat7GiwYsOzBD/e83OAbwBr8axtqurKCsplGGUzntIZqwyhuNU3dorc0egEx78zmKI5GmQg\nkeGCs5zYz7xS/u0lzYXp855jQwUH06KmSKEt8Fu9cUKBwIQW4lSXLozK42d6fjdwG3AH8H7gc/hz\nIBnGjFNKuYQDwrO7+xjJ5miKhFi1sJ5QIHCa1VfMWg0EavjqDWtHxUtuPxpjIJmhORokGgpzuH8Y\nFtQRS4wQDtSQyioXLW0BYM+xOL2DKTZEQ2YhzgP8KM2oqj4lIqKqB4HbReQV4NYKy2YYZTFeYPjR\nWJJ4MkNDOEBiJMOmvf2saK/nxg+OjgGdqLBy/v1Q/zBntdYVFGu+clGN1IAIFy1tLhQQqQ0G2GAF\nOuYNfpRmSkRqgLfclrxvAw2VFcswymO8fkGPb+tlWVv9qHz1xkiQzqbwpKbCxdZH85WL/v4Plzkl\n7IIBegcS7OwZ5ORwmitXtU1r+16jeviZZn8JqMMpRHwxTmD7TZUUyjDKZbx2wvn4x47GCBtWtvGB\ntZ1ctbpjVLGOchivLXHeUh3JZHlhbz8A7121gFAwMK3te43q4cd7/rK7GcdZzzSMWcd4vX38dM30\nMlGe90T3W9PlTM2vOW/hqHPAGqjNB0pamiLyyHivmRTSMCZivBqYpVI3i7WY8LbGKJXF4+d+41mj\nxtxmPEtzA3AYuB/YjNNUzTBmJePFaPrtmgmlS9d5LUQ/9yvXujXmDqJafF1HRAI4KZOfAC4Afgnc\nr6rbZ06801m/fr1u2WKF443T8VM+baJzbnnodbqaI6e1uuiJJfm7j72nLFnyjqnGSJBDJ4bY1Rvn\nrNYo5y9uttJusxAReUVV1094XimlOeZmYRzl+b+Br6rq96Yu4uQwpWlMlrGKLG+NeguD3LFxd9FU\nzMn09Mkr6B09MQ6fSLC6s2FUULyVdptd+FWa43rPRSQsIr8P3Ad8Hvguo/PQDWPOMJ6HPU85658T\nsabLKXy8tquZy89uY0VHg5V2mweUXNMUkXuBdcBjONblthmTyjDKwG9V8/E87HnKWf/0i5/nGnOH\n8RxBnwaGcOI0vyin1ngEp6FaU6kLDWOmGC+ofayi8+ucme48b3MKzS9KKk1VtfxyY9bjx9udZzwP\neyWZ6LnW/2duYYrRmNOUGw8ZDdWwef8Jntp5jHQmOyPOmFLV6vN58RPFhRqzCz+554Yxa/E79fVO\n469d01mw9maKUlP+cixlY3ZglqYxp/Hr7fbjOa8Gljk096io0hSR60Vkl4jsEZG/Gue8PxARFZEJ\nY6QMw8t4U18vs1U5jZf+acxOKjY9dzOKvo+TVXQEeFlEHlHVHWPOa8Tx0G+ulCzG/MaPt3u2erCr\n5ZwyJk8lLc1LgT2quk9VR4AHgA8XOe/rwDeBZAVlMc5w/Ezjd/bEuGPjbl89yqcLv5ayMXuopCOo\nG6fgR54jwGXeE0TkImCJqv5SRL5c6kYicjNwM8DSpUsrIKox35koaL2ceM9KyGZKcu5QNe+5Ww3+\n28BnJzpXVe8E7gQn97yykhnzlfGUk3mxDb9UUmm+DSzx7J/lHsvTiJOm+YybbbQIeEREblBVq8hh\nzCjFUh2T6Qyb95+YlqBzC2CfP1RyTfNl4BwRWSEitcDHgULxYlWNqWq7qi5X1eXAi4ApTKMqjPVi\n9w0meXn/SWoDNVMOOrcA9vlFxZSmqmaALwBPADuBf1XV7SLyNRG5oVLPNYzJMNZRtP3oAAqs626a\nclznbI0RNSZHRdc0VfUxnCpJ3mNFW/+q6tWVlMUwxmOsoyidVS5b2Up7w6kp+2TjOq3K0fzC0igN\nw8XrKMoXI/Yy2bjO6YgRtTXR2YOlURqGh3ys5o6eGC/uPcH+vviUixFPtbCxrYnOLkxpGoaLVzm1\nN9SSU+WZt/p49PWjU6qINNUAdlsTnV3Y9NwwXPLKKZ3N8uqhGNHaACsW1IEIw+nclO49lQB2WxOd\nXZilaRgu+aIee44NEQ7WEAkFCIcCpLNaVcvOinrMLkxpGoZLXjkNJNOEg86fRiqToyESrKplN53N\n3oypY0rTMFzyyqk2UEMynSWZzpLK5FjVUV9Vy86KeswubE3TMFzyyum+TQf59Z4TtNaFuHBJM7XB\nQNXLtVlRj9mDKU3D8LCmq5n/+fsXjIqLbI6GptzG15g/mNI0jCKYZWeUwtY0DcMwysCUpmEYRhmY\n0jQMwygDU5qGYRhlYErTMAyjDExpGoZhlIEpTcMwjDIwpWkYhlEGpjQNwzDKwJSmYRhGGZjSNAzD\nKANTmoZhGGVgStMwDKMMTGkahmGUgSlNwzCMMjClaRiGUQamNA3DMMrAlKZhGEYZWLsLwziD8PY+\n6m6Jcv26TmvrUSZmaRrGGcLOnhh3PrufWCJNV3OEWCLNnc/uZ2dPrNqizSlMaRrGGcLj23ppjoZo\njoaoESlsP76tt9qizSlMaRrGGcLb7yZojIxekWuMBHn73USVJJqbmNI0jDOE7pYog8nMqGODyQzd\nLdEqSTQ3MaVpGGcI16/rJJZIE0ukyakWtq9f11lt0eYUFVWaInK9iOwSkT0i8ldFPv8TEXlTRF4T\nkedEZG0l5TGMM5k1Xc3cfNUKmqMhemJJmqMhbr5qhXnPy6RiIUciEgC+D3wAOAK8LCKPqOoOz2n/\noqo/dM+/Afg2cH2lZDKMM501Xc2mJKdIJS3NS4E9qrpPVUeAB4APe09Q1QHPbj2gFZTHMAxjylQy\nuL0bOOzZPwJcNvYkEfk88OdALXBNsRuJyM3AzQBLly6ddkENwzD8UnVHkKp+X1XPBv4S+JsS59yp\nqutVdX1HR8fMCmgYhuGhkkrzbWCJZ/8s91gpHgA+UkF5DMMwpkwllebLwDkiskJEaoGPA494TxCR\nczy7vwe8VUF5DMMwpkzF1jRVNSMiXwCeAALAXaq6XUS+BmxR1UeAL4jItUAaOAncVCl5DMMwpoOK\nVjlS1ceAx8Ycu9Wz/aVKPt8wDGO6qbojyDAMYy5hStMwDKMMTGkahmGUgSlNwzCMMjClaRiGUQai\nOrfSvUWkDzhYbTk8tAPHqy1EhZnvY5zv44P5P8bpGN8yVZ0w5XDOKc3ZhohsUdX11Zajksz3Mc73\n8cH8H+NMjs+m54ZhGGVgStMwDKMMTGlOnTurLcAMMN/HON/HB/N/jDM2PlvTNAzDKAOzNA3DMMrA\nlKZhGEYZmNIcBx/dNMMi8qD7+WYRWT7m86UiEheRW2ZK5nKYyvhE5AIR2SQi292OopGZlN0vkx2j\niIRE5Efu2HaKyF/PtOx+8DG+q0Rkq4hkROSjYz67SUTecl+ztizjZMcoIhd6fkffEJEbp0UgVbVX\nkRdODdC9wEqc/kWvA2vHnPNfgB+62x8HHhzz+cPAQ8At1R7PdI4Pp6TgG8B73P02IFDtMU3zGD8J\nPOBu1wEHgOXVHtMkxrccuAC4F/io5/gCYJ/73uput1Z7TNM8xtXAOe72YqAHaJmqTGZplmbCbpru\n/o/c7YeB3xURARCRjwD7ge0zJG+5TGV81wFvqOrrAKp6QlWzMyR3OUxljArUi0gQiAIjwACzCz8d\nXw+o6htAbsy1HwQ2qmq/qp4ENjI722dPeoyqultV33K3jwLHgCk3GTOlWZpi3TS7S52jqhkgBrSJ\nSANOo7ivzoCck2XS48P5D64i8oQ7LfqLGZB3MkxljA8DQzjWySHg71S1v9ICl4mf8VXi2plkWuQU\nkUtxLNW9UxWoopXbz2BuB+5Q1bhreM43gsAVwCXAMPCUiLyiqk9VV6xp5VIgizOtawV+LSJPquq+\n6opllIuIdAE/Bm5S1bEWd9mYpVkaP900C+e407hm4AROf/dvicgB4L8C/93tlzSbmMr4jgDPqupx\nVR3GaWlyUcUlLp+pjPGTwOOqmlbVY8DzwGzL3S634+t0XTuTTElOEWkCfgl8RVVfnA6BTGmWZsJu\nmu5+3uv4UeBpdbhSVZer6nLgO8Dfqur3Zkpwn0x6fDjN8n5LROpcRfM+YMcMyV0OUxnjIeAaABGp\nBy4HfjMjUvvHz/hK8QRwnYi0ikgrzjr1ExWScypMeozu+T8D7lXVh6dNomp7x2bzC/gQsBtnHeQr\n7rGvATe42xEc7/ge4CVgZZF73M4s9J5PdXzAp3GcXNuAb1V7LNM9RqDBPb4d5x/Cl6s9lkmO7xKc\nmcEQjgW93XPtH7vj3gN8rtpjme4xur+jaeA1z+vCqcpjaZSGYRhlYNNzwzCMMjClaRiGUQamNA3D\nMMrAlKZhGEYZmNI0DMMoA1OacwAR+Q8i8tqYV05E/l21ZRuLiNxeTlUnEVkuIttKfPZPIrLW3T4g\nIu3u9gueaz85CRnvd6ve/Fm515b5nGfc6jyvuZWSbq7k83zI85iItLjbcfe98P2LyHoR+W41ZZwL\nWBrlHEBVf4YTpAuA+8f3KXwGI7sFKESnIYXMvV9QnTztiqKq/7HE8d9xN5fjZO78i997isgi4BJV\nXVXks0qM61OqukVEFgB7ReQedQpP+JE1oNNYCEVVPzTB51uALdP1vPmKWZpzDBFZDdwK/FFeCYrI\nl0XkZdd6+qp7bLlr5dyLE4C+REQ+4daH3CYi3yxx/wMi8i33vJdEZJV7/B4R+aGIbMZJEV0gIv/m\nPvNFEbnAc5v3uHUM3xKR/+Re3yAiT7kFPt4UEW+lmqCI/MS1xh4WkTr3mmdE5LTUxbyVBPwv4ErX\nkvszEXlWRC70nPeciLxnzOX/F+h2r7nSfcZ3RGQL8CX3e3vaHddTIrLUM/4fuGPdJyJXi8hdrsz3\n+PjRNeAEX2fd+13nfkdbReQhcYq85L//b4rIVuBjrnzfdH8Wu0XkSve8iIjc7X6Xr4rI+93jnxWR\nQvaZiDwqIld77t1eSkB3TI/6GMuZTbWj/e1VVmZECMcSuNFz7DqcplKC80/wUeAqHCssB1zunrcY\nJzWwA2eG8TTwkSLPOMCprIvPAI+62/e49w64+/8A3OZuXwO85m7fjlPzMAq041SoWew+s8k9px0n\nC0VcORV4r/vZXbgZVMAzwHqPXO3udtx9vzovn7t/E/Add3s1sKXI+JYD2zz7zwD/6Nn/BU5hB3Ay\nZv7NM/4HXJk/jFMm7rfc7/wVimSauPfehVN7NAH8Z8/4nwXq3f2/BG71jPMvxtzj793tDwFPutv/\nDbjL3T7P/dlGgM8C3/Nc/yhw9TjfYeH7GPt92qv4yyzNucXXcVLEHvQcu859vQpsxfkDOsf97KCe\nKlJwCfCMqvapMwX9CY5yLcb9nvcNnuMP6anp4hU4lWNQ1adxSuI1uZ/9XFUTqnoc+BVOxSAB/lZE\n3gCexCnv1emef1hVn3e373PvPRkeAv69iIRwFN49Pq/zfp8bODXd//EYWX6hjnZ5E+hV1TfVsfa3\n4yifYnxKVS8AlgK3iMgynDz2tcDzIvIajrJfVkIegJ+67694nnMFzneFqv4GOIjzj8KoMLamOUdw\np1h/wOnVhAT4hqr+nzHnL8eZDk4GLbHt935jc3MVZw22A7hYVdPiVICKjHN+2ajqsIhsxLEE/xC4\n2OelfseVct9znu38/rh/S6ra5065L8OxOjeq6id8ypN/Vnai5wAZRi+7zco2JHMZszTnAOJUobkb\n+IyqDo75+Angjz1rYt0isrDIbV4C3ici7SISAD4B/L8Sj7zR876pxDm/xlGEeYV+XFXzlc0/7K65\nteFM+V7GKbl2zFWY72e0ZbVURPIW7SeB50o8cyyDQOOYY/8EfBd4WZ2K5OXyAk4lHXDG9+tJ3OM0\n3HXa38YpOvEi8F7PenG9u1ZdDt7vfzWOJbsLZwp+oYjUiMgSHCvfmEbM0pwb/AmwEPiBjC5q/A1V\nfVBE1gCb3M/iONVdRnldVbVHnKZUv8KxTn+pqj8v8bxWdxqdwlGuxbgduMs9b5hT5dXAWcP7Fc7a\n3ddV9aiI/AT4hYi8ibMu6y2ztgv4vIjchVNR6Aclv4nRvAFkReR14B5VvUNVXxGRAZx/MpPhT4G7\nReTLQB/wuUneJ89PRCQBhF0ZXwHHYQPcLyJh97y/wank45d/xPl9eBPHuvysqqZE5HmcNis7gJ04\nSzbGNGJVjoxRuNPm9e565JxDRBbjOE/O02kKsTIMLzY9N+YNIvIZYDOO998UplERzNI0DMMoA7M0\nDcMwysCUpmEYRhmY0jQMwygDU5qGYRhlYErTMAyjDP4/piWzhRTdz+8AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f24ec4ec110>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(5, 5))\n",
    "plt.scatter(np.mean(process_proba(results[\"dropout\"]), axis=1),\\\n",
    "        qc[:, 4], alpha=0.5)\n",
    "#plt.ylim((0.045, 0.27))\n",
    "plt.xlabel(\"Zero probability from Bernouilli\")\n",
    "plt.ylabel(\"Mapped reads QC in each cell\")\n",
    "plt.savefig(\"unnormalized_zero_B_mapped_reads.pdf\", dpi=300)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}