visualize_predictions_wfpt.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Plotting Model Predictions "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Import necessary packages\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import tensorflow as tf\n",
    "import scipy as scp\n",
    "import matplotlib.pyplot as plt\n",
    "import os\n",
    "import dnnregressor_predictor as dnn_pred\n",
    "import dnnregressor_model_and_input_fn as dnn_model_input\n",
    "import make_data_wfpt as mdw\n",
    "import make_data_sin as mds\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "\n",
    "\n",
    "hyper_params = pd.read_csv('/Users/admin/OneDrive/git_repos/nn_likelihoods/tensorflow_models/dnnregressor_mse_from_simulation_mix_08_16_18_19_21_26/dnn_training_results_mse_from_simulation_mix_08_16_18_19_21_26.csv',\n",
    "                          converters = {'hidden_units':eval,\n",
    "                                        'activations':eval})\n",
    "\n",
    "model_params = hyper_params.to_dict(orient = 'list')\n",
    "for key in model_params.keys():\n",
    "    model_params[key] = model_params[key][0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "datapoint 0 generated\n",
      "datapoint 1000 generated\n",
      "datapoint 2000 generated\n",
      "datapoint 3000 generated\n",
      "datapoint 4000 generated\n",
      "datapoint 5000 generated\n",
      "datapoint 6000 generated\n",
      "datapoint 7000 generated\n",
      "datapoint 8000 generated\n",
      "datapoint 9000 generated\n",
      "label 0 generated\n",
      "label 1000 generated\n",
      "label 2000 generated\n",
      "label 3000 generated\n",
      "label 4000 generated\n",
      "label 5000 generated\n",
      "label 6000 generated\n",
      "label 7000 generated\n",
      "label 8000 generated\n",
      "label 9000 generated\n"
     ]
    }
   ],
   "source": [
    "# Generate a dataset wfpt\n",
    "dataset,_, __ = mdw.make_data_rt_choice(v_range = [1,1], # uniform [0.6, 0.6]\n",
    "                                        a_range = [2, 2], # unifor\n",
    "                                        w_range = [0.5, 0.5],  # uniform\n",
    "                                        rt_params = [1,2], # gamma(1,2)\n",
    "                                        n_samples = 10000,\n",
    "                                        mixture_p = 0.1,          \n",
    "                                        write_to_file = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Make dataset feedable to tensorflow (wfpt)\n",
    "features, labels, __, ___ = mdw.train_test_split_rt_choice(data = dataset,\n",
    "                                                           p_train = 1,\n",
    "                                                           write_to_file = False,\n",
    "                                                           from_file = False)\n",
    "\n",
    "# Make dataset feedable to tensorflow (sin)\n",
    "# features_sin, labels_sin, __, __ = mds.train_test_split(features_sin, \n",
    "#                                                        labels_sin, \n",
    "#                                                        p = 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Make feature columns (wfpt)\n",
    "feature_columns = dnn_model_input.make_feature_columns_numeric(features = features)\n",
    "model_params['feature_columns'] = feature_columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Reinstantiate model\n",
    "my_predictor = dnn_pred.get_dnnreg_predictor(model_directory = '/Users/admin/OneDrive/git_repos/nn_likelihoods/tensorflow_models/dnnregressor_mse_from_simulation_mix_08_16_18_19_21_26/',\n",
    "                                             params = model_params)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Get Predictions (wfpt)\n",
    "model_checkpoint= 'model.ckpt-420000'\n",
    "my_predictions = dnn_pred.get_predictions(regressor = my_predictor,\n",
    "                                         features = features,\n",
    "                                         labels = labels,\n",
    "                                         checkpoint = \"/Users/admin/OneDrive/git_repos/nn_likelihoods/tensorflow_models/dnnregressor_mse_from_simulation_mix_08_16_18_19_21_26/\" + model_checkpoint)\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "# concatenate predictions with dataset\n",
    "new_col = pd.DataFrame(my_predictions, columns = ['pred_likelihood'])\n",
    "data_pred = pd.concat([dataset, new_col], axis = 1)\n",
    "\n",
    "# Prepare for plotting\n",
    "data_pred.sort_values('rt', axis=0, ascending=True, inplace=True, kind='quicksort', na_position='last')\n",
    "data_pred_long = data_pred.melt(id_vars = ['rt', 'choice'], value_vars = ['nf_likelihood', \n",
    "                                              'pred_likelihood'], \n",
    "                                               var_name = 'group', \n",
    "                                               value_name = 'likelihood')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(-1, 1)]"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZQAAAEKCAYAAAA1qaOTAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzsvXd4XOWZ8P17pqt3S7K6e5WbbDA2YCAYCMUkoSahJiGQJcmX3X2vZfd7N4V997vybshuwoYUShaSEDqh2QQwNQaDLWO5d1u2ZdlW723K8/1xz2hG0qhYbVSe33XNNTrPec4590ijc5/7uZvSWmMwGAwGw1CxRFoAg8FgMEwMjEIxGAwGw7BgFIrBYDAYhgWjUAwGg8EwLBiFYjAYDIZhwSgUg8FgMAwLEVUoSqnfK6UqlFK7e9mvlFIPK6UOK6V2KqWWhuy7Qyl1yP+6Y/SkNhgMBkM4Im2hPAlc2cf+q4CZ/tc9wG8AlFLJwI+A84AVwI+UUkkjKqnBYDAY+iSiCkVr/RFQ08eUdcAftPApkKiUygSuAN7RWtdorWuBd+hbMRkMBoNhhLFFWoB+yAJOhmyX+cd6G++BUuoexLohJiZm2Zw5c0ZG0rFCWz24WyEqEWyuSEtj6IuOZmhvBHsUNFWAzSmv2HTZ394EHU3UWpJAKZKiHV0Or2hsJ8puJc411v+NDeOdbdu2VWmt0/qbN9a/iSrMmO5jvOeg1o8CjwIUFRXp4uLi4ZNuLOL1wNEPIG8lOGIiLY2hL45/AkfeB1c8nN0LU+ZBRxNc9I9gtcPpHbB/A3/wXUlGRiZr52d0Ofzlz8tobvdw28r8yMhvmDQopY4PZF6kfSj9UQbkhGxnA+V9jBusNpj5BVEmHc3QWhdpiQy9oTU0nILy7WJVxqTKeEezvLsS8WmNp7mWOJe9x+HZSdFUNXXQ0uEZRaENht4Z6wrlNeB2f7TX+UC91vo08BawVimV5HfGr/WPGQJoDTufg10vgLst0tIYwqLl74SC5gqw2KHobnDEyu6oJDo8Plye+rDLWtlJUQCcqm0dRZkNht6JdNjwM8BmYLZSqkwp9Q2l1L1KqXv9UzYAR4HDwGPAdwC01jXAvwFb/a8H/WOGAErB9EuhpQb2vgI+b6QlMnSne6XvtlqISxcrE8AZR5u2EuWpIz6MhZIe78JuVZQZhWIYI0TUh6K1vrWf/Rr4u172/R74/UjINWFIyofZV8L+DXDoHZh1hSgawxihm0LxuuHwRkidBYm5oBRtjhSiO6qJcVp7HG21KKYmRlFW2zJK8hoMfTPWl7wMQyVzEeSeL+v0ZRM8IGG8oTVdlIqywMmt0BB0BzY7Uohx1xDl6KlQIOhHae0wFqgh8hiFMhmYtgbSZoPbPMmOLbQokQAWq0R3dTR1DrVYYrD5OnApX9gzZAX8KHVm2csQeYxCmQwoBfOuh2kXy7bXRAWNCbSGpAIovInOSHibC6oOdU5p9dmxWRUWb/jAivQ4JzaLMgrFMCYwCmWyYPH/qc/uha2PSUKdIbJon+gRdwvEZUBcJkQlSai3P3S4RduxWy0SVhwGm9VCRoLLRHoZxgRGoUw2olPkZrXrRfB0RFqaSY6GupOw7w3wdkB8JkxdIrvaGgBo9jmwWRR4elcYWUlRVDS20e4xfhRDZDEKZbIRly7LX01nYd9rPUNXDaOHBjz+pay2erEaU2bABfd3ll9p9tmxWS195hJlJ0ajNZyuM/lGhshiFMpkJHUGTL9M1uqPvh9paSYxIVWEfB6oPgw2BzjjOpcom7w27FbVZ0BFRoILizJ+FEPkMQplspJdBFlLobzE+FMiQXO1hHF7uy07ag37XofTO9Ba0+izY7dYgpZMGBw2C+nxTuNHMUQco1AmK0rBjMth2Z3yRGwYXaoOynv3JUeloLYU6sto9/jwYcNqd/TqlA+QlRTFmYY23N7w4cUGw2hgFMpkxmKB6GQJIz7wV3lqNowOlkCiYhgfliMW2ptod4tysDii+lcoiVF4fZoz9caPYogcRqEYwN0MVQdg1/PQYZIfRwXlVyhJ+TDzCrp0ZHDGQUdTZ9SWxRHd55IXwNTEKJQyCY6GyGIUigFcCbDgBmnotPslk/g4GqiQREZXvLwCVYb9rQfaPWKhWJ3R/VooLruV1FjjRzFEFqNQDEJCFsy5GurL4OCbJpx4pAmUXGk8Iy0GbC5InSljjhhwt9DuFsVuHcCSF4gf5XR9K16f+dsZIoNRKIYg6fOg4CI4sxtOl0RamomN9ichtlTJu9cd3DdlHiy4gQ6/hWJzxvSZ2BggOzEKt1dT0Wj8KIbIYBSKoSt5F8CstZC+INKSTGwC/WkClmBrjVSEBuncmDoDao7i9DRgd0ZLYmM/VuPURNNwyxBZjEIxdEUpyFomVW9baqDqcKQlmpj01fCsowU2/5r4Q38hv/YTbK5oqfvVVNlnxnyM00ZyjMM45g0RwygUQ3i0hoNvyfr+/g3gaY+0RBMLXyDwIYzV4fNAWz1enybWW4/VES1/j8+fgsbynvNDyEqM4lRdKz7jRzFEAKNQDOFRChbeKM25zuyE4t9LIUPD8BBQKF2WsfyRX654WP4NqlKWEe1rlF7z3nZorYWW2j5Pm5UURbvbR1WzeQAwjD6R7il/pVLqgFLqsFLqgTD7/0spVeJ/HVRK1YXs84bse210JZ8kWG0w/RJY/DXZLnkajn8SWZkmCoElr/ipwQrDocROoS4qn6qU5bL86G4Dn1sy7CsP9nrazoZbxo9iiAAR6ymvlLICjwCXA2XAVqXUa1rrvYE5WusfhMz/LhD6n9eqtV48WvJOahJzoOhuOPwuuBJlTGvTn34oBKK8YtKkwnAYGpxptE1JgSh/lJeyQnMFHN8EabPCHhPvshMfZedUXStLcpNGSnqDISyRtFBWAIe11ke11h3As8C6PubfCjwzKpIZemJzwpwvSmgxwIENcHij8a0MlsCSV2sdlG6Szo3dLJV2t48ET7VEgLU3ARryVkHjWXn1QlZiFKdqW9Eml8gwykRSoWQBoYvyZf6xHiil8oAC4L2QYZdSqlgp9alS6vqRE9PQA63lafnkVtjyGFQeMImQ50pAodQcgcbT0FbXY0qH10dW1d/gxKdSvj46BdLnSx2wMzt7PXV2UhQtHV5qW9y9zjEYRoJIKpRw6yW93ZVuAV7UWofGWuZqrYuArwK/UEpND3sRpe7xK57iysrKoUlsEJSC2VfC0tvA7oLdL0sHyNaeN0VDL3SP8mqtlX4oIXR4fHiip0BzJeSvhrS5YI+C1FlwdnevJXKy/Pko5SZ82DDKRFKhlAE5IdvZQG8xkbfQbblLa13ufz8KfEBX/0rovEe11kVa66K0tLShymwIJSEblt0NMy6DuuPS38MwMHz+MvOhj1DtTV2meHwaX3SqLCtarOKUB7FS3G3QcCrsqROj7cQ4rZQZx7xhlImkQtkKzFRKFSilHIjS6BGtpZSaDSQBm0PGkpRSTv/PqcAqYG/3Yw2jgMUCOStgxbfkKRqgYj/UnYisXGOdLhZKeMPc4/Xhi/I/BJ3eCR1+hZOYB8u/KRFiYVBKkZUYbRIcDaNOxKK8tNYepdT9wFuAFfi91nqPUupBoFhrHVAutwLP6q4exrnA75RSPkQp/jQ0OswQAVwJ8q41HP8YmiogsxCmXQKO6MjKNhYJm4cSRGuN26vRMVOgDonqCnTWtDkgtm9rOyspioNnG6lvdZMQZR9GwQ2G3omYQgHQWm8ANnQb+2G37R+HOe4TYOGICmcYHErBkttEqZzcIn3rp18CGYUmzDiUgEKJSoS4TKk6HILHn+luczjhov8FZVvh6Af+5S+bhHAnT5OaX2EI9aMYhWIYLUymvGH4sTlEiRTdLZFJ+zfAoXciLdXYQnshKQ+u/42EAncj0MrXZrVIgmnAAmyrl9L35duhoazX06fGOnDaLSbB0TCqRNRCMUxwYtNgydclxDU2XcZa68AeLUpnMuPzyu8BLc71hJwuCY5ur99CsQTKsQQUSgPETpHfn6ej19OLHyXK+FEMo4pRKIaRRSnIXCQ/aw17XxXn8ozLe832nhT4PLJ0tfkRqNgrUVvO2M7dHr+FYrf6FxFCLRSQhlz9tAXOToriaGUzjW1u4lxm2csw8pglL8PooRRMvxSsDmk1PJlzVwIKJeCUb6vr4kcJ+FDsVr+F4oiR+YEESKuj3yoFeSkxAByvbhle2Q2GXjAKxTBwmqv8FsYQblCBumDTL4XaUtj6GJRtGzYRxw0+r+SWhIYMNwUVSqBbY6eFopRUIW5vkG2bSyoQ90FKjIM4l43S6ubhlNxg6BWz5GUYOI2npcxK7XGY/cVeI4z6xWKF3PNgyhypBxaI/ppMBScDCkX7QgaDn70zyssa8vtwxgeXvHJW9HsJpRR5KTEcPNuI16exWibJ79YQMYyFYhg4GQth6R2SV7LrBTjwZp+O4X5xJcCCrwSLIh5+F/a9AR0T/Ilaa1nyUtZe81ACPhSbJeRf1JUgTnmA1Jny6of8lGg6PD5O1xvnvGHkMQrFcG7EpcPSO6Xx1ukdUPzE0BtvKSU3VqtdHNRbHoXykolbcDLQC0VZpIqztWfEWyDKyx5qobgSRNl63VJt+Mzufi+VkxyNRSnjRzGMCkahGM6dcI23jrzXa7HCAaEUTLtY/CsxaWL9bP+jZNxPNAI1Tm1OuPDvYe51PaZ4fN18KCA+FBArpXI/7H+jX6XrslvJTHQZP4phVDAKxTB4Ag72zEVw4jP4/Mk++3QMiJhUUVRzroaWGsm2n2gEsuSVVd49bRCbIdafn2BiYzcLBaC9XpzyWoM3ZMlRa6jYJyHIIeSnxFDR0E5z+xAUvsEwAIxCMQwNmxNmXyX95zta4POn4PjmYDXdwaCU1AE779sSDQZDiywbawQUis8DWx8XS8PTKiVY/HQueYX6UJwBC6U+mBgamovSUA57XoGj73e5XH6K1FI7VmWsFMPIYhSKYXhInSEVcFNmSM2pkj+JhTEU7FESANBcBZ/9VsqNTAQCS4MWKzT5e/S01QeLPwIer0RlWSzdoryURZa8bC7/xBALpd5fisXd1QGfFuckzmXjSGXX8vgGw3BjFIph+HBEw/wvwdxrRQkU/16UwFCd61FJ0nvl4FtSbHK8E+hrovxR+4FfT/WRzilun6+r/wSkVYAzTpSPNYyFUlvqn9c1G0ApxfQpsZyobunMbzEYRgKjUAzDi1KQsQCWfwPis+DAXyXEOOTp+5yxWGHe9RCXAXtfgfrwjaXGDV6/QrH0/u/n9vi6RngFcPlzUVyJ0mgrYKmAhHXPuRrm9XTyz0iLxePTnKgxy16GkcMoFMPI4EqARbfAzLXSzXHr4+IwHiw2h/hpHHGioJqrh0/W0abTQrH2OsXj08HCkKG4EkShxKSI4gjti5I+T3xPYchKjMJlt3K4wigUw8hhFIph5FAKspdJm+CoJHEY73hWlnYGswzmiIHCm2RJp2mI0WSRJNSH0gtur09K13cnKkmsPU+H+FIC/pLqI1B1WJYEP3u0Rzthi8/NQucZjlXU4fVN0PweQ8QxCsUw8sSkwJLbJWKruRJ2Pi8WS3lJcPlnoEQnS/RX+jzZHko0WaQIWChRSRJ2PfeaHlM8Xh1+ySsqSd6bq6RScXmJbB//BI5vAhS0VAeLSAaoO8686o04ms+YHimGEcMoFMPoYLFI/a7zvyM3UGWR5MXNj8Cxj3o8UfeJ1V+KvXQT7HwumHk+XggoUbtLck+sDlnKCi1fH84pD0GF0t4oFo6nTayVhnLpNR+VKPsDNb8CuBJIiLaT7D7N4coh+LMMhj4wCsUwulis4jwuuhsWf1Wit0o/hk9/DfvXB8NoB4IzXiKbBpAxPqYI5KG426ST5eF3wWLv0mCrw6vDL3k54/zHNoki8nZA/UkpMpmUH0x+7N4WYMczWJViur2ag2ebzLKXYUSIqEJRSl2plDqglDqslHogzP47lVKVSqkS/+ubIfvuUEod8r/uGF3JDUNGKWmBu/AGWcLKXCR1vLY+PnA/S2ahlGs5u7dHMt+YJmCheNuhrFisCW/XIpserw97OKe8PUZ+dx3NwSZbtcfEr5SQLdabI6brkpe7TRJDlYUsSy3tbW2crJlAiaKGMUPEytcrpazAI8DlQBmwVSn1mtZ6b7epz2mt7+92bDLwI6AIieLf5j+2dhRENww30ckw6woouEh8AqeKxc8SkwrZRZC+ILjM1Z3clbJcduIziQDLWT66sg8Gn1uUggp5nmtvgPIdnZWXPb1ZKBaLtA7uaJYqBZ52qXcWUCYgy16hFkqr/99i6mKSyraR2naW/WdSyE+NGaEPaJisRLIfygrgsNb6KIBS6llgHdBdoYTjCuAdrXWN/9h3gCuBZ0ZIVsNoYI+CvJXS66NiH5RtkTyWox9C1lKYurSLnwGQG/OML0BHoyiiqYt7Vz5jBa+7R/IhENoOBbfP17WOVyiOGFGiUYlyUO75/v70fuZcI8omQKu/YkHGQixeNznNKeyqbMLt7cVPYzAMkkgqlCwgtO55GXBemHlfUUpdBBwEfqC1PtnLsVnhLqKUuge4ByA3N3cYxDaMOBarJEemz4e6E1C2VaKYTnwKU+aJwomdEjLfAnPXyfLPWFcmIArFau/WXKsrHq/G0dvN3hkHHU1QdFf4/dHJXbcDFkpMGsy9hoKaFrZtK+NYVTOz0uMG8QEMhvBE8vEk3ONX90Xz14F8rXUhsBF46hyOlUGtH9VaF2mti9LS0sJNMYxVQv0sK+6BzMVQuQ+2PgElz3T1s1htYr10tMhy2Vgue+9zixM+oFC6dan0+jTe3hIbQSyUjj6i4hrPSKvmzkgvJVUGrHbwdJDlOUmCzcO+0w1D/ywGQwiRVChlQE7IdjZQHjpBa12ttQ40zn4MWDbQYw0TjOhkmLUWVt4P09ZIrsXO52HLY3Dq86Cj2+cWZbLzuZ6hs2OFgIUSlSTLddMu6bI7WLq+l39PR4wozt5ycDztEqgQKM6ZvypozTRXYtnzMotiaymtaqHJlLQ3DCORVChbgZlKqQKllAO4BXgtdIJSKjNk8zogULvjLWCtUipJKZUErPWPGSY6AT/L+fdJ6RGrXYpGbn5EfC3KKtn03g5ROO4xmMTn84gPxZUgQQSB3BE/gX7yYRMbQYIPtE9K3oejMxfF75gPjZaLywSbk9mOanxaGyvFMKxETKForT3A/Ygi2Ac8r7Xeo5R6UCkVqG73PaXUHqXUDuB7wJ3+Y2uAf0OU0lbgwYCD3jBJsFjFx7LsTljyNYlyOrFZ8llOboGCNeI72PXiuWfjjzQBC6W9ESoPSLh0Yp6ETtNLP/lQHP7orN6SQR2x8vtpqRZL5qOH4PRO2WexQFIesS0nyU50sftUPXo85fAYxjSRdMqjtd4AbOg29sOQn/8Z+Odejv098PsRFdAw9lEKEnPl1VIDp7ZJr/szu+SmfXonpM6SLP2+8PmkE2JLjSiiuExICBvnMXR8brBFSXb77pclMbE1GPHe4VcoDlsfPhTw+1HSu+5rb5IKBHUnIH6qnNfnEcsuQFI+VB5kUb5m/WE3ZbWt5CRHYzAMlYgqFINhWIlOhpmXQ/5qUSplxfJEXl4CaEjI8VfrrRPF0VItIbUtNZK34QvxJzjjZFmtjwKOg8brlvOHRnm1N0iQwdQlePzdGnu1UAKh091bApzcIst+Po+UY6ktFQc9QFRI5FdSAQDTbJU47YnsPlVvFIphWDAKxTDxsEdJbkb2cllSKtsilY7bm7paHcoijvHoZEieJu/RKVLFd9/rkguTsWD45fN5ukZ5BYIWO6S0fKdC6c2H4kwQH0xLVXCs8QwceR+SCyS0eufzkLNS8nOU6uqniUqCrGXYYlOZmxHNrlP1XNzhIdphbgeGoWG+QYaJi8UqVYnT58Gul+DEJ7Lck71clIcrMXyTK63FH3PyM/HTqF5u7IOlex5Kt/O7/dFbvSYdWixSRSBQ98znlTpojmiYe53U+IpOAncztLSIVRZqaSklEXPAIkcHJSfr2FlWz/nTUob1YxomHyZN1jA5mH89ZBWJb0EpUSi9LSkpJcmTTRXBtrrDSWceit8ZrrrKEbBQ+sxij02XnjBaSx2zpgppZmZ3yedKnh5sD5C1rOfxXg9UHSLZ0kxBagw7y+o6gwEMhsFiFIphcmCxwoIvy5P9nr+IQ7wvpswX5/fJz4ZXDq3lZm61iR8ldSakze0yJZiH0odlFDtFQqJPfgYnt4rSSJsd3J86Uz5zznmiHLvjbZcIuLN7WZKbSHO7lwNnTVl7w9AwCsUwebA5JUfFHg27X+o7nNhqk8KUNceGN+te++RlsYu/Y+ENEiwQQkCh2PvoOU+Mv+rDkfclwm3GZV33JxXIslpDWfjjHTHSi6W2lNzkaFJjHWw/UWdCiA1DwigUw+TCGQeFN8PsL/Zf92vqEpkznFZKoEy91S4Kzd0KzWchPksCAwgmNvZroQDYHLDgKz2j0ewuUVRVh3s/R1I+NJxCed0syU2isrGdkzVjMBHUMG4wCsUw+YhJgZTpsvx06nMJsQ2HPUqSDc/ulciv4aBToTjEx7HpF1C5X8rE+Jtnub0+lKL3Wl4B2ZZ/U0rR2F3h56TOlNDoll5yfpPyxaFff5I5GXHEuWx8eqx68J/NMOkxCsUweWmqgENvi0+ltzbC2f6WO6eKh+eagWU2q6NrHkpHU2fOiNursVstqP6iy2LTupap706gA2TVwfD7E3Ik/Lj2GDarhWV5SZyqbTXNtwyDxigUw+QlLl0ae9UclezycP6DqCRImwPl26Xo4lAJtVC6l69vOgtI6ZU+rZOBEpUIiTlQuim8H8hql8KRiXkALMhKIMZp5bNjpoqRYXAYhWKY3ExdIpn1Z3bBsY/Cz8k5T5bFykuGfr1QH0pnHkrXf0N3b90aB8Pc68SK2fVieCss7wJZGkPClJflJXOypoWyWmOlGM4do1AMhvzV4is5/kmwiGIo8ZkSSVW2tfelsYESuuTV2cKnqzXi8fl6rzR8rrjiYdaV4qOpOdpzv88njvuG0wAUZicQ67Tx8eEqE/FlOGeMQjEYlJKbbs4KCeUNR855UjurYl/4/QMlsGxmc0qpfZtTlt5Cp3h173W8BkPyNHHin97Rc59SsP91KaqJWCkrp6dQXtfG4Yo+mngZDGEwCsVgAMkun3GZRFq5Wzuf2DtJmS5JkSc/C+9rGSihS155K+HCv4fkGV2mdHiH0UIBCSnOXg5Vh6RSQChKSbRXbWnn55qXGU9qrINNh6vw+oyVYhg4RqEYDN3Zvx52PgvNIcUXh6scS5clr8BYO8RmSDkVxELps+zKYMhZIctfh9/tqRCT8sX6apGQYYtFsXpmGnUtbnaU1Q2vHIYJjVEoBkN3Zlwmy1Elf+6aGNhZjmXL4M8dsFAsdvHZbP+TRJB1NIqvBvGh9JnUOBisdmmd3HhGAhBCScqX9xBFmZ8STV5KNJuPVJs2wYYBYxSKwdCdqCRYdKv4HXa9AHtfk86HneVYjg6+HIu3Q85jsYijPNBYq6O5s7+Je7h9KAGmzJOmW8c+7JrMGZUkrxCFopTiktlT8Pk0Hx2sHH5ZDBMSo1AMhnDEpkHR3RIBVrEPqg/J+FDLsXjdweWuQG/5QPiwPwrLPdw+lABKifXV3gQnP+26L28lTOlapDIpxsHygmQOnGmktKp5+OUxTDgiqlCUUlcqpQ4opQ4rpR4Is//vlVJ7lVI7lVLvKqXyQvZ5lVIl/tdroyu5YVJgsULBhbDiW5BRKGNVByVvY7DlWLwdPRVKNzxe3/DloXQnIVv6w5z8rLOhFyBh0+nze0wvyksiOcbBe/sr6PCY8vaGvomYQlFKWYFHgKuAecCtSql53aZtB4q01oXAi8B/hOxr1Vov9r+uGxWhDZOT6GR5um9vklItp3eIxXLsw3M/l7cjWJTS5/UXdQxaI1prf+mVEbBQAuStlhL6pz7vOl59RF4h2KwWvjAvnYY2Nx+apS9DP0TSQlkBHNZaH9VadwDPAutCJ2it39daB1J2PwWyR1lGgyGIM1YKMmYvF2Ww5THY/XLXaLD+8LrB6q+/FbBQQupxuQfSXGuoxKRIna/yz0WxBCj9mwQKdCMrMYqivGR2n6rnSKXJTTH0TiQVShZwMmS7zD/WG98A3gzZdimlipVSnyqlru/tIKXUPf55xZWV5gnLMESikmD2VbDmXyTM99iHcNpfkmUg+Sme1qACmXUlzLkGMhcHd/fX/ne4yFkugQZndwfHkvKl8ViYmmUrp6eQFufknb1nTdSXoVciqVDC2fRh/yOVUl8HioCfhQznaq2LgK8Cv1BKTQ93rNb6Ua11kda6KC0tbagyGwxC2kwJw03Klyx6kGzznc9LYmRvdDRL6DFI8cboZEBJ33dHLG6PvxfKcBSH7IvEPOmpUrY1qAiT8iVAoHvyI2C1KK5akIHH62P9znKT8GgISyQVShkQ2qouG+jRl1Up9QXg/wWu01p3Pjpprcv970eBD4AlIymswdCDnPPA3Qa1x2XbYpXQ272vhbdWfD6/QomV7dKPxR9zpgRQkDIdt99CcdhG+F8zkKjZXBWs8RWfLSHNvSRupsQ6uXxeBuV1bSaU2BCWSCqUrcBMpVSBUsoB3AJ0idZSSi0Bfocok4qQ8SSllNP/cyqwCtg7apIbDNCzHMvUJWK11ByFhlM957tbZF5AoZzZJTdvT0dnwmNnP/mRtlBA8lKcscFETasNEnL7rAQwOyOOpXlJlJysY095/cjLaBhXREyhaK09wP3AW8A+4Hmt9R6l1INKqUDU1s+AWOCFbuHBc4FipdQO4H3gp1pro1AMo4tS4qAPLceSuVja8naPoIJgmK7Tr1C8HZIxr72ibE7vxDMaTvkAFitkFYnsDf7Fgaxl8pn68AddOCOV3ORoNu6t4Hj1MOanmOrG456I5qForTdorWdpradrrf/dP/ZDrfVr/p+/oLVO7x4erLX+RGu9UGu9yP/+RCQ/h2ESk76gazkWmwPSF0pb345uPUU6/BFSAR+Kz921phdSGBJGSaEAZC2VigCBXjCpM2DqYlGWvWCxKK5KKugEAAAgAElEQVQuzCQ51sEbO09T0dA2NBm0hkMbYf8bRqmMc/r81iqllvb1Gi0hDYYxS7hyLFlL5Um/e4xJqELR2h9C3DWxMWihjMKSF0jEWe5KqDkW9AXVlsKZ3X0e5rJb+dKSLJw2C3/ZforqpkF2syzfLuVn7C65Zvn2wZ3HMCbo7zHo5/7XI8BnwKPAY/6fHx5Z0QyGcUL3ciwxqVLiJGCJBGgPKJQ4yWPR2m+hBJVHwIdiH2mnfChZS2UZ7thHItPpHXDkvX6thVinjS8vzUYpeHFbGVXnqlRqjsHBt6CsGPJWiU/q8Mbg8pth3NHnt1ZrfYnW+hLgOLDUH367DImoOtzXsQbDpMEeJaVLupdjKd8uSzkBOprFIrDa/E291kJSQZc8lE6FMhLFIXvDapcben2ZWFpJ+SLrABI2k2Mc3LAsB4tSvLitjIrGAS5/dTTDvtchOkUCGZSCuddKwMKev/RcLjSMCwb6rZ2jte6sea213g0s7mO+wTC5yC4CNJwqDo611kmeR32Zf7tW8k3A7xBfJiXrbQE/igrJlB+lJa8AmYskL+bYh5KjAlB7bECHilLJxmZRvFBc1n8hSa1h3xuSQDlvXfDz26Ng/pdEmRz9oOdxp3eIgvaZmmJjlYEqlH1KqceVUmuUUhcrpR5DIrMMBgNIBn3abLFKApnmeaukA+SeV8RyaamCGH9yrbtNqhi3N4rvIj4bMgvxeH0oJYmE3Wlze/n8RC1HKpuGv1CjxSqVlRvPQuNpsRzOoZFYUoyDm5fnkBBl59WScnaV9RFSfPIzsYRmXCbJlaHEZ8KCL8P0S7qO15+C/RtEQR99f+CfyzCqDFSh3AXsAb4P/D9IzsddIyWUwTAuyTlfckoCvdttDlh4o3RkLHkaWmrEvwJirex5RRpetVRBq3RLlPa/FpTqWjASwOPTfHigktdKynn0oyO8t/8sDW3u4ZN/ynyR79jfIDFXMuZ93gEfHueyc2NRNrkpUWzcd5Z39p7tXMLrgsUuFY+n9pKLnDJdrJX2RpHB3Qb7XhPrLn2+RNRVmOfZsUjP2tlh0Fq3KaUeATYioSsHtNbD+E02GCYA8ZlyIy7bKstZFivEpcP8L8PWx+XpP2ChhPaW1z4p11J9BI8lvctyV21zB2/sLOeKBRmkxTr59sXTqG7qYN/pBnafkteqGSksy0seuvwWCxRcJAUvU6aLBaF9gHXAp3DarKxblMXmo9VsOVbDmYY2rl6YSXKMQ5aqLBbIXgZ6aZ+hyYBYJA2npGZaWwMs+RrEZUJbnbRpjk6VvjWGMcOALBSl1BrgEPAr4NfAQaXURSMol8EwPsk5T25+oU/QyQXio4jLkKUkCPaWt9iD89zNeHy+zhwUrTXvH6igsd2D02ZFKUW0w0ZOcjRr52dw56p8ZmfEEeeSc7R7vPiGWmMrdZbIWXlA8mms9v6P6YbFolg1I5Xrl2TR1Obhz58dZ1tpFb6dz0m5GehfmYAELTSegb2vSgOwhGxR0vO/JNFxe14W68UwZhjoktfPgbVa64u11hcBVwD/NXJiGQzjlEA5llDnPEgRSFcCuBJl2+dXKN0SG9s9QYVyqq6V49UtnFeQQkJUzxt7vMvOFfMzmJUeB8B7+yr485YTnKgeQoSUUmKltNVLCO/JrYM+VUFqDF8/P5ec5GhKt2xg7+4d1Hqc/R8YwOcTBWKxdY2ec8aJUmmtM8mQY4yBKhS71vpAYENrfRA490cXg2Gio5R0d2w4LT6TAM3VUmLlwHrZDl3yCkmAbPf4cPpzUHaV1eO0WyjMThjQpaelxdLu8fHS52W8/Pk5hPB2J3kaJObAwb9KQ7EhWAFxLjvXZdRwcdRRjkfP46nSeN7dd5bm/krgez2w71VZ7lpym9Q9Ky8J7k/MkSW5qkNw/ONBy2cYXgaqUIqVUk/4o7zW+KO8to2kYAbDuGXKXFEsFSHl5VqqAAVn94jj3hkvUWE2l+Si+Gn3+HDYLLS5vRyqaGJuZvyAy7DMzojjjpV5XDQrjbMN7fz5sxO8s/fsucsfsFJsLkkyrNx/7ucI0HAadeCvpGbP5NIv3sKinER2n2rgyU9K2XykmjZ3L07/Yx+Iz2nO1TDrClFy3dsCZC0TJ33pph6dJg2RYaAK5T4kyut7SKTXXuDekRLKYBjXuOIhIUcSHbUWa6WtQfwTWouiSS6Q8Fi7Kxj5BXR4fDhtVpraPaTGOpmbEX9Ol7ZZLSzLS+KuVfksy0siyi4OdbfXR3ldH31aupOYKwmX7fWSgDiYqCqt4fA74IiG+dcT5XJwyewp3L4yj9zkaD49Ws0Tm47xt0OVXS2W6iOy1Ja1DFJniiN/4Y3iRwmcF0Txzb5KAh32vtrVIjREhIFGebUrpX4FvIOJ8jIY+id9Hhz4q+R0HPtIMuTzVsnT/oE35SYYlSQ325YaiJsKSdNor/fitFtIjXXy1fNyB315l93KhTODEVD7Tjfw7r4KpsQ7WZiVwOyMOJy2fqK3pl0Mp7dLpNXeV8XyOheUkgg3d2uXMjRJMQ6uXTSVysZ2tpbWsO14LZ8fr2PGlFgWTbGSdfgNVGwaTL80eK5A5YCyYsmPWfAVOb/VLop525PipF9ye0iiqGG0MVFeBsNIkDZHHMrbnpKaVdPWBENcLTZ54v/017Jddxxaa9DOGDo8PuwWRU1zR2f+yXAwJyOeS+dMwafh3X0VPP63Y7y150zffpb4TIlas9hh9tUydupzcdZ7+3mePLVNMt6dsb2G9qbFOfniwkzuvCCfxbmJnKhuZve7T7P92Fm2RF9IVWuY5TBl6ek3iUqCuddJqZiDbxonfQQZkIVCMMrrAIBSahbwDLBspAQzGMY19igpunjqc1EmWf7i3Od9W5IFTxUHI7y0Bk8b7vozaF1AY5ubpz4pZd3iqUxLix0WcRw2C4tyEinMTuBMQxu7yuo5VNHEnAyJECuva6W53UNOcjQue4jlUnARVB2EZn8l5fYGWY6qPgpzr4H4qT0vVl4CB9+WigF5F/QrW2K0g4tnpbHKcYjapnpKoi5iz2nFx+XHSY11MH1KLPkpMWTEu7BMXSIWU+kmyUlJ8Xf+TpkO+ReKNRif5S+FYxhtBqpQekR5KaVMlJfB0BczvgDTLhFLJYArETb/tzy9B3wnWrLJPfVngAKqm9xYLYqspKhhF0kpRWZCFJkJUVw6x4fFnw+y/UQdB882SpBavIvclGiyE6PJSEjGkT5fFGP2clGOibmSdPj5HyH3fCnZEviM9ackMiy5QCoHDJSGcmylfyOtoJDL51/BBR0SlHDwbCNbjtXw2dEanHYLOUnR5CaezzTrKWL2voal6C6pQQaivBpPw+F3paRL4uCXDA2DY6AKpVgp9QTwR//21zBRXgZD/1i6+SksluCSTLfy9u3++lyVTW3kJsf07+MYIraQ6LGrFmSwJDeR0upmTlS3yE1c1/CVpdnk5q+mrnQHtZ+/g3PeF0mJz8O5/Juy9HX8E7Fa5l4rpVL2vCx5IvPWBf0e/eFph72vyfLY7KtAKWKcNhbnJLI4J5E2t5cTNS0cr27heHUzhyua+MS9lMWVr2Ct/jPuBTeRFuckJcZB0uyrsWz/g5S1WXanBEgYRo2BKpT7gL9DorwU8BHiSzEYDOeKM7aHoxpEobS5vfi0piAtppeDRwaLRTE1MYqpiVFcMF0KUZ5taCMjwQW2aI7aZ9C271NK6vNot8UT57KRFreE8/PySJ+SQUt7B9697xDd0Yq16E5Z8hsoB9+SciqLvxb2OJfdyqz0OGalx6G1pqHVQ3l9K3Wnvkxpm5PTpbX4/EraalFk2lew6OwrWBr/SPuCW0iMjSYhyk6Mw9qlRpph+BlwlBfwn/7XsKGUuhL4JVIs6HGt9U+77XcCf0B8NdXAzVrrUv++fwa+AXiB72mt3xpO2QyGEcMRC1RKLgpITSrkJl7b0kFqrJOClNFVKN1x2a3khciwZPUXaVelTIs9xokpl1Hd1E5VUzut2dMgNob9x6up2nOE0/GLcW9vIsbZRrTDyuyMOOZkxNPm9nKqrpVohxWnzYrLbsFps2Kt2C25OQUXSrJiPyilSIi2kxBth8wiVgKejjbqzp6gwppBdXM7VU1R7GxeTfaptzhd/xqlyasAsFkUcS4b0U4bsU4b0Q6r/91GjNNKlF1kc9otOG0Wo3wGwYAUilJqFfBjIC/0GK31tMFeWCllRTpBXg6UAVuVUq9prUOywfgGUKu1nqGUugX4v8DNSql5wC3AfGAqsFEpNUtrPfDSqAZDpHDESBmWQIn2pHwAmtq9xDhszJsaT1LM2Ap9Va54XHnLySjbSkbaGijI7LJ/WlocrjXfIL5DU9/qprndQ0Obh+Z2+ZesbGzntZKunRhd7joubHiDBbNnoXNX8vK2MmxWhd1qwWZR2G0W7BYLq2akoJTiWFUz7R4vNosFq0VhVQqrVTGl/D1SK/cSs/BrTE1MxWJRWOdlYDniY+GJLVTnt1PryqGu1U1Tm4fmDg9nG9po6fD22gZAKToVn8tuxWmzYLNacFgVNovFL5tfRr+8DluIXBaFxaKwWRQW/7ZVKSwWsFksWCzItlIoxYRRXgNd8noC+AHiNxmum/YK4LDW+iiAUupZYB2SNBlgHaLIAF4EfqXkN78OeNZvOR1TSh32n2/zMMk27vjgQAWNbT3LWVw8O414l51dZfWUVkvjo9CgyoVZCRSkxlBe10rx8VrZHxJ2OTUxiuX5ybR7vPzp0+OcrGmR4/1TbFYLi3PEKbr7VD1tbi+62zXmT40n2mGjtKqZSn+Yauj+3ORo0uNd1DR3cKTS3yZXB9+Soh3MmBKL2+uj5GRd53HaP8lmsbAkNyhDS0fPr2ioDBWNPVvV5iZHM6W7DIHraN2HDP7fg0WxJDepU4bWPmTQR6qJrz7MxhPbQSkW1x1mfqOTnRYbiwoTuaYwTOTUWCB3JZwugdKPJA8khMRoB4nRvVc8To93ceuKXFrdXtrcXto73MTve5uouCiYey0+LHi1prXNi8frw+3VuH0+fD7N6pliwW05Vk15Xc8w55sXn8fUumOc+eRpXrddgdfqAkDpHAortjLds4FFV3yfEw0O/rL9FBZF503cZlGkxTlZPTOV5nYvzxefxOvTeH0aj8+H16fxaU1RXjLN7R52naqnodWNT4M3pBBnQWoMMU4bZxvaqPS3Qg5VEVPiXKTFOWlq90ittZCdCmmnnJ8aA1qz/2wjgCgbRFaLUizISvAr1iZaO7wo/36UnGPGlFhinXZO17dS2dhNBgWZCS4y4qNobHNzxN8ELVTGOJedmVNi8WnNzjL5jisFSdEDr782UIVSr7V+c8BnHRhZwMmQ7TLgvN7maK09Sql6IMU//mm3Y7PCXUQpdQ9wD0Bu7sSN+mho81Df0tFj3OvvANjq9lLX2jN3oN0jNz6310dDmP2BG6PWUN/ipqa56zVsFgt1/uvWtbppDVNKo6HVjceraWxz09BN6Smgqd1DTLuHJv9LxoNfdafNS0uHF7fXJ/J0e5izWTRtbh8KaHf7wj51ur0at9eHxyev4PXlZF6fxud/6cCNIuQ6OuQVevnAz8F//uATaffPabXIeFnScnbaC3EocbrnuI8Rq5uIS83igukpPWQfMziiIXuFhOw2nJY8lYEearOIPybA4XfB2ShJia4ErMBNRX0veV1TOJV2jw+P14dXa//fDJLjnTD/etI+e4ovu7ZTN20dXg1erbFmf4WkEy/B4XeIy72CovwkfFqjNfKOPLBkJ0Xj9WnOK0iWv3PnHLmpfnGhfNaPDlZS29KB1jLHqzUen+a8ghTiXDb2ljdw6GwjPv/5fVq+UwVpseSlRFPZ0M42h/h8vD45h09rkqMdLMxOxOvzUdvq7ozbCMiqFOQkR0uhhYY2vD4ZD8gKEGW3Eu2wYrWo4Hc85MFMa/GVaej5wKOl7bTH58OndefDqUb+xweK6it5SinlD57nJsTP8TLQ+Xintf58wFfqee4bgSu01t/0b98GrNBafzdkzh7/nDL/9hHEEnkQ2Ky1/pN//Algg9b6pb6uWVRUpIuLi/uaYjCMPu/8UDLAL/nfkN9/3kZE8bRLQmZcJiy6pes+d5vk1vR3A6o+Ajufl9ycWVcMn2xlxXDoHcmdyV8VHD/2kZTNX/AVSJs1fNebRCiltmmt+03u6c9C+Xm37dATauBSBk8ZEPpIkg2U9zKnTCllAxKAmgEeazCMD/x5KNQeG/sKxeaE3AvgyHtSAsXv/8HdJorG6pBclSnzwiuW9iYpOR+T2rW0ynCQtUySHtvq6XysByl5U3VIsugTssXSMowIfT5KaK0v6eM11G/DVmCmUqpAKeVAnOyvdZvzGnCH/+cbgPe0mFSvAbcopZxKqQJgJrBliPIYDIaBkLVMck2OfhDMqak6KNZLe6MUkyx+Air2dy2DorV0WvR0wLzrB9W8q0+UgjnXwpwvdm3gZbFKnoynHQ6ZYNCRpE8LRSn1da31n5RSfx9uv9Z60GHEfp/I/cBbyHLa77XWe5RSDwLFWuvXkGCAP/qd7jWI0sE/73nEge8B/s5EeBnGPeMl0sdqkzDf/RtEkaTNltpkUYlw3r1SALN0E+z5i2SsF1wEKTOkF3zNUVnmGqnWvQGr6OxesVQCFYpjp0hG/9EPIXWvFO80DDv9LXkFAtHjRuLiWusNwIZuYz8M+bkNuLGXY/8d+PeRkMtgGFWsYytEeECkLxQFcfRDqZ1VWwo5K0QpTpkLqbOlTH/pJtj1orQVbq4UH8bUJSMvX20pnNkJSXnBemM558vS16G3JOfFOSK3tUlNnwpFa/07//tPRkccg2ESkr4QTnza/7yxhMUCBRfD7pdg94viB5oyr+v+jAWiXM7skurAjliY/cXRscSmXyrW0P71sOwusaosFphzDRT/XloLLLxh/FiF44T+lrwe7mu/1vp7wyuOwTAJUUiioyOy2fHnTOpMefpvKJeaWbFTes6xWGHqYmmLrL3D7zfpDbtL6oLtfF7yZgIBADEpEjRweKNYMJmLRkeeSUJ/S16mAKTBMNJU7JMn/ORBF56IDErJzbnkz9KNsq+nfYuFgTeIHSZSpovCOLlFluAS/Klq2UXi+zm8UaLUXAmjK9cEpr8lr6dCt5VSMVrr5pEVaXRxu92UlZXR1tZHoyHDmMTlcpGdnY3dPs47KXg7JAJpPJKUB0u+JnkpY5EZl0k1ZBWizJSSSLCtT0hgwaJbzNLXMDHQWl4rkYirWCBXKbUI+LbW+jsjKdxoUFZWRlxcHPn5+ROmns5kQGtNdXU1ZWVlFBQURFqcoeHzilI5swsyFkZamnNnLPcdsTmDCZihuSlRSaJsDvwVyj+XUGjDkBmoDfoL4Aqk4i9a6x3AhGgB3NbWRkpKilEm4wylFCkpKRPDsjQta0eetgbY8Yw0AAuQuViWGY+8By01kZNtAjHgRU2t9cluQxMm78Mok/GJ+bsZBozNCa21EvXl9desU8rf0Msq477wlYcNA2egCuWkUuoCQCulHEqpfwT2jaBcBsPkwSjGkcfmFOXRUi21vQK44mHm5VBfBmVbIyffBGGgCuVepGNjFlJHa7F/22AwDJUpJmt7VEieJkmVZVtFgQRIXyAh0Mc+guaqyMk3ARioQvFprb+mtU7XWk/RWn8dMM2aRwmPp2efE8MEYjxmyo9Xpl8inTK7L33NulJyZPa9bpa+hsBAFcrrSqlOBaKUmgu8PjIiTT7+7d/+jTlz5nD55Zdz66238tBDD7FmzRr+5V/+hYsvvphf/vKXHD9+nMsuu4zCwkIuu+wyTpw4AcCdd97Jiy++2Hmu2NhYAD744AMuuugivvSlLzFv3jzuvfdefOYfZWzSUA5xU8dnhNd4w+aUkOHMxeI7CeCMFaXSeAZOTNo+fUNmoArl/0OUSqxSahnSPfHrIyfW5KG4uJiXXnqJ7du38/LLLxPar6Wuro4PP/yQf/iHf+D+++/n9ttvZ+fOnXzta1/je9/rv0jBli1b+PnPf86uXbs4cuQIL7/88kh+FMNgaa2BFrPUMmok5UPueZJs6QuJLZoyR4pGlm6CxrMRE288MyCForVeD/wX8DbwJHC91rpkBOWaNGzatIl169YRFRVFXFwc1157bee+m2++ufPnzZs389WvfhWA2267jU2bNvV77hUrVjBt2jSsViu33nrrgI4xRIBAHkrN0UhLMrkoK5a6Xt6QTqUz14I9Cva/Dl6z1Hyu9KlQlFL/rZR62F/T61LEb3IM+G5/db4MA6OvjpkxMb3XdgqEzNpsts6lLK01HR0dPeb0tm0YK/i/Ax2tkRVjshGTKk74Yx8Gx+xRUsCyqRKOmwewc6U/C6UYqecVeP0MeClk2zBEVq9ezeuvv05bWxtNTU2sX78+7LwLLriAZ599FoCnn36a1atXA5Cfn8+2bfKnePXVV3G7g09bW7Zs4dixY/h8Pp577rnOYwwGA7L0lbVULJW6kDS71BmQWSgVoEMTIQ39ck61vAzDz/Lly7nuuutYtGgReXl5FBUVkZDQs1jdww8/zN13383PfvYz0tLS+J//+R8AvvWtb7Fu3TpWrFjBZZdd1sWqWblyJQ888AC7du3qdNAbDIYQpl0iPe73r4fl3whWQ55+mfRU2b8eiu4avSrJ4xzV15KLUup5rfVNSqlddNrlQbTWhSMp3HBTVFSkQ53eAPv27WPu3LkRkkhoamoiNjaWlpYWLrroIh599FGWLl06pHN+8MEHPPTQQ7zxxhvDJOXYZCz8/YbMlseku+GS22HxLZGWZvJRWwolz4ijPrTPfc0x2PEs5CyHGV+ImHhjAaXUNq11UX/z+isO+X3/+zVDF8nQG/fccw979+6lra2NO+64Y8jKxDDOiEqOtASTm6R8mH2ltCkOJblAikaWFUPKTKmsbOiT/pa8Tvvfjw/nRZVSycBzQD5QCtykta7tNmcx8BskEMAL/LvW+jn/vieBi4F6//Q7x3PU2Z///OdhP+eaNWtYs2bNsJ/XMAK01UNshtzYDJEh0JbY0y6l7gNLXNPWSPTdgQ1Q9A2wmSTUvugvyqtRKdUQ5tWolGoYwnUfAN7VWs8E3vVvd6cFuF1rPR+4EviFUioxZP//0lov9r/GrTIxGGgokxpTzthISzK58bRLj5SjIVFfNgfMuVqU/pH3IifbOKFPhaK1jtNax4d5xWmth1J6ZR0QcPg/BVwf5toHtdaH/D+XAxVA2hCuaTCMTbQPfG5oMsl0EcXmlGWvU8VQdyI4npgD2cuhfLs48A29Mso9OTtJD1lOOw2EaUYdRCm1AnAAoX/Nf1dK7VRK/ZdSytnHsfcopYqVUsWVlZXDIbvBMLwEsrWbzPcz4kxbIy2B928ATzCni4KLJW/lwJvgngA9eEaIEVMoSqmNSqndYV7rzvE8mcAfgbu01oFiVP8MzAGWA8nAP/V2vNb6Ua11kda6KC3NGDgGg6EPbA5JbGyt7ZrwaLXBnGugo1l60RvCMqAWwINBa91rnJ1S6qxSKlNrfdqvMCp6mRcPrAf+t9b605Bzn/b/2K6U+h/gH4dRdINhdDEVDMYWSXmQXQSntsHUpRCTIuPxmZC3Eko/htRZkDYrsnKOQSK15PUacIf/5zuAV7tPUEo5gL8Af9Bav9BtX6b/XSH+l90jKm2E2L9/P4sXL2bJkiUcORJ+7TZQXbi8vJwbbrgBgCeffJL7779/wNcJnf/b3/6WP/zhD4BEinXP2xkORuq845aopEhLYOhOwcVQeFNQmQTIWwWxU+Dgm9DREhnZxjAjZqH0w0+B55VS3wBOADcCKKWKgHu11t8EbkL61qcope70HxcID35aKZUGKKAEaQA2ZH7y+h72lg8leK0n86bG86Nr5w/q2FdeeYV169bxk5/8pN+5U6dO7VLGfrDce++w/CoN50JCdqQlMHTH5pCGXFpLSfv4TBm3WGHutbDtSTj0Fsw31SdCiYiForWu1lpfprWe6X+v8Y8X+5UJWus/aa3tIaHBneHBWutLtdYLtdYLtNZf11o3ReJzDBelpaXMnTuXb33rW8yfP5+1a9eyfv16fvGLX/D4449zySWXDOgcCxYs6DG+fv16Vq5cSVVVFZWVlXzlK19h+fLlLF++nI8//rjH/B//+Mc89NBDndsvvPACK1asYNasWfztb38DoK2tjbvuuouFCxeyZMkS3n///T7HW1tbueWWWygsLOTmm2+mtdUUQeyCzwux6RBrfHxjjtM7RHnUhqTixU6B/NVQsR/O7o2YaGORSFkoY5LBWhLDwaFDh3jmmWd47LHHuOmmm6itreXee+8lNjaWf/zHwbmI/vKXv/Cf//mfbNiwgaSkJL761a/ygx/8gNWrV3PixAmuuOIK9u3b1+c5PB4PW7ZsYcOGDfzkJz9h48aNPPLIIwDs2rWL/fv3s3btWg4ePNjr+G9+8xuio6PZuXMnO3fuNJUAulN1SPJQYtMjLYmhO1PmSZHI7omNOefL3+3QWxJW7IyLrJxjBKNQxggFBQUsXrwYgGXLllFaWjqk873//vsUFxfz9ttvEx8vKUMbN25k797gE1VDQwONjY19nufLX/5yD5k2bdrEd7/7XQDmzJlDXl4eBw8e7HX8o48+6mwIVlhYSGHhuCoBNwpo8HnAbdbkxxw2h3R4LPkzHP0AZq2VcYtFor6Kfw8H/goLbzDBFUTOKW/ohtMZTKWxWq1D7iM/bdo0GhsbOXjwYOeYz+dj8+bNlJSUUFJSwqlTp4iL6/vJKiBXqEy9FRTtp9DouX6EyYPP/7euORZZOQzhScyFLH/UV21pcDwmRfJWqg/DmV0REm5sYRTKBCUvL4+XX36Z22+/nT179gCwdu1afvWrX3XOKSkZXMWaiy66iKeffhqAgwcPcuLECWbPnj2g8d27d7Nz586hfDSDYfSZdlnR5tIAABj/SURBVLFE453c0nU8u0gUzuF3pDzLJMcolAnM7Nmzefrpp7nxxhs5cuQIDz/8MMXFxRQWFjJv3jx++9vfDuq83/nOd/B6vSxcuJCbb76ZJ598EqfT2ev4fffdR1NTE4WFhfzHf/wHK1asGOZPajCMMFa7hBHP/3LXcaVkSUxrya7vw0qfDPTZD2WiMVb7oRgGz4T4+733f+D4J7D0Dlh0c6SlMfRHSw142iB+anCsfLv4UmatlZL3E4yB9kMxForBEGnSZkdaAsNA0Vqaoe15pWutr8zFkrdy5D1ROJMUo1DGAdXV1SxevLjHq7q6OtKiGYYFixQkdMT0P9UQWZSCmWuhvQGOvt91fPZVoKzSNtjn6/0cExgTNjwOSElJGbQD3TAOqD4EXrd0CDSMfRJzxBl/cqtYl4HGaK54mHk57HtDSuDnTD5fobFQDIZI43VLcyfD+KHgYohO9pe5D/nbpS+A1JnSpKu5KnLyRQijUAyGSONzg/bCmQlZ43RiYrVLJ0d3CzSUB8eVgllXyv59r0+6pS+jUAyGsYJJ/hxfJGTDyr/ruVTpjBWl0ngGTmyOjGwRwigUg8FgGCz2KPB64PjmrktfU+ZA+jw4/jE0Tp7WzkahTEACPVLCEVqVuLi4uLPGVvcqw/0ROv+HP/whGzdKF7v8/HyqqoZ/7XikzmswDJnmSunueOS9ruMz14LNBftfF6UzCTBRXqG8+cDw1+TJWAhX/XTIp/F6vVit1mEQKEhRURFFRf3mKvXLgw8+OAzSTGJSZ0GZaTg2bonPlIiuE59J1FfyNBm3R0k74V0vwPFNUvdrgmMslDFAaWkpc+bM4Y477qCwsJAbbriBlpYW8vPzefDBB1m9ejUvvPACR44c4corr2TZsmVceOGF7N+/H4Bjx46xcuVKli9fzr/+678O+LoffPAB11xzTY/xxx57jKuuuorW1tZerxnKnXfe2aW513//93+zdOlSFi5c2Dm/pqaG66+/nsLCQs4///zOel69jVdXV7N27VqWLFnCt7/97T4LT457rI5IS2AYKvkXQXQKHHiz69JX6gzILJQS+PWnIiffKGEslFCGwZIYLAcOHOCJJ55g1apV3H333fz6178GwOVysWnTJgAuu+wyfvvb3zJz5kw+++wzvvOd7/Dee+/x/e9/n/vuu4/bb7+9syfJYPnVr37F22+/zSuvvILT6eSee+4Je82+SE1N5fPPP+fXv/41Dz30EI8//jg/+tGPWLJkCa+88grvvfcet99+OyUlJb2O/+QnP2H16tX88Ic/ZP369Tz66KND+lxjmqYK6YWS3rNBmmGcYLVJ1Nf2P8rS1+yrgvtmfEGqFO9fD0V3SQTYBMUolDFCTk4Oq1atAuDrX/86Dz/8MAA33yy1nZqamvjkk0+48cYbO49pb5cnoY8//piXXnoJgNtuu41/+qd/GpQMf/zjH8nOzuaVV17Bbrf3ec2+CO2h8vLLLwPSQyUg46WXXkp1dTX19fW9jn/00Uedx1599dUkJU3gvutt9bIObxjfJGTJ0ldLjXThtPiXqG1OWfra8az4WmZ8IbJyjiARUShKqWTgOSAfKAVu0lrXhpnnBQJOjRNa6+v84wXAs0Ay8Dlwm9a6o/vx44nu/UIC2zExUo7D5/ORmJjYa8b8cPQbWbBgASUlJZSVlVFQUNDvNXtjoD1UlFK9joe+T3h8btA+6YeSPi/S0hiGQsEaCf/u/t1NLpCikWXF4jNLzI2IeCNNpHwoDwDvaq1nAu/6t8PRGtJP/rqQ8f8L/Jf/+FrgGyMr7shz4sQJNm+WmPVnnnmG1atXd9kfHx9PQUEBL7zwAiA36B07dgCwatUqnn32WYDOviODYcmSJfzud7/juuuuo7y8vM9rniuhPVE++OADUlNTiY+PH9D4m2++SW1tj+eNiUNAqZqOjeMfi0WUSc1RyZYPZdoacCXK0pdnXD//9kqkFMo64Cn/z08B1w/0QCWPrZcCAS/wOR0/Vpk7dy5PPfUUhYWF1NTUcN999/WY8/TTT/PEE0+waNEi5s+fz6uvvgrAL3/5Sx555BGWL19Off3QmvysXr2ahx56iKuvvpqqqqper3mu/PjHP+7sxfLAAw/w1FNP9Tn+ox/9iI8++oilS5fy9ttvk5s7MZ/oDBOU2uPSkqD6SHDM5hA/S1t9zxDjCUJE+qEopeq01okh27Va6x6L5EopD1ACeICfaq1fUUqlAp9qrWf45+QAb2qtw3o0lVL3APcA5ObmLjt+/HiX/WOhn0ZpaSnXXHMNu3eb0hvnylj4+w2Zd34MZVtg2Z3SxMkw/vF6YNv/SMTX8m+C3RXcd/hd6fy46OZgiPEYJ+L9UJRSG5VSu8O81v3/7d1/cNT1mcDx95MEiAH5EZJAcIOA0FghgY6RH1qJQrUtFZS72JbOWTzlHDzP/gA82vLH2Zk6o9Zipz1unEzVwp16eNYeeoJUBEvRwYIYBRRIKlQDAZIAVUQlJM/98dlNsmE3CbK7n/3xvGYym/3ud795dgn75PPr+ZzDZUYGX8R3gF+KyCVApI71qFlRVatVtUJVKwoLC8/xVRiTAAOLfUdgYi006+v0ybNbI6MroX+BKyzZ8qmf+OIkboPyqhp1KoOIHBGRYlVtEJFi4GiUaxwK3r4nIq8AXwJ+BwwWkRxVPQMEgEORnp8qRo0aFfPWyc6dO7nlllvCjvXr14/XX389pj/HxEDu4J7PMaln4AgomeLWoBSWwtBL3PFQstnxn/CXl933acLXtOHngPnA/cHbszrmRWQIcEpVPwt2c10FPKiqKiKbgCrcTK+Iz890ZWVltodKqjj9sVuHMsT2Q0k7o65264ykS2fQwBEwcoqrAVZ4aUeySXG+BuXvB64TkVrguuB9RKRCRH4TPOeLwHYReQvYhBtDeSf42FJgkYjUAUOBRxMavTGxdPKw+9CxHRvTT3ZOcKwkwh8LF3/ZdX3tXZc2XV9eWiiq2gzMjHB8O7Ag+P1rQFmU578HZN52aCY9tbUC6pKKrUNJTy2fuOnCI77UpevrBtixKm26vqyWlzG+aau7tdXy6SurD3xyHPaudcklZGAxjJwKDW9DU52/+GLEEkoS6lpssSeHDh2iqqoqjhEZY85LqDVy+pSbNtzZxVe5rq9968KTTQqyhJIGRowYcU4JyBjjQag1cnhneGuku2STYiyhJIFVq1ZRXl7OxIkT26f6bt68mSuvvJIxY8a0JwtV5Z577mHChAmUlZWxevVqIHzTrNbWVpYsWUJZWRnl5eX8+te/BuCNN96gsrKSyy+/nK9+9as0NDR4eKUmotxBviMwiRKtNRIt2aQYqzbs2e7du7nvvvt49dVXKSgo4NixYyxatIiGhga2bNnCnj17mDNnDlVVVTz77LPU1NTw1ltv0dTUxBVXXMH06dPDrlddXc3+/ft58803ycnJ4dixY7S0tHD33XezZs0aCgsLWb16NcuWLeOxxx7z9KpNmAttYWPGyM6BL86Go++6cZXORn0Zmmtdshm0wG3QlWIsoXi2ceNGqqqqKCgoACA/Px+Am266iaysLC677DKOHHF7Um/ZsoV58+aRnZ3NsGHDqKysZNu2bZSXl7dfb8OGDSxcuJCcnJz26+3atYtdu3Zx3XXXAa4VU1xsH2JJQ9tgQBH0t0oOGeHC4e4LwsvcZ2W7rq83VkLdBpd4UowlFM9UNWKZ9lAJ+NA5nW/P9Xqqyvjx49urGZskc/wAnGx0ScVkjiO7Yf+f4PL5Ha2RC4fDxdPgwKtuwWPBOL8xniMbQ/Fs5syZPP300zQ3NwNuS9xopk+fzurVq2ltbaWxsZHNmzczeXL4cpzrr7+eRx55pH0fkmPHjlFaWkpjY2N7QmlpaWH37t1xekXmnGkboG7FvMkceUNd5eG6DeHHL74KBhQGFzym1qwvSyiejR8/nmXLllFZWcnEiRNZtGhR1HPnzp3bPng/Y8YMHnzwQYYPHx52zoIFCxg5cmT7eU8++SR9+/blmWeeYenSpUycOJFJkybx2muvxfulmd5qbXG3xw94DcMkWKg1cngXNNV2HM/Khktnu2RS+5K/+D4HL+XrfamoqNDt27eHHUuL8ucZLC3+/V76N6jfBpf/I5Tf3PP5Jn20tboy96dPweR/Ch+I3/8nOLAFyqq8d315L19vjDGmB6GB+JZPIix4vNKNq6VQ15clFGN8y+7T8zkmfV04HEq/7hJIZ52TTYp0fVlCMca3oWPdbYTZfiZDFJdDXr7b6fHMZx3HLxzmEs2R3dC4z198vWQJxRjfRNxq+T55viMxPrW1wo6VsO/F8OOhrq99L7qxliRmCcUY3068D2c+jbxnhskcWdlQ8AU48k54ayTU9XXmU6hL7q4vSyjG+NbaEt7NYTJXtNZIe9fXO9C41198PbCEkkT27NnDtGnT6NevHw899FDU8/bv38+UKVMYN24c3/rWtzh9+nQCozQxF0omR2yxacYLm/XVpTUycppLLEnc9WUJJYnk5+fzq1/9iiVLlnR73tKlS/nhD39IbW0tQ4YM4dFHbQfk9GCD8gaXNEZd5VojJz7oON7e9fUZ1P7BX3zd8FLLS0TygdXAKOAA8E1VPd7lnGuBhzsduhT4tqr+r4j8FqgE/hZ87FZVrTnfuH7xi1+wd29sm5OlpaUsXry4V+cWFRVRVFTECy+8EPUcVWXjxo08+eSTAMyfP597772XO++8MybxGg8yaHGx6aWR06B/EQwKhB8fUORKs+zf7Gp9FV3qJ74ofLVQfgS8rKrjgJeD98Oo6iZVnaSqk4AZwCmgc1q+J/R4LJJJqmhubmbw4MHt1YQDgQAHDx70HJUxJqaysqHwC24G4EdHwh8bOdW1YmrXJ139N1/Vhm8Ergl+vxJ4BVjazflVwDpVjWvHYW9bEj5FKpUTqVqxSSFDx8LB7T2fZzJPUx3s/B8YP7ejNdJe5v63rutr/FyvIXbmq4UyTFUbAIK3PdXt/jbwVJdj94nI2yLysIj0i/SkVLBixQomTZrEpEmTOHToUI/nFxQUcOLEifZqwvX19YwYMSLeYZp4spXyJpr80ZFbIwOK3IZcR/e4ryQRt4QiIhtEZFeErxvP8TrFQBmwvtPhH+PGVK4A8ummdSMid4jIdhHZ3tjY+DleSXzddddd1NTUUFNT06vEICJce+217dsCr1y5khtvPKe31CSbU83uA2LYZb4jMcmmu4H4kqmubEsSdX3FLaGo6ldUdUKErzXAkWCiCCWMo91c6pvA71W1pdO1G9T5DHgcmBztyaparaoVqlpRWJjcO+IdPnyYQCDA8uXL+dnPfkYgEODDDz8EYNasWe0tmAceeIDly5czduxYmpubuf32232Gbc7Xp3+Dk939FzAZLTQQ37U1kpWVdLO+fI2hPAfMB+4P3q7p5tx5uBZJOxEpVtUGcYMHNwG74hVoIg0fPpz6+vqIj61du7b9+zFjxvDnP/85UWGZeOu8H0pRipfiN/Excio07YX3X4PC0o66bwMKXdfXe3+Ewne9//74Sij3A0+LyO3A+8DNACJSASxU1QXB+6OAEuCPXZ7/hIgU4ibu1wALExO2MXGgbe42SRermSSQle0G33MuOLuIaMlUaNoH+9bD4JHQt7+fGPGUUFS1GZgZ4fh2YEGn+weAiyKcNyOe8RljTNK5YIi7/ewj+OS4Sx7Q0fW1/TGXVMbP9Va52lbKG5M0bPq36YU9L8CuZ8MH4vsXwKirXZ2vo+96C80SijG+DSz2HYFJJZfMhNbTrjXSWckU97tU+wf47KSX0CyhGONb7iDfEZhUEhqI79oaCXV9tba4qcQeSvpYQjHGt9On3NTQISN9R2JSRclU1xrZt/7srq/RV7v9VDx0fVlCSQK33XYbRUVFTJgwIeLjqsr3vvc9xo4dS3l5OTt27EhwhCauTh5x61D6DvAdiUkVodZIWwsc/2v4Y4HJMHCEa6UkuOvLEkoSuPXWW3nxxRejPr5u3Tpqa2upra2lurraKgunm9bgfjYfJ18lB5PE+hfA1H8+u8JCe9fXGbd3SgK7vnytQ0lad9xxR8Tj1dXVQPQS94sXL6a0tJTnn3+e559//qzndWf69OkcOHAg6uNr1qzhu9/9LiLC1KlTOXHiBA0NDRQX22BuWjj2nrv96LDfOEzq6dsf2trg0A5Xzr5fsJXbf6jr+vrLJjj6Dgwbn5BwrIWSAg4ePEhJSUn7fStZn6ZCCxyNORefnnCJo+tAfHvXV+JmfVkLpYueWhQ9lbifPXs2s2fPjmVIVrI+U4j9fWc+h7z8yK2RsAWPL8KEv4/7gkf7DU4BgUCADz7o2ArUStanKWuhmM8rWmuk/1AYPR2aauHI7riHYQklBcyZM4dVq1ahqmzdupVBgwbZ+Ek6ajvjOwKTqrobiA9cAYMugrqXXNmWOLIuryQwb948XnnlFZqamggEAvz0pz+lpcVVoF24cCGzZs1i7dq1jB07lry8PB5//HHPEZuYumCI2xOlf3Jvr2CSXGggvqnWzRzMCe47GEo22x5161bi2PVlCSUJPPVU180ow4kIK1asSFA0JuFOB7soBvS0cakxPQhMdl9ZXTqf8vJhTCXUvQxHdsHwsrj8eOvyMsa30H4oSbLrnklhWVnu68NDULchvOvrogoYFIDa+HV9WUIxJlmE1qMYc75OfAAfbAsfiM/Kgku/AW2tsDc+Cx4toRB5Wq5Jfmn375aV7TsCky6iDcTn5cOYa6C5znV9xVjGJ5Tc3Fyam5vT78Mpzakqzc3N5Obm+g4ldrL7+o7ApIuwWV9dFzx26vr69MOY/tiMH5QPBALU19fT2Gh1lFJNbm4ugUDAdxgxZH/UmBgKG4jfDcODxWdFXNfX9kfdFOOym2M26yvjE0qfPn0YPXq07zBMJgt1ddme8ibWLqqAE+9DVpeP+rx8GH2NG7g/vBOKy2Py47x0eYnIzSKyW0TaRKSim/O+JiJ7RaRORH7U6fhoEXldRGpFZLWIWF+BSV2hsvV98/zGYdJPVhaUVUHRpWc/FqiAwSUuqcSo68vXGMou4O+AzdFOEJFsYAXwdeAyYJ6IhOo0PwA8rKrjgOPA7fEN15h4CnZ1DbGWsomT1ha3F/3hnR3HRKB0FmhrzMrce0koqvquqp5dAz7cZKBOVd9T1dPAfwM3iquKOAN4JnjeSuCm+EVrTJyFSq60WS0vEyeSDaeOnT0Qn5cPY66F5r/A4bfP+8ck8xjKRcAHne7XA1OAocAJVT3T6fhF0S4iIncAoU1OTopIT4ks3gqAJs8xJAt7LzoUwDp7Lxz7vegQh/ciWsX0n3T3pIt7c+W4JRQR2QAMj/DQMlVd05tLRDim3RyPSFWrgZ53uUoQEdmuqlHHjTKJvRcd7L3oYO9Fh1R7L+KWUFT1K+d5iXqgpNP9AHAIl60Hi0hOsJUSOm6MMcajZF7YuA0YF5zR1Rf4NvCcuhWIm4Cq4Hnzgd60eIwxxsSRr2nDc0WkHpgGvCAi64PHR4jIWoBg6+NfgPXAu8DTqhoqTLMUWCQidbgxlUcT/RrOQ9J0vyUBey862HvRwd6LDin1XoiVHDHGGBMLydzlZYwxJoVYQjHGGBMTllA8EpElIqIiUuA7Fl9E5OciskdE3haR34vIYN8xJVq0EkOZRkRKRGSTiLwbLM30fd8x+SQi2SLypoj8n+9YessSiiciUgJcB7zvOxbPXgImqGo5sA/4sed4EqqHEkOZ5gywWFW/CEwF7srg9wLg+7gJSSnDEoo/DwP/SobXLFfVP3SqerAVt64ok0QsMeQ5Ji9UtUFVdwS//wj3YRq1CkY6E5EA8A3gN75jOReWUDwQkTnAQVV9y3csSeY2YJ3vIBIsUomhjPwQ7UxERgFfAl73G4k3v8T9wZlSBd6SuZZXSuuu9AyuaM71iY3In96U4RGRZbgujycSGVsSOKdSQplARAYAvwN+oKqx3VIwBYjIDcBRVX1DRK7xHc+5sIQSJ9FKz4hIGTAaeMsVTiYA7BCRyap6OIEhJkxPZXhEZD5wAzBTM29hVLQSQxlJRPrgkskTqvqs73g8uQqYIyKzgFxgoIj8l6r+g+e4emQLGz0TkQNAhapmZHVVEfkasByoVNWM24dZRHJwkxFmAgdxJYe+06kqRMYIbk2xEjimqj/wHU8yCLZQlqjqDb5j6Q0bQzG+/TtwIfCSiNSIyCO+A0qkHkoMZZqrgFuAGcHfhZrgX+kmRVgLxRhjTExYC8UYY0xMWEIxxhgTE5ZQjDHGxIQlFGOMMTFhCcUYY0xMWEIxJkmIyE98x2DM+bBpw8YkgeCivo9VNc93LMZ8XtZCMcYTERkV3PvjP4AdwAXBxXyZVs/MpAlroRjjSbCi7nvAlaq6VUROquoAv1EZ8/lZC8UYv/6qqlt9B2FMLFhCMcavj30HYEysWEIxJnm0BMu3G5OSLKEYkzyqgbdtUN6kKhuUN8YYExPWQjHGGBMTllCMMcbEhCUUY4wxMWEJxRhjTExYQjHGGBMTllCMMcbEhCUUY4wxMfH/FxIs2Eqiz4MAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot\n",
    "ax = sns.lineplot(x = \"rt\", \n",
    "                  y = \"likelihood\", \n",
    "                  data = data_pred_long, \n",
    "                  hue = 'group',\n",
    "                  style = 'choice',\n",
    "                  alpha = 0.5)\n",
    "\n",
    "\n",
    "# g.add_legend(title=\"Meal\")\n",
    "# g.set_axis_labels(\"Total bill ($)\", \"\")\n",
    "ax.set(xlim=(-5, 5))\n",
    "ax.set(ylim=(-1, 1))\n",
    "# g.despine(trim=True)\n",
    "# g.fig.set_size_inches(6.5, 3.5)\n",
    "# g.ax.set_xticks([5, 15, 25, 35, 45, 55], minor=True);\n",
    "# plt.setp(g.ax.get_yticklabels(), rotation=30);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'hidden_units': [150, 40, 40],\n",
       " 'activations': ['relu', 'relu', 'relu'],\n",
       " 'output_activation': 'linear',\n",
       " 'optimizer': 'adam',\n",
       " 'learning_rate': 0.005,\n",
       " 'loss_fn': 'mse',\n",
       " 'beta1': 0.9,\n",
       " 'beta2': 0.9990000000000001,\n",
       " 'rho': 0.9,\n",
       " 'l_1': 0.0,\n",
       " 'l_2': 0.0,\n",
       " 'batch_size': 10000,\n",
       " 'mse_test': 32.600433,\n",
       " 'mae_test': 1.8839523999999999,\n",
       " 'training_steps': 14000,\n",
       " 'training_time': 2664.7,\n",
       " 'time_started': '08_15_18_18_57_28'}"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_params"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "tf",
   "language": "python",
   "name": "tf"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}