Hill-Valley Concrete.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "from torch import nn\n",
    "from torch.nn import functional as F\n",
    "from torch import optim\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from stochastic_rnn import StochasticLSTM\n",
    "from variational_module import MCSingleMixin\n",
    "from train_utils import train_model, BCLogitMetric"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import minmax_scale\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import roc_auc_score, precision_score, accuracy_score, recall_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from classification_uncertainty_utils import aleatoric_uncertainty, epistemic_uncertainty, uncertainty_avg, score_avg"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<torch._C.Generator at 0x7f948414ca10>"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.random.seed(1234)\n",
    "torch.manual_seed(1234)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "training_table = pd.read_table(\"./data/hill-valley/Hill_Valley_with_noise_Training.data\", sep=',', dtype=np.float32)\n",
    "testing_table = pd.read_table(\"./data/hill-valley/Hill_Valley_with_noise_Testing.data\", sep=',', dtype=np.float32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "dim = 1\n",
    "seq_length = 100\n",
    "# DROP = 0.5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train = minmax_scale(training_table.drop(\"class\", axis=1).values, axis=1)\n",
    "y_train = training_table[\"class\"].values\n",
    "x_test = minmax_scale(testing_table.drop(\"class\", axis=1).values, axis=1)\n",
    "y_test = testing_table[\"class\"].values\n",
    "\n",
    "x_train = np.expand_dims(x_train, axis=-1)\n",
    "x_test = np.expand_dims(x_test, axis=-1)\n",
    "\n",
    "x_train, x_val, y_train, y_val = train_test_split(x_train, y_train, test_size=0.3, random_state=42)\n",
    "\n",
    "N = len(y_train)\n",
    "\n",
    "train_dl = torch.utils.data.DataLoader(\n",
    "        [(x_train[i], y_train[i]) for i in range(len(y_train))],\n",
    "        batch_size=10,\n",
    "        shuffle=True\n",
    ")\n",
    "val_dl = torch.utils.data.DataLoader(\n",
    "        [(x_val[i], y_val[i]) for i in range(len(y_val))],\n",
    "        batch_size=10,\n",
    "        shuffle=False\n",
    ")\n",
    "test_dl = torch.utils.data.DataLoader(\n",
    "        [(x_test[i], y_test[i]) for i in range(len(y_test))],\n",
    "        batch_size=len(y_test),\n",
    "        shuffle=False\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Net(nn.Module, MCSingleMixin):\n",
    "    def __init__(self):\n",
    "        super(Net, self).__init__()\n",
    "        self.rnn = StochasticLSTM(1, 10)\n",
    "        self.fc = nn.Linear(10, 1)\n",
    "\n",
    "    def regularizer(self):\n",
    "        # Weight and bias regularizer\n",
    "        weight_sum, bias_sum, dropout_reg = self.rnn.regularizer()\n",
    "        \n",
    "        return weight_sum + bias_sum + dropout_reg\n",
    "    \n",
    "    def get_output_shape(self, x: torch.Tensor):\n",
    "        B = x.shape[1]\n",
    "        return (B,)\n",
    "\n",
    "    def forward(self, x: torch.Tensor):\n",
    "        out, _ = self.rnn(x)\n",
    "        out = out[-1]\n",
    "        out = torch.relu(out)\n",
    "        out = self.fc(out).flatten()\n",
    "        return out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "net = Net()\n",
    "criterion = nn.BCEWithLogitsLoss()\n",
    "\n",
    "optimizer = optim.Adam(net.parameters(), lr=1e-3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Original dropout rate = 0.51\n"
     ]
    }
   ],
   "source": [
    "print(\"Original dropout rate = %.2f\" % net.rnn.first_layer.p_logit.sigmoid().item())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "metrics = [BCLogitMetric()]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1, 100.00%] train loss: 0.698511\n",
      "[1, 100.00%] validate loss: 0.725518\n",
      "Accuracy = 0.484\n",
      "[2, 100.00%] train loss: 0.689498\n",
      "[2, 100.00%] validate loss: 0.710731\n",
      "Accuracy = 0.484\n",
      "[3, 100.00%] train loss: 0.665226\n",
      "[3, 100.00%] validate loss: 0.669350\n",
      "Accuracy = 0.484\n",
      "[4, 100.00%] train loss: 0.606342\n",
      "[4, 100.00%] validate loss: 0.610949\n",
      "Accuracy = 0.510\n",
      "[5, 100.00%] train loss: 0.548160\n",
      "[5, 100.00%] validate loss: 0.547669\n",
      "Accuracy = 0.599\n",
      "[6, 100.00%] train loss: 0.484353\n",
      "[6, 100.00%] validate loss: 0.495227\n",
      "Accuracy = 0.658\n",
      "[7, 100.00%] train loss: 0.436392\n",
      "[7, 100.00%] validate loss: 0.446316\n",
      "Accuracy = 0.700\n",
      "[8, 100.00%] train loss: 0.395179\n",
      "[8, 100.00%] validate loss: 0.373522\n",
      "Accuracy = 0.737\n",
      "[9, 100.00%] train loss: 0.345471\n",
      "[9, 100.00%] validate loss: 0.353101\n",
      "Accuracy = 0.762\n",
      "[10, 100.00%] train loss: 0.298964\n",
      "[10, 100.00%] validate loss: 0.287983\n",
      "Accuracy = 0.785\n",
      "[11, 100.00%] train loss: 0.265760\n",
      "[11, 100.00%] validate loss: 0.281546\n",
      "Accuracy = 0.802\n",
      "[12, 100.00%] train loss: 0.252254\n",
      "[12, 100.00%] validate loss: 0.246042\n",
      "Accuracy = 0.816\n",
      "[13, 100.00%] train loss: 0.217434\n",
      "[13, 100.00%] validate loss: 0.199949\n",
      "Accuracy = 0.830\n",
      "[14, 100.00%] train loss: 0.212227\n",
      "[14, 100.00%] validate loss: 0.172479\n",
      "Accuracy = 0.841\n",
      "[15, 100.00%] train loss: 0.164067\n",
      "[15, 100.00%] validate loss: 0.155442\n",
      "Accuracy = 0.851\n",
      "[16, 100.00%] train loss: 0.163447\n",
      "[16, 100.00%] validate loss: 0.133153\n",
      "Accuracy = 0.861\n",
      "[17, 100.00%] train loss: 0.150549\n",
      "[17, 100.00%] validate loss: 0.143220\n",
      "Accuracy = 0.868\n",
      "Retry 1/5\n",
      "[18, 100.00%] train loss: 0.143039\n",
      "[18, 100.00%] validate loss: 0.122868\n",
      "Accuracy = 0.875\n",
      "[19, 100.00%] train loss: 0.127921\n",
      "[19, 100.00%] validate loss: 0.137693\n",
      "Accuracy = 0.881\n",
      "Retry 1/5\n",
      "[20, 100.00%] train loss: 0.122556\n",
      "[20, 100.00%] validate loss: 0.098189\n",
      "Accuracy = 0.887\n",
      "[21, 100.00%] train loss: 0.101202\n",
      "[21, 100.00%] validate loss: 0.113990\n",
      "Accuracy = 0.892\n",
      "Retry 1/5\n",
      "[22, 100.00%] train loss: 0.118158\n",
      "[22, 100.00%] validate loss: 0.119124\n",
      "Accuracy = 0.896\n",
      "Retry 2/5\n",
      "[23, 100.00%] train loss: 0.100725\n",
      "[23, 100.00%] validate loss: 0.092031\n",
      "Accuracy = 0.901\n",
      "[24, 100.00%] train loss: 0.093123\n",
      "[24, 100.00%] validate loss: 0.096982\n",
      "Accuracy = 0.904\n",
      "Retry 1/5\n",
      "[25, 100.00%] train loss: 0.087164\n",
      "[25, 100.00%] validate loss: 0.074929\n",
      "Accuracy = 0.908\n",
      "[26, 100.00%] train loss: 0.097499\n",
      "[26, 100.00%] validate loss: 0.082780\n",
      "Accuracy = 0.911\n",
      "Retry 1/5\n",
      "[27, 100.00%] train loss: 0.081154\n",
      "[27, 100.00%] validate loss: 0.084333\n",
      "Accuracy = 0.914\n",
      "Retry 2/5\n",
      "[28, 100.00%] train loss: 0.061270\n",
      "[28, 100.00%] validate loss: 0.073728\n",
      "Accuracy = 0.916\n",
      "[29, 100.00%] train loss: 0.081510\n",
      "[29, 100.00%] validate loss: 0.110253\n",
      "Accuracy = 0.919\n",
      "Retry 1/5\n",
      "[30, 100.00%] train loss: 0.087376\n",
      "[30, 100.00%] validate loss: 0.049478\n",
      "Accuracy = 0.921\n",
      "[31, 100.00%] train loss: 0.059790\n",
      "[31, 100.00%] validate loss: 0.055895\n",
      "Accuracy = 0.924\n",
      "Retry 1/5\n",
      "[32, 100.00%] train loss: 0.056534\n",
      "[32, 100.00%] validate loss: 0.073080\n",
      "Accuracy = 0.926\n",
      "Retry 2/5\n",
      "[33, 100.00%] train loss: 0.062851\n",
      "[33, 100.00%] validate loss: 0.046636\n",
      "Accuracy = 0.928\n",
      "[34, 100.00%] train loss: 0.059067\n",
      "[34, 100.00%] validate loss: 0.055006\n",
      "Accuracy = 0.930\n",
      "Retry 1/5\n",
      "[35, 100.00%] train loss: 0.057432\n",
      "[35, 100.00%] validate loss: 0.050422\n",
      "Accuracy = 0.931\n",
      "Retry 2/5\n",
      "[36, 100.00%] train loss: 0.058070\n",
      "[36, 100.00%] validate loss: 0.046524\n",
      "Accuracy = 0.933\n",
      "[37, 100.00%] train loss: 0.048201\n",
      "[37, 100.00%] validate loss: 0.042580\n",
      "Accuracy = 0.935\n",
      "[38, 100.00%] train loss: 0.042664\n",
      "[38, 100.00%] validate loss: 0.075692\n",
      "Accuracy = 0.936\n",
      "Retry 1/5\n",
      "[39, 100.00%] train loss: 0.051120\n",
      "[39, 100.00%] validate loss: 0.044134\n",
      "Accuracy = 0.938\n",
      "Retry 2/5\n",
      "[40, 100.00%] train loss: 0.057094\n",
      "[40, 100.00%] validate loss: 0.034216\n",
      "Accuracy = 0.939\n",
      "[41, 100.00%] train loss: 0.041763\n",
      "[41, 100.00%] validate loss: 0.052857\n",
      "Accuracy = 0.940\n",
      "Retry 1/5\n",
      "[42, 100.00%] train loss: 0.035857\n",
      "[42, 100.00%] validate loss: 0.059964\n",
      "Accuracy = 0.942\n",
      "Retry 2/5\n",
      "[43, 100.00%] train loss: 0.024225\n",
      "[43, 100.00%] validate loss: 0.044314\n",
      "Accuracy = 0.943\n",
      "Retry 3/5\n",
      "[44, 100.00%] train loss: 0.049212\n",
      "[44, 100.00%] validate loss: 0.050175\n",
      "Accuracy = 0.944\n",
      "Retry 4/5\n",
      "[45, 100.00%] train loss: 0.034007\n",
      "[45, 100.00%] validate loss: 0.021000\n",
      "Accuracy = 0.945\n",
      "[46, 100.00%] train loss: 0.048493\n",
      "[46, 100.00%] validate loss: 0.023278\n",
      "Accuracy = 0.946\n",
      "Retry 1/5\n",
      "[47, 100.00%] train loss: 0.034920\n",
      "[47, 100.00%] validate loss: 0.020451\n",
      "Accuracy = 0.948\n",
      "[48, 100.00%] train loss: 0.030065\n",
      "[48, 100.00%] validate loss: 0.040968\n",
      "Accuracy = 0.949\n",
      "Retry 1/5\n",
      "[49, 100.00%] train loss: 0.029009\n",
      "[49, 100.00%] validate loss: 0.047160\n",
      "Accuracy = 0.949\n",
      "Retry 2/5\n",
      "[50, 100.00%] train loss: 0.027068\n",
      "[50, 100.00%] validate loss: 0.038507\n",
      "Accuracy = 0.950\n",
      "Retry 3/5\n",
      "[51, 100.00%] train loss: 0.029070\n",
      "[51, 100.00%] validate loss: 0.018035\n",
      "Accuracy = 0.951\n",
      "[52, 100.00%] train loss: 0.041065\n",
      "[52, 100.00%] validate loss: 0.039269\n",
      "Accuracy = 0.952\n",
      "Retry 1/5\n",
      "[53, 100.00%] train loss: 0.029021\n",
      "[53, 100.00%] validate loss: 0.040957\n",
      "Accuracy = 0.953\n",
      "Retry 2/5\n",
      "[54, 100.00%] train loss: 0.054076\n",
      "[54, 100.00%] validate loss: 0.036984\n",
      "Accuracy = 0.954\n",
      "Retry 3/5\n",
      "[55, 100.00%] train loss: 0.027186\n",
      "[55, 100.00%] validate loss: 0.063007\n",
      "Accuracy = 0.954\n",
      "Retry 4/5\n",
      "[56, 100.00%] train loss: 0.046302\n",
      "[56, 100.00%] validate loss: 0.013519\n",
      "Accuracy = 0.955\n",
      "[57, 100.00%] train loss: 0.037685\n",
      "[57, 100.00%] validate loss: 0.020249\n",
      "Accuracy = 0.956\n",
      "Retry 1/5\n",
      "[58, 100.00%] train loss: 0.027079\n",
      "[58, 100.00%] validate loss: 0.034836\n",
      "Accuracy = 0.956\n",
      "Retry 2/5\n",
      "[59, 100.00%] train loss: 0.040661\n",
      "[59, 100.00%] validate loss: 0.013346\n",
      "Accuracy = 0.957\n",
      "[60, 100.00%] train loss: 0.040413\n",
      "[60, 100.00%] validate loss: 0.017863\n",
      "Accuracy = 0.958\n",
      "Retry 1/5\n",
      "[61, 100.00%] train loss: 0.025801\n",
      "[61, 100.00%] validate loss: 0.016339\n",
      "Accuracy = 0.959\n",
      "Retry 2/5\n",
      "[62, 100.00%] train loss: 0.025162\n",
      "[62, 100.00%] validate loss: 0.013947\n",
      "Accuracy = 0.959\n",
      "Retry 3/5\n",
      "[63, 100.00%] train loss: 0.023497\n",
      "[63, 100.00%] validate loss: 0.021759\n",
      "Accuracy = 0.960\n",
      "Retry 4/5\n",
      "[64, 100.00%] train loss: 0.017226\n",
      "[64, 100.00%] validate loss: 0.012038\n",
      "Accuracy = 0.960\n",
      "[65, 100.00%] train loss: 0.022109\n",
      "[65, 100.00%] validate loss: 0.012501\n",
      "Accuracy = 0.961\n",
      "Retry 1/5\n",
      "[66, 100.00%] train loss: 0.021894\n",
      "[66, 100.00%] validate loss: 0.031182\n",
      "Accuracy = 0.962\n",
      "Retry 2/5\n",
      "[67, 100.00%] train loss: 0.029992\n",
      "[67, 100.00%] validate loss: 0.027236\n",
      "Accuracy = 0.962\n",
      "Retry 3/5\n",
      "[68, 100.00%] train loss: 0.039598\n",
      "[68, 100.00%] validate loss: 0.024561\n",
      "Accuracy = 0.963\n",
      "Retry 4/5\n",
      "[69, 100.00%] train loss: 0.019219\n",
      "[69, 100.00%] validate loss: 0.015293\n",
      "Accuracy = 0.963\n",
      "Retry 5/5\n",
      "CPU times: user 7min 14s, sys: 4.44 s, total: 7min 18s\n",
      "Wall time: 7min 55s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "net = train_model(net, train_dl, val_dl, criterion, optimizer, \"hill-valley-concrete.pt\", 100, metrics=metrics, patience=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimized dropout rate = 0.24\n"
     ]
    }
   ],
   "source": [
    "print(\"Optimized dropout rate = %.2f\" % net.rnn.first_layer.p_logit.sigmoid().item())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy = 1.00\n",
      "Precision = 1.00\n",
      "AUCROC = 1.00\n"
     ]
    }
   ],
   "source": [
    "net.eval()\n",
    "with torch.no_grad():\n",
    "    for i, (x, y) in enumerate(test_dl):\n",
    "        pre_sigmoids = net.sample(10, x.transpose(-2,-3))\n",
    "        scores = torch.sigmoid(pre_sigmoids)\n",
    "        score_average = score_avg(scores)\n",
    "        predictions = torch.round(score_average)\n",
    "        uncertainties = uncertainty_avg(scores)\n",
    "        print(\"Accuracy = %.2f\" % accuracy_score(y, predictions))\n",
    "        print(\"Precision = %.2f\" % precision_score(y, predictions))\n",
    "        print(\"AUCROC = %.2f\" % roc_auc_score(y, score_average))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "A = aleatoric_uncertainty(scores)\n",
    "E = epistemic_uncertainty(scores)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f9488297c88>"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAY4AAAEGCAYAAABy53LJAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAgAElEQVR4nO3de3hU5bnw/+89E0KOQEIC2xgQ3KIVkESJ1HejAroVqigesKJYwGPVqlW3fbGv3ajIbmt1a0vVVlRUKipKX5BfPdAiUvdPpRIUFajIQZQQhZAAkkAOM3O/f6w1YRImMJPMZCbJ/bmuuTLzrEOelcC6s57D/YiqYowxxkTKk+gKGGOM6VgscBhjjImKBQ5jjDFRscBhjDEmKhY4jDHGRCUl0RVoD3l5eTpgwIBEV8MYYzqU1atX71LV/OblXSJwDBgwgNLS0kRXwxhjOhQR+SpcuTVVGWOMiYoFDmOMMVGxwGGMMSYqXaKPI5yGhgbKysqora1NdFW6hLS0NAoLC+nWrVuiq2KMaaMuGzjKysrIzs5mwIABiEiiq9OpqSqVlZWUlZUxcODARFfHGNNGXbapqra2lt69e1vQaAciQu/eve3pzph2EggoFfvq2L57PxX76ggEYpvMNq6BQ0TGicgGEdkkIneH2X6niKwXkU9F5G0ROSZk21QR2ei+poaUDxeRz9xzzpY23PktaLQf+1kb0z4CAWXDjn1c/MR7jHzwHS5+4j027NgX0+ARt8AhIl7gceAHwGDgChEZ3Gy3j4ESVR0GLAR+4x6bC9wLfB8YAdwrIjnuMX8ArgcGua9x8boGY4zpaCpr6rl+Xilluw8AULb7ANfPK6Wypj5m3yOeTxwjgE2qukVV64GXgQmhO6jqO6q63/24Eih0348F/qaqVaq6G/gbME5EjgJ6qOpKdRYSmQdcFMdr6FQWL17M+vXrj7jfH//4R+bNm3fYfdasWcMbb7wRq6oZY2Kk3udvDBpBZbsPUO/zx+x7xDNwHA1sC/lc5pa15FrgzSMce7T7PtJzGpfP54s4cNx4441MmTLlsPtY4DAmMY7Uf5Ga4qUwJ71JWWFOOqkp3pjVISk6x0XkKqAEeCiG57xBREpFpLSioqLN54tHZ9PWrVsZOnRo4+eHH36Y++67j9GjRzN9+nRGjBjB8ccfz//8z/8A4Pf7ueuuuxg6dCjDhg3j97//PQCrV69m1KhRDB8+nLFjx/LNN98AMHr0aG6//XZKSkp48MEHWbJkCT/72c8oLi5m8+bNPPXUU5x66qkUFRVx6aWXsn+/8/B333338fDDDzeeo3ld6uvrmTFjBgsWLKC4uJgFCxYwaNAggj/nQCDAcccdRyx+7saYgyLpv+idmcpTU0oag0dhTjpPTSmhd2ZqzOoRz+G424F+IZ8L3bImROTfgXuAUapaF3Ls6GbHrnDLC5uVH3JOAFWdA8wBKCkpadNdPvjLCrYbBn8RJ/TNxuOJT6evz+fjww8/5I033uD+++9n2bJlzJkzh61bt7JmzRpSUlKoqqqioaGBW2+9lddee438/HwWLFjAPffcw9y5cwGor69vzNO1ceNGxo8fz8SJEwHo1asX119/PQC/+MUveOaZZ7j11lsjqsvMmTMpLS3lscceA+Dzzz9n/vz53H777SxbtoyioiLy8w/JjWaMaYOW+i8W3TyS/OzuAHg8wgl9s1l080jqfX5SU7z0zkyN6b0qnk8cq4BBIjJQRFKBScCS0B1E5GTgSeBCVd0ZsmkpcK6I5Lid4ucCS1X1G+A7ETnNHU01BXgtjtcAtE9nU3OXXHIJAMOHD2fr1q0ALFu2jB//+MekpDjxPjc3lw0bNrB27VrOOecciouLmTVrFmVlB1vzLr/88ha/x9q1aznjjDM46aSTmD9/PuvWrYu4Ls1dc801jf0ic+fO5eqrr47qeo0xRxZp/4XHI+Rnd+fonAzys7vH/A/cuD1xqKpPRG7BCQJeYK6qrhORmUCpqi7BaZrKAl51h2t+raoXqmqViDyAE3wAZqpqlfv+ZuA5IB2nT+RN4ixenU0pKSkEAoHGz6HzHLp3d/568Hq9+Hy+Fs+hqgwZMoQPPvgg7PbMzMwWj502bRqLFy+mqKiI5557jhUrVoTdL5K69OvXj759+7J8+XI+/PBD5s+f3+L3Nca0TrD/IvR+FOv+i0jEtY9DVd9Q1eNV9V9V9b/cshlu0EBV/11V+6pqsfu6MOTYuap6nPt6NqS8VFWHuue8xR1dFVfx6mzq27cvO3fupLKykrq6Ov7yl78cdv9zzjmHJ598svHmXVVVxQknnEBFRUVj4GhoaGjxySE7O5t9+/Y1ft63bx9HHXUUDQ0NUd/om58L4LrrruOqq67isssuw+tt33/IxnQF7dF/EYmk6BxPdvH6ZXXr1o0ZM2YwYsQIzjnnHL73ve8ddv/rrruO/v37M2zYMIqKinjxxRdJTU1l4cKFTJ8+naKiIoqLi3n//ffDHj9p0iQeeughTj75ZDZv3swDDzzA97//fUaOHHnE793cmDFjWL9+fWPnOMCFF15IdXW1NVMZEyeh/RfvTR/DoptHxrWvtSXSDn+wJ1xJSYk2X8jpn//8JyeeeGLE5wgElMqa+rh1NnUGpaWl3HHHHY2jwJqL9mdujEksEVmtqiXNy7tsksNoBTubTHi//vWv+cMf/mB9G8Z0AdZUZWLi7rvv5quvvuL0009PdFWMMXFmgcMYY0xULHAYY4yJigUOY4wxUbHAYYwxJioWOJLMgAED2LVrV6uO/e1vf9uYqDAa1113XURZc40xBixwdCqtCRx+v5+nn36awYObr7FljDHhWeCIVCAA1Ttgzzbna0iOqda66KKLGD58OEOGDGHOnDmHbH/hhRcYMWIExcXF/PjHP8bvd3Jj3XTTTZSUlDBkyBDuvfdeAGbPnk15eTljxoxhzJgxALz00kucdNJJDB06lOnTpzeeNysri//4j/+gqKiIDz74gNGjRzdm0H3rrbc45ZRTKCoq4uyzz27zNRpjOiFV7fSv4cOHa3Pr168/pKxFfr/qN5+pPjpU9d4eztdvPnPK26CyslJVVffv369DhgzRXbt26THHHKMVFRW6fv16HT9+vNbX16uq6k033aTPP/98k+N8Pp+OGjVKP/nkE1XVxmNVVbdv3679+vXTnTt3akNDg44ZM0YXLVqkqqqALliwoLEeo0aN0lWrVunOnTu1sLBQt2zZ0uT7xEpUP3NjTMLhJKQ95J5qTxyR2F8BL18Be752Pu/52vm8v20LFc2ePZuioiJOO+00tm3bxsaNGxu3vf3226xevZpTTz2V4uJi3n77bbZs2QLAK6+8wimnnMLJJ5/MunXrwvZPrFq1itGjR5Ofn09KSgqTJ0/m3XffBZwst5deeukhx6xcuZIzzzyTgQMHAk7admOMac5SjkTCV38waATt+dopb6UVK1awbNkyPvjgAzIyMhg9enSTtOqqytSpU/nVr37V5Lgvv/yShx9+mFWrVpGTk8O0adOaHBeJtLQ0y15rTCcW0ABVtVXU++tJ9aaSm5aLR2L3nGBPHJFISYVe/ZuW9ervlLfS3r17ycnJISMjg88//5yVK1c22X722WezcOFCdu501reqqqriq6++4rvvviMzM5OePXuyY8cO3nzz4HIkoanOR4wYwd///nd27dqF3+/npZdeYtSoUYet02mnnca7777Ll19+2fg9jTEdS0ADbNy9kcmvT2bsn8cy+fXJbNy9kYC2vV82yAJHJDLyYdJLB4NHr/7O54zWL406btw4fD4fJ554InfffTennXZak+2DBw9m1qxZnHvuuQwbNoxzzjmHb775hqKiIk4++WS+973vceWVVzJy5MjGY2644QbGjRvHmDFjOOqoo/j1r3/NmDFjKCoqYvjw4UyYMOGwdcrPz2fOnDlccsklFBUVHXb1QGNMcqqqreK25bdRXlMOQHlNObctv42q2tj9IRjXtOoiMg74Hc4KgE+r6q+bbT8T+C0wDJikqgvd8jHAoyG7fs/dvlhEngNGAXvdbdNUdc3h6hGLtOoEAk6fhq/eedLIyAePxd1oWFp1Y+KvvLqcsX8ee0j50kuXUpBVENW52j2tuoh4gceBc4AyYJWILFHV0J7cr4FpwF2hx6rqO0Cxe55cYBPw15BdfhYMMu3G44Gsvu36LY0xJlqp3lQKMgsanzgACjILSPXGbpXAeP7JPALYpKpbVLUeeBlo0laiqltV9VPgcI1vE4E3VTX6KdHGGNPF5KblMvus2RRkOk8XBZkFzD5rNrlpsRslGc9RVUcD20I+lwHfb8V5JgGPNCv7LxGZAbwN3K2qdc0PEpEbgBsA+vfv33yzMcZ0Sh7xMChnEPPPn981R1WJyFHAScDSkOKf4/R5nArkAtPDHIqqzlHVElUtyc9vfSe2McZ0NB7xkJeeR0FWAXnpeTENGhDfwLEd6BfyudAti8YPgUWq2hAsUNVv3EmNdcCzOE1ixhhj2kk8A8cqYJCIDBSRVJwmpyVRnuMK4KXQAvcpBBER4CJgbQzqaowxJkJxCxyq6gNuwWlm+ifwiqquE5GZInIhgIicKiJlwGXAkyKyLni8iAzAeWL5e7NTzxeRz4DPgDxgVryuIZnMmDGDZcuWtbh98eLFMU2Nft5557Fnz56Ync8Y03nEdR5HsojJPI4kN23aNMaPH8/EiRMTXZUWdbafuTGdXUvzOJK6czyZBDTArgO7KK8uZ9eBXTGZvh8ubXpWVhZ33HEHQ4YM4eyzz6aiwkmkOG3aNBYudKau3H333QwePJhhw4Zx11138f7777NkyRJ+9rOfUVxczObNm9m8eTPjxo1j+PDhnHHGGXz++eeN57nppps47bTTOPbYY1mxYgXXXHMNJ554ItOmTWusW+iCUvPmzWPYsGEUFRXxox/9qM3XbYzp4MKlzO1sr7amVfcH/Pp55ed67qvn6tDnhuq5r56rn1d+rv5A69Oqt5Q2HdAXXnhBVVXvv/9+/clPfqKqqlOnTtVXX31Vd+3apccff7wGAgFVVd29e3eT7UFnnXWWfvHFF6qqunLlSh0zZkzjfpdffrkGAgFdvHixZmdn66effqp+v19POeUU/fjjj1X1YIr2tWvX6qBBgxrTtbcl1bqlVTemY6GFtOqWHTcCLeV+mX/+fPLS81p1ztC06QAHDhygT58+eDyexhxRV111FZdcckmT43r27ElaWhrXXnst48ePZ/z48Yecu7q6mvfff5/LLrussayu7uBUlwsuuAAR4aSTTqJv376cdNJJAAwZMoStW7dSXFzcuO/y5cu57LLLyMtzrtNSrRtjLHBEoN5f32T6PjjBo97f+rTq2kLa9AceeKDJZ2fw2EEpKSl8+OGHvP322yxcuJDHHnuM5cuXN9knEAjQq1cv1qwJn8Kre/fuAHg8nsb3wc8+n6/V12SM6RqsjyMCwdwvodqa+6WltOmBQKCxL+PFF1/k9NNPb3JcdXU1e/fu5bzzzuPRRx/lk08+AZqmVO/RowcDBw7k1VdfBZwgFdwvWmeddRavvvoqlZWVjfU0xnRtFjgiEI/cLy2lTc/MzOTDDz9k6NChLF++nBkzZjQ5bt++fYwfP55hw4Zx+umn88gjTjaWSZMm8dBDD3HyySezefNm5s+fzzPPPENRURFDhgzhtddea1U9hwwZwj333MOoUaMoKirizjvvbPU1G2M6BxuOG6F4r6gVlJWVRXV1dczPmwxsOK4xHUu7p1XvbIK5X4wxpquzpqok01mfNowxnUeXDhxdoZkuWdjP2pjOo8sGjrS0NCorK+2G1g5UlcrKStLS0hJdFWNMDHTZPo7CwkLKysoaU3qY+EpLS6OwsDDR1TDGxECXDRzdunVj4MCBia6GMcZ0OF22qcoYY0zrWOAwxhgTFQscxhhjohLXwCEi40Rkg4hsEpG7w2w/U0Q+EhGfiExsts0vImvc15KQ8oEi8g/3nAvcZWmNMca0k7gFDhHxAo8DPwAGA1eIyOBmu30NTANeDHOKA6pa7L4uDCl/EHhUVY8DdgPXxrzyxhiTpOKxqFy04vnEMQLYpKpbVLUeeBmYELqDqm5V1U+BiK5cnBzjZwEL3aLngYtiV2VjjEleAQ2wcfdGJr8+mbF/Hsvk1yezcffGdg8e8QwcRwPbQj6XuWWRShORUhFZKSLB4NAb2KOqwUUjWjyniNzgHl9qczWMMZ1BS4vKVdW273IHyTyP4xhV3S4ixwLLReQzYG+kB6vqHGAOONlx41RHYzqVQECprKmn3ucnNcVL78xUPB458oGmXcRjUbnWiOcTx3agX8jnQrcsIqq63f26BVgBnAxUAr1EJBjwojqnMaZlgYCyYcc+Ln7iPUY++A4XP/EeG3bsIxCwv7uSRTwWlWuNeAaOVcAgdxRUKjAJWHKEYwAQkRwR6e6+zwNGAuvdxdPfAYIjsKYCrVuhyBjTRGVNPdfPK6Vs9wEAynYf4Pp5pVTWtO9fs6Zl8VhUrjXi1lSlqj4RuQVYCniBuaq6TkRmAqWqukRETgUWATnABSJyv6oOAU4EnhSRAE5w+7WqrndPPR14WURmAR8Dz8TrGozpSup9/sagEVS2+wD1Pn+CamSa84iHQTmDmH/+/LgvKnc4ce3jUNU3gDealc0Ieb8Kp7mp+XHvAye1cM4tOCO2jDExlJripTAnvUnwKMxJJzXFm8BameaSYVE5mzlujAGgd2YqT00poTAnHXCCxlNTSuidaXNsTVPJPKrKGNOOPB7hhL7ZLLp5pI2qModlgcMY08jjEfKzuye6GibJWVOVMcaYqFjgMMYYExULHMYYY6JigcMYY0xULHAYY4yJigUOY4wxUbHAYYwxJioWOIwxxkTFJgAa0wHYOhkmmVjgMCbJBdfJCKY8D+aQOqFvtgUPkxDWVGVMkrN1MkyysScOY5JQaNOUX9XWyTBJJa5PHCIyTkQ2iMgmEbk7zPYzReQjEfGJyMSQ8mIR+UBE1onIpyJyeci250TkSxFZ476K43kNxrS35ku4bt5Z05jqPMjWyTCJFFHgEJHe0Z5YRLzA48APgMHAFSIyuNluXwPTgBeble8HprirAY4DfisivUK2/0xVi93XmmjrZkwya940NfvtjTw0cZitk2GSRqRNVStFZA3wLPCmu/b3kYwANrkr9iEiLwMTgOASsKjqVndbIPRAVf0i5H25iOwE8oE9EdbXmA6r+RKuH2/bw2/e2sCCG04DsFFVJuEibao6HpgD/AjYKCK/FJHjj3DM0cC2kM9lbllURGQEkApsDin+L7cJ61ERCbt4gIjcICKlIlJaUVER7bc1JmGCS7iGqqiuIzXFy9E5GeRnd7egYRIqosChjr+p6hXA9cBU4EMR+buI/K94VU5EjgL+BFytqsGnkp8D3wNOBXKB6S3UeY6qlqhqSX5+fryqaEzM2RKuJtlF1FTl9nFchfPEsQO4FVgCFAOvAgPDHLYd6BfyudAti4iI9ABeB+5R1ZXBclX9xn1bJyLPAndFek5jOgJbwtUku0j7OD7A+cv/IlUtCykvFZE/tnDMKmCQiAzECRiTgCsj+WYikgosAuap6sJm245S1W9ERICLgLURXoMxHYYt4WqSWaR9HL9Q1QdCg4aIXAagqg+GO0BVfcAtwFLgn8ArqrpORGaKyIXuOU4VkTLgMuBJEVnnHv5D4ExgWphht/NF5DPgMyAPmBXNBRtjjGkbiWSAlIh8pKqnHKksWZWUlGhpaWmiq2GMMR2KiKxW1ZLm5YdtqhKRHwDnAUeLyOyQTT0AX2yraIwxpiM4Uh9HOVAKXAisDinfB9wRr0oZY4xJXocNHKr6CfCJiLyoqg3tVCdjjDFJLNJRVSNE5D7gGPcYwZnecWy8KmaMMSY5RRo4nsFpmloNWEpOY4zpwiINHHtV9c241sQYY0yHEGngeEdEHgL+L1AXLFTVj+JSK2OMMUkr0sDxffdr6HheBc6KbXWMSV627rcxjogCh6qOiXdFjElmtu63MQcdaQLgVar6gojcGW67qj4Sn2oZk1xaWvd70c0jLaeU6XKO9MSR6X7NjndFjElmzRdXAlv323RdR5oA+KT79f72qY4xySm4uFJo8LB1v01XFel6HGnAtcAQIC1YrqrXxKlexiSV4OJKzfs4bHEl0xVFOqrqT8DnwFhgJjAZJ1W6MV2CLa5kzEGRBo7jVPUyEZmgqs+LyIvA/8SzYsYkG1tcyRhHpIEjmOBwj4gMBb4F+sSnSsa0D5uXYUzrRLoC4BwRyQF+gbPW+Hog7Mp/oURknIhsEJFNInJ3mO1nishHIuITkYnNtk0VkY3ua2pI+XAR+cw952x3CVljohKcl3HxE+8x8sF3uPiJ99iwYx+BwJEXNjOmq4s0cLytqrtV9V1VPVZV+wB/PdwBIuIFHgd+AAwGrhCRwc12+xqYBrzY7Nhc4F6cGesjgHvdwAXwB+B6YJD7GhfhNRjTqKV5GZU19QmumTHJL9LA8ecwZQuPcMwIYJOqblHVeuBlYELoDqq6VVU/BQLNjh0L/E1Vq1R1N/A3YJyIHAX0UNWV6qx5Ow+4KMJrMKaRzcswpvWONHP8ezhDcHuKyCUhm3oQMiy3BUcD20I+l3Ew59WRhDv2aPdVFqb8ECJyA3ADQP/+/SP8tqarsHkZxrTekZ44TgDGA72AC0Jep+A0FyUtVZ2jqiWqWpKfn5/o6pgkE5yXUZiTDmDzMoyJwpFmjr8mIn8BpqvqL6M893agX8jnQrcs0mNHNzt2hVte2MpzGtPI5mUY03pH7ONQVT+t60dYBQwSkYEikgpMwhmRFYmlwLkikuN2ip8LLFXVb4DvROQ0dzTVFOC1VtTNmMZ5GUfnZJCf3d2ChjERinQex3si8hiwAKgJFh5uISdV9YnILThBwAvMVdV1IjITKFXVJSJyKrAIyAEuEJH7VXWIqlaJyAM4wQdgpqpWue9vBp4D0oE33ZcxBALKrpo6ahv8eEVIT/XSK92eIoyJNXEGJx1hJ5F3whSrqnaIhZxKSkq0tLQ00dUwcRRuvYyHJg6jb480BvTOtOARoYAGqKqtot5fT6o3ldy0XDwS6eBL09mIyGpVLWlebgs5mU4h3LyMny38lAcmDCU7rZulColAQANs3L2R25bfRnlNOQWZBcw+azaDcgZZ8DBNRPSvQUT6isgzIvKm+3mwiFwb36oZE7mW5mVkpHptbkaEqmqrGoMGQHlNObctv42q2qojHGm6mkj/jHgOp6+iwP38BXB7PCpkTGsE52WEKsxJZ3+93+ZmRKjeX98YNILKa8qp99tsetNUpJ3jear6ioj8HBo7vu3POJMQ4ZITBudlPLrscyaemkPvbC+5GRmkaKbNzYhQqjeVgsyCJsGjILOAVK/9/ExTkT5x1IhIb0ABROQ0YG/camVMC1pKTggwqE8md57Xk/9eewvTlk3gxrensp8yEEtcGInctFxmnzWbgkynYSHYx5GblpvgmplkE+moqlOA3wNDgbVAPjDRzTOV9GxUVedRsa+Oi59475BUIYtuHom3Ww3rdq0jPSWdvfV7mfvZXHYd2MX88+eTl56XwFp3HDaqyoRq66iqj0RkFE4KEgE2qGrDEQ4zJuZa6gQPBPzsqtnBrJWzGkcE3T/yfn7/0e+tjT4KHvFYkDVHFOmoqp8AWaq6TlXXAlkicnN8q2bMoVrqBA94q7n9ndubjAi69717ubHoRmujNybGIn0GvV5V9wQ/uKnOkzrJoemcWkpOqPjCjgg6pscx1kZvTIxFOqrKKyLiroERXKTJ/owz7a6l5IRVdfVhRwRlpGRYG70xMRbp/6i3gAUicraInA285JYZ0+7CJSdscURQuj1tGBNrkY6q8gA/Bs52i/4GPO1mzk16Nqqqa7ARQcbEVltHVQVw1vr+Q6wrZkys2IggY9pHRIFDREYC9wHHuMcITnbcY+NXNWPC6xJPFoEA7K8AXz2kpEJGPng62TWaDivSzvFngDuA1UCHaJ4ynVOXyOAaCMDO9fDyFbDna+jVHya9BH0GW/AwSSHSf4V7VfVNVd2pqpXB15EOEpFxIrJBRDaJyN1htncXkQXu9n+IyAC3fLKIrAl5BUSk2N22wj1ncFufKK7XdHAJzeAaCED1DtizzfkaCMTn++yvOBg0wPn68hVOuTFJINInjndE5CHg/wJ1wcLDrQDoDtl9HDgHKANWicgSVV0fstu1wG5VPU5EJgEPAper6nxgvnuek4DFqrom5LjJqmq93V1QwjK4tudTgK/+YNAI2vO1U25MEog0cHzf/Trc/So4CQ8PtwLgCGCTqm4BEJGXgQlAaOCYgNN3ArAQeCx0vojrCuDlCOtpOrmEZXBt6SngumWQ1Te23ysl1QlMocGjV3+n3JgkEOmfSivc19/dV/Dz4RwNbAv5XOaWhd1HVX04GXd7N9vncpx5I6GedZup/lNEwq4JKiI3iEipiJRWVNgjfmeRsAyu7fkUkJHvPM306u98Dj7dZOTH/nsZ0wqRPnFUh7xPA8YD/4x9dZoSke8D+938WEGTVXW7iGQDfwZ+BMxrfqyqzgHmgDOPI951NU2FWzMjFut+e8TDoJxBzD9/fvuOqmrPpwCPx2kCu27ZwVFV6b1tlJVJGpHO4/jv0M8i8jDOioCHsx3oF/K50C0Lt0+ZiKQAPYHQTvdJNHvaUNXt7td9IvIiTpPYIYHDJE5wzYzgGuDBfFIn9M2OWfBo9/kawaeA5n0c8XoK8HgONoHZKCuTZFr7ry4DJxAczipgkIgMFJFUnCCwpNk+S4Cp7vuJwPKQfFge4IeE9G+ISIqI5Lnvu+E8+azFJJXKmvrGoAFO2vPr55VSWdOBO3dDnwJuX+t8ba8bt42yMkkm0gmAn+Gu/gd4cRZymnm4Y9zlZW/BeTLxAnNVdZ2IzARKVXUJzvyQP4nIJqAKJ7gEnQlsC3auu7oDS92g4QWWAU9Fcg2m/bS0Zka9r4NPAQp9CmhPNsrKJJlI+zjGh7z3ATvczuzDUtU3gDealc0IeV8LXNbCsSuA05qV1XBwZJdJUsE1M5qv0pea4k1grTowG2VlkkxEz9mq+lXIa3skQcN0XZm6bpUAABi8SURBVC2tmdE70250rWKjrEySiSg7bkdn2XHbX7xGVcVSh8p5ZbmrTAK0KTuuMdEKrpmRrDpczqtE9a8YE0YS/g8xJv4SmvPKmA7OAofpkhKW88qYTsACh+mSgjmvQrVLzitjOgELHKZLSljOK2M6AescN11SwnJeGdMJWOAw8dEBho92pDXKO9TQYdPpWeAwsWdJ+WKqww0dNp2e/aszsWdJ+WLKhg6bZGNPHCb2s7wtKV9M2dBhk2wscHRxcVk7w5LyxVS7LpfbAfqmTOLZv4guLi5rZ1hSvphqt6HDwb6pp/8dfjvU+bpzvVNuTAh74uiigs1T++t9sV87I9zSp/aXa6u129DhlvqmrltmebJME/Y/uQvy+QL889vvuPiJ9/j8232N6c+DYrJ2RjApX69+zlcLGm0SHDpckFVAXnpefEZTWd+UiVBc/zeLyDgR2SAim0Tk7jDbu4vIAnf7P0RkgFs+QEQOiMga9/XHkGOGi8hn7jGzRSS5cnUnsUBAqaqpY/veA/z4T6sp232AP67YzIOXDrO1M8zBvqlQ1jdlwohbU5WIeIHHgXOAMmCViCxR1fUhu10L7FbV40RkEvAgcLm7bbOqFoc59R+A64F/4KwuOA54M06X0WkEO8G/3VtLdlpKY/PUx9v28PDSDfzn+MGc+C/ZpKemJOXaGV1NQib8Bfumms+/sb6p5JIEAxji+d1GAJtUdYuq1gMvAxOa7TMBeN59vxA4+3BPECJyFNBDVVeqswLVPOCi2Fe98wl2gmekeqmsqW/SPPXxtj088Jf1pKemkJ/d3YJGggUn/E1+fTJj/zyWya9PZuPujQQ0zp3UoX1Tt691vtqkzeSSJAMY4vkv4mhgW8jnMrcs7D7ucrR7gd7utoEi8rGI/F1EzgjZv+wI5wRARG4QkVIRKa2osIln9T4/ZbsPsOdAA39eve2Q5qknfzTcmqeSREIn/FnfVHJLksm1yTqq6hugv6pWishwYLGIDInmBKo6B5gDztKxcahjh5Ka4qUwJ50/rtjMXWNP4Pn3v+Q/xw+md2YqfbK7U9Az3Z40koRN+DMtSpIBDPH8c2I70C/kc6FbFnYfEUkBegKVqlqnqpUAqroa2Awc7+5feIRzmjB6Z6by1JQSKqrreHjpBq4YcQwn9M2mf+8MCnMySEmxvyyTha0VYlqUJAMY4nm3WAUMEpGBIpIKTAKWNNtnCTDVfT8RWK6qKiL5buc6InIsMAjYoqrfAN+JyGluX8gU4LU4XkOn4fEIJ/TNZtHNI3nsypMZenRP+udm0Cc7zZ40koytFWJalCSTa+PWVKWqPhG5BVgKeIG5qrpORGYCpaq6BHgG+JOIbAKqcIILwJnATBFpAALAjaoabOC9GXgOSMcZTWUjqiLk8Qj52d0P3ZAEozTMQbZWiGlRkkyuFWdwUudWUlKipaWlia5GcgkGi0AAaipgwWRLgW6MaUJEVqtqSfNyuzN0RX4f7FjrDOXbXnowaIClQDfGHJEFjq4mEIC92w4Gi/ScpBilYYzpOJJ1OK45jEBA+a62npo6P76A0s3roU9W98hGRu2vgOodB4PFgd2WAt0YExV74uhgAgFl+579fF11gMvnrGTUQyv44ZMf8PmOffh8R5g9GghAwwGnTyM4KuO938KFjyV8lIYxpuOwwNHBVNbUU+dTbp7/UZM1NG58YTU7q+taPjCYqmDXF7DmpYPBoqwU/vEkTFliaSaMMRGxpqoOJBBQDjT48AjkZ3Xjd+ML6JMh7NyvzFpRgc9/mCeOYKqCrD5w1gwnWIz9JWTmO6klevYDr/1zMMYcmd0pOojQ7LbH9clk3vgsshddBnu+pl+v/sy7eB613Q7zlBBMVbDna1g+E0be7nSM9zjaeXXFJ4wo568kJGOtMUnI/tV3EMHstrPf3ki+7CV70ZQmQ2izF00hT/a2fILQVAVlpbDgKlh8k1PeVYNGFFlGE5ax1pgk1AXvGB1TMLvtx9v24PHXhh1CK77alk+QJKkKkkaUWUYTmrHWmCRjTVUdRDC7bdnuAyje8ENo5TDLvSZJqoKkEWWWUctYa8xBXfSukbwCAWXnvlq+rqph++79VNXUEQhoY3bbwpx0vtwbgAlPNH16mPAEdEs//MltrYWDoswyahlrjTnIclUlkUBA+aqymkD1TgqzBPF4afCksUezOKpXBuD0dXjxk1NbhuzZCt0yoGE/5BwLucd27WAQjWAfR/NlUlsYihzs4wg2VwUz1g7KGWQd5KbTailXlQWOJBAIKJU19WjAT/qeDQc7vt0niYbMvlRn9icnM+3gDe+dX0LxFTacti1sVJUxh9VS4LA7TYIFnzKqq75lSJ/ueJqNluK1m+l2/n+TkZoNgb6wr/zgX8kbXnf269Xf6bvI6pu4C+mIgk13ke4uHvLS8+JYIWM6BgscCba/ro5jvLvwpFdCIM+ZoBfaabvna+iWQYo2OE8a9dWWlNAYk1Bxfc4WkXEiskFENonI3WG2dxeRBe72f4jIALf8HBFZLSKfuV/PCjlmhXvONe6rTzyvIV58DT58+yrIrN2Jx9/g5JD66ww4+z4oDHky7NUfGvbj8XicJ43QPFOh+1hSQmNMO4lb4HCXfn0c+AEwGLhCRAY32+1aYLeqHgc8Cjzolu8CLlDVk3CWlv1Ts+Mmq2qx+9oZr2uIF199Pd59ZaR8tw15/nx4/FT4/26D//UTWPkHGDXd2dHt49CcYxHxOk8WlpTQGJNg8WyqGgFsUtUtACLyMjABWB+yzwTgPvf9QuAxERFV/Thkn3VAuoh0V9XDZPFLfj5fgOraOnoG9iABH7w6tWl/xuIbnfxROQPR29YgnhTolo6k5zqduMGkhMtnHswz1bMQsgtsNJUx5qA4Lwcdz7vN0cC2kM9lblnYfVTVB+wFejfb51Lgo2ZB41m3meo/RUTCfXMRuUFESkWktKIi8avZ1dX58Ptq6emrcGZ4ezzh+yoy81FPCtLrGGe+RWaes2/ozO+yUlj6fyA1y4KGMaapKNPptEZSd46LyBCc5qtzQ4onq+p2EckG/gz8CJjX/FhVnQPMAWc4bjtUt0W+2lpSfd9BfQ1SvcPpp0jPCTv7W7P6oj36Ic2Dgc38NsZEoqV0OjEceRnPu852oF/I50K3LOw+IpIC9AQq3c+FwCJgiqpuDh6gqtvdr/uAF3GaxJKWr7YW73dbkb1lyLwLYe5Y52khJR0untOkr0Ivn4+/RyGelBbiuc38NsYcSZTpdFojnneeVcAgERkoIqnAJGBJs32W4HR+A0wElquqikgv4HXgblV9L7iziKSISJ77vhswHlgbx2toE19tLd7aCkQ8sH+XM9QWnF/iwmkQaICpf4FrlsKPFkH2v5CS0i2hdTbGdHBRptNpjbgFDrfP4hZgKfBP4BVVXSciM0XkQne3Z4DeIrIJuBMIDtm9BTgOmNFs2G13YKmIfAqswXlieSpe19BaPp+PQG0N3t1fIM+d54yaev0/nAWUgkNt93wN3lQQcZqu/joDofPP4jfGxFk7ZMK2lCMx5qutxUsdUvsdPHfeoRlsx/7SWQujV39nuVZ/A8y/9LB5kowxJioxGlVlKUfaga+2Fm9dpfPk4G+hnTHYKX75fLRbBpIScDqtrKPbGBMrUabTiZYFjhjx1dbi3fe1k7E291io2hJ+zYye/dBpb0D3bCQ1yxITGmM6HPsTNwYaO8G9KdBrAPjq4O8PHjrD+4fzUF8d/rR8JL2XBQ1jTIdkd642coLGTsRXDwEffDwfhk2EzL4HZ3in50DDfjTrX/B3zyGle1qiqx13loLcmM7LAkcrBfx+pKEW7+5NyIKrDq6fcdk8+HQhjJ0F8yY0doTr5S+iGXkJHW7b0s081jd5W/TImM7NRlW1gq+hHq+/Adm/E+ZdeGg/xpWvgrcbVG2G3oPQlDQkMz+hTVOhN/O89DxuLLqR/j36k90tm4oDFTG9ye86sIvJr09uskZ3QWYB88+fH5v1LOKch8cY42hpVJX9b4uSr6Eeb0MNUlsJ1TvCj5zypEDAD1n/gmb0QbL6Jrw/o6q2qjFo3HrKrby64VU27dnEvoZ9jUEDoLymnNuW30ZVbVWrv1e9v75J0Aiet94fg5mr7ZCHxxhzeBY4ouCrq8VbtwfZ607fP9zaGGk90Jx/RVLTY/bXcEAD7Dqwi501O/m25lvKq8vZdWAXAT3yTTN4M7/mpGt4cf2LXDn4Sn7z4W+o2F8R85t8qjeVgsyCJmUFmQWkemMwc7WlPDz7E5/I0piuwgJHhHy1tXhrvkUaDjj9FgEfrHnp0JFTl7+ApvVEu2XHPGhs3L2Rme/PZMt3W5j65lTG/nksk1+fzMbdG48YPII3856pPZkwaAL3vncv5TXl7K3fG/ObfG5aLrPPmt143mDzV25abqvP2agd8vAYYw7P+jgi0DjcNtAAkgLvuutNlVwH7z4ExVc46dCz+kJGHn66kZIW25FTwX6D/z3if/ObD38Tdf9BMPDs3L+TrG5ZTHlrCgDD8oZx6ym3NgaSWHVkx21UVfUOp3mqeb+SrbluTMzZzPFWahxuG/A5fRcVG+HU62DV01D6NJw9AzwpaEoq/rQ+pKSlxeWHGmxq6pnas1VNSx7xMChnEPkZ+dTU11CQWUB5TTmf7vqU33/0e35x2i8Y2HMgaSlpMbnJe8QTm47w5oJ5eILNVaF5eKzT3Jh2YYGjBU7OKT/eqo3IKz86eJP64Z+gphL+7adO8sIv30UvfwF/VmHMnzJCBZuagk1LzZ84Imla8oiH3LRceqT24HdjfsdP3/kp5TVOP0mfjD4UZBUk/3DZltYlAaeTPDSgXD7f2dcmWhoTU9ZUFUYwaEhtJTx3/qHNItNed/o4qneiPQrwZ/QhJTW+k/qCTU2Pf/w4Vw6+khfXv8iEQRPI7Z5LXkYeR2UeRYon8htkp5ug11IT1pQl0OsYe/IwphWsqSoKXv8+5Lty6JYZviM24Ae/D+2Wjj+9dUEj2ht3sKlpxr/NQFS4ffjtbK/ejl/9bN27FV/AxzE9jon45h+3pqREaanTvHoHpGZY/4cxMWR/hoUTOnIq3HBbjxfNyMGfc0JU6UOCw2nLq8vZXr2dme/PjGpkVPBmLx6h8kAls1bO4uqlVzNr5Sx27d/Fnro9rbnazqGlxWtqKmzElTExZk1VYWjVl8jsYjj5Kmfk1KtTmvRxaO4g/Hib9GkENEDVgSpq/bV4xEM3TzdUlVRvKgd8Bxr3q/PXAU5nd25aLnWBOgKBAB7xkOpNJScthz21exrPk5aSRq/uvZo8SZRXl3P1W1cf0s/x7LhnKchqOrS2ywgEYMdaWDD54O/qwsfgH0/CBY/aE4cxrZCQpioRGQf8DvACT6vqr5tt7w7MA4bjrDV+uapudbf9HLgW8AO3qerSSM4ZE54Uaqd/zd5ALV7x0nDdX/EFfHg9XrJSsvBpHXX+OlL2p4BCgzbgD/hJ8aSQ4c3Apz5Q5wmhxldDQ6CBFEnhNx/+hnfK3qEgs4BHRj3CzgM7ueOdOxqHwT545oN8V/8dv1v9u8b9Zo2cRV5GXpNmqIAGwo6simQiYKcV7DSfssRpnqqpcILGmP8T05XPjDFxbKoSES/wOPADYDBwhYgMbrbbtcBuVT0OeBR40D12MM4a5UOAccATIuKN8JxtVpeWz9b93/DCuhfYsX8H096axnmLzuPqt66mrKYMVeWTHZ/wXf13bP1uK1e/dTXnLTqPaW9N49v931Lnr2Nfwz6+3Psl17x1DRcsuoBrl17LlCFTGJY3jPKacqrqqhqDBjg3/unvTqe8upwJgyY0lv3ivV9Qtq+sSQqQNG9a2El7ad7On3X3sLwpTkd47kA4qth50rBVFY2JuXj+jxoBbFLVLapaD7wMTGi2zwTgeff9QuBsERG3/GVVrVPVL4FN7vkiOWeb7fbt5qfv/JSLjr+IO1fc2eTmfsc7d1AXqGNY32GUVzs39ibbV9xBQ6AB4JBt9/z/93DNSdcAkJ6SHvapIT0lnZ6pPQ8pC52nkZvewszs9BjMzO7ogiuf9ernfLWgYUzMxbOp6mhgW8jnMuD7Le2jqj4R2Qv0dstXNjv2aPf9kc4JgIjcANwA0L9//3C7tMgX8FFeU45XvGFv7v6AH2j55h9sUgq3LRgUDvgOhJ2PccB3gPpA/SFlofM0giOs5p8/v/MMpzXGdBid9k6jqnNUtURVS/Lzo2vjTvGkUJBZgF/9YZuEvB4vXo+38ebffHtAAwQ0EHZbcAJfbvdcHh39aJOnhgfPfJCjMo/itY2vNZbNGjmLwuzCQ/I8BUdYFWQVkJeeZ0HDGNNu4nm32Q70C/lc6JaF3UdEUoCeOJ3kLR0byTnbLCcth9+N+R2Lv1jMI6MfaXJzf3TMo3T3dOfTHZ9SkOXc2JtsH/0o3TzOYk3Ntz0y+hFOyDmBuePmktEtg+zUbJ4b9xxvXPIGc8fOpd5fz2sbX2P6iOm8ecmbPDvuWf4151+jmp9hjDHxFrfhuG4g+AI4G+fmvgq4UlXXhezzE+AkVb1RRCYBl6jqD0VkCPAiTp9GAfA2MAiQI50znNYkOaz11bK3bq8zqirQ0GxUlc8ZVeUJGVWlflIkhe6e7vjwISog0BBoIKABvOIFQFHSvGnU+etI86ahovRI7cGeuj3W7GSMSSrtPhzX7bO4BViKM3R2rqquE5GZQKmqLgGeAf4kIpuAKpyRVLj7vQKsB3zAT1TV717IIeeMR/3TUtJIS2m/UUqdaha3MaZTswmAxhhjwrKlY40xxsSEBQ5jjDFRscBhjDEmKhY4jDHGRKVLdI6LSAXwVQS75gG74lydROrM12fX1nF15uvr6Nd2jKoeMoO6SwSOSIlIabgRBJ1FZ74+u7aOqzNfX2e9NmuqMsYYExULHMYYY6JigaOpOYmuQJx15uuza+u4OvP1dcprsz4OY4wxUbEnDmOMMVGxwGGMMSYqXSZwiMg4EdkgIptE5O4w27uLyAJ3+z9EZEDItp+75RtEZGx71jsSrb02ERkgIgdEZI37+mN71z0SEVzfmSLykYj4RGRis21TRWSj+5rafrWOTBuvzR/yu1vSfrWOTATXdqeIrBeRT0XkbRE5JmRbR/+9He7akvr3FhFV7fQvnBTsm4FjgVTgE2Bws31uBv7ovp8ELHDfD3b37w4MdM/jTfQ1xejaBgBrE30NMbi+AcAwYB4wMaQ8F9jifs1x3+ck+ppicW3utupEX0Mbr20MkOG+vynk32Vn+L2FvbZk/71F+uoqTxwjgE2qukVV64GXgQnN9pkAPO++XwicLSLilr+sqnWq+iWwyT1fsmjLtXUER7w+Vd2qqp8CgWbHjgX+pqpVqrob+Bswrj0qHaG2XFuyi+Ta3lHV/e7HlTgrekLn+L21dG2dQlcJHEcD20I+l7llYfdRVR+wF+gd4bGJ1JZrAxgoIh+LyN9F5Ix4V7YV2vLz7wy/u8NJE5FSEVkpIhfFtmptFu21XQu82cpj21tbrg2S+/cWkbitAGg6hG+A/qpaKSLDgcUiMkRVv0t0xUxEjlHV7SJyLLBcRD5T1c2JrlS0ROQqoAQYlei6xFoL19bhf29d5YljO9Av5HOhWxZ2H3e99J5AZYTHJlKrr81tfqsEUNXVOO22x8e9xtFpy8+/M/zuWqSq292vW4AVwMmxrFwbRXRtIvLvwD3AhapaF82xCdSWa0v231tkEt3J0h4vnCerLTid28HOrCHN9vkJTTuQX3HfD6Fp5/gWkqtzvC3Xlh+8FpyOvu1AbqKvKdrrC9n3OQ7tHP8Sp4M1x32fNNfXxmvLAbq77/OAjTTroE32a8O5YW4GBjUr7/C/t8NcW1L/3iL+GSS6Au34yz4P+ML9Zd7jls3E+WsAIA14Fafz+0Pg2JBj73GP2wD8INHXEqtrAy4F1gFrgI+ACxJ9La28vlNx2plrcJ4S14Uce4173ZuAqxN9LbG6NuDfgM/cm9ZnwLWJvpZWXNsyYIf7728NsKQT/d7CXltH+L1F8rKUI8YYY6LSVfo4jDHGxIgFDmOMMVGxwGGMMSYqFjiMMcZExQKHMcaYqFjgMKYdich9InKXiMx0J4ghImeIyDo3W2q6iDzkfn4o0fU1JhxLOWJMAqjqjJCPk4FfqeoLACJyA86EN39CKmfMEdg8DmPiTETuAaYCO3GS460GhgJ/AXoBv8FJPPk+kA2cjzM57FequiARdTbmcOyJw5g4cpNHTgKKcf6/fYQTOABQ1adF5HTgL6q60D2mWlWLE1FfYyJhgcOY+DoDWKTu2gwddsU3Y0JY57gxxpioWOAwJr7eBS5yR0tlAxckukLGtJU1VRkTR6r6kYgswMmGuhNYleAqGdNmNqrKGGNMVKypyhhjTFQscBhjjImKBQ5jjDFRscBhjDEmKhY4jDHGRMUChzHGmKhY4DDGGBOV/weOmSzfAD/YRAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "dist = torch.abs(y - score_average)\n",
    "sns.scatterplot(x=dist, y=uncertainties)\n",
    "sns.scatterplot(x=dist, y=A)\n",
    "sns.scatterplot(x=dist, y=E)\n",
    "plt.xlabel(\"diff\")\n",
    "plt.ylabel(\"uncertainty\")\n",
    "plt.legend(labels=['uncertainty', 'aleatoric', 'epistemic'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Spread pre and post sigmoid analysis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "T, B = pre_sigmoids.shape\n",
    "pre_sigmoid_df = pd.DataFrame(columns=[\"entry\",\"value\"])\n",
    "post_sigmoid_df = pd.DataFrame(columns=[\"entry\",\"value\"])\n",
    "for i in range(T):\n",
    "    for j in range(20):\n",
    "        pre_sigmoid_df.loc[i*B + j] = (j+1, pre_sigmoids[i,j])\n",
    "        post_sigmoid_df.loc[i*B + j] = (j+1, scores[i,j])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAwgAAAFgCAYAAAAfPV4JAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAgAElEQVR4nOzdd5xc5X3v8e9zpmzV9qZt2lUvoMYKJED0ZsCAIbFxwXFJXBPbsRPX5CZOHPsmcXztxL5OfO0krtjgGowBm94RQgUhod5X2tVqe5+Zc577x4iRRruSVlhzzkj7eb9evF6c58zu+WnL7PmepxlrrQAAAABAkpygCwAAAACQPQgIAAAAAFIICAAAAABSCAgAAAAAUggIAAAAAFLCQRdwOm644Qb74IMPBl0GAAAAcLYzJzpxVvUgHD58OOgSAAAAgHPaWRUQAAAAAGQWAQEAAABACgEBAAAAQAoBAQAAAEAKAQEAAABACgEBAAAAQAoBAQAAAEAKAQEAAABACgEBAAAAQAoBAZiARHxA8dHeoMsAAADIuHDQBQAT5cYHFYoU+H7dHa/8u/Zu+aE8L67qhms0e8kn1NW+SpGcEpVVLZMx5GwAyAbDMau2Lk9VJY4Kco2stdrd7ql/2GpmbUj5OSboEpHlvK5R2YMjchrzZaZEgi4nxSZcmXDIt+tNyoBgY3FJkolmzzceJzbat1f7nvprjXRtUU7RNNVf+nnllc/15dq9na9o96v/lTpu3/c7HT74jNzEkCSpYuqlWnTpv/hSCwD/be0b0N7BIdXl5WlWUYHCTvY8EBhxPX17W4fWdA5qdlGu3j+7SqU5k/LPuiRpZ5urnz0dU8KVjJGuvyCs3e1Wm/e5kqTcSFzvuDpHlcXZ8z1Edkm82KnEbw4kD0JGoSurpGFPTkWOnPNLZEL+B0zvYK/i/7NB9vCATF2JIrctlFOSn/HrTqp3Emut3Eefl7tmo+RZKRqRs2iuwpcvkwn5l8pweg688E8a6doiSRrt26P9z3xes26525drD/btHNP2WjiQpMMHn1Zv5wYVl5/vSz1AJm3t79Q3tq9W63C/rqicpvfPWKqIM3nfG7+xZZd+vKc1dVwcCetLS+br/JKiAKtKcq3Vx1bt0abeEUnS9v5RtQ3H9dULp/lex4YuV3khozklwf6sPLwmGQ4kyVrpodUJ2WPOj8SlVVsSuunCaCD1DQ142rQqob5uT9UNIc1dGlYoTI9GtrCuVeLRtqMNrpX7cHvyfyU5O/oVvaPR97ri//Oy7OHBZI2tPUo8uEnRO1syft1JFRC8HXvlrn7laEMsLu/FDXKjEYUvvSC4wnBSw0fCwWtGe3fJc2NyQpl/ky+tWiZjwrI2ccLXJI4JDMDZKuF5+sv1j+hwLPnzfPe+jcoPR/Se5sUBVxaMztGY7t3bmtbWG0/oy5u267sXLw2oqqO+v+NwKhy8Zk3XkIYSnvLD/jwhH4hbfey5Qe3o8yRJDQWO+uOe8sNG756Tq2vq/O2l7zvurdiO85r4id/KM27VwzH1diar6u9JyFrpvIsYyZA1PCuNeic+/Uqv7PVxmUL/vmc2lkiFg1QdB3tlYwm5q/fKOzyg0KwqhebVnPFrT6p+Nnuoc9x2b+d+nyvB6SisSU/K+ZULfQkHkpRXMFWLLv0XFZcvVEHxTDXMulPGHH1Klj+lSaWVhEuc/XYOdqfCwWtWdR0IqJrgDSZcuePcYe4dHPa/mHE81tY/pq08J6RcH4dA3L83lgoHkrRv0FNPTDowZPXFtcN6YO+oPDvebXpm1JSN/bcXHzMSwxhpyYxgejmGB20qHLymba8bSC0Yn4k4cs4vOckLJDn+9viYaFhmanqPpdNYpvjP1inx2FZ5Gw4o/vN1Sry094xfe1IFBKepftx2U1XmcyU4HbUXfUrF065RKLdUU+ouUf3Kv/P1+uU1y1Ves1zDA/u1f/u9Kq9ZofqZb9aM8z6oliu/JceZVB1xOEfV5RUpL5T+szyzsDSgaoLXWJCnRaVjhxJlw/AiSZqaN/YpZnNhjg4MxX2roXv0xDf/VtI/vTyqv17tX6C6bUVUucd8WQpypFtWRHTVorCWzQ7pnVfnaFp1MAEhJ1eK5qa3TSlheFG2idxSp/D1U+UsKlFoRUVaIAgtK5fJ9//vfeS2RXKay6XciJy51QpfOlPezsNpr3HXnfkH3ZMrINRWKXzT5dKUguSjBEmmvkbhlZkfy4XXL5xboobLvqB5f/iAqpd8SO5on6/X7+vapJ0bvyXPHZG1rg4ffFqeO6KaaW9QJKfY11qATGkbGdCfzmxRSSRHkjS9oERvbTgv4KqC9Y9L5utd0xuUGzr6p3Jtd69+e+BQgFUlvX92lSqPm5C8unNIf7pqt0bcEw+TOJOuqovoVB0Wz7YntLnHnyflhXmOPnhzrpqrk9+vwVHpR4/FVVXi6JolUdWWB3fL44SMFl8a1ZFfLxWWGC24kOFF2caEHIWXVyh6W4Mi101V9MOzFH5DrSLvbFbkDbXB1FSaL6ehVKYwKo0mZIdjUij9Z9nkn/mfJWN97P77fbW0tNjVq1efkc9lXU8ajcnk5576xQic9RLa+8Rn1b//SUlSQfUFmnb1V+SEcjJ+7f07fq4ta/5xTLsxYS285J9UMfWSjNcAZEpffFR/vu632tzfKUdGMwpKtG2wW5JUEsnRN5a+QU0FJ+l2P8cdGBrRW54e+3fnqpoK/dV5sxUJcFWjhGf12TX79PxxY5T/cWm9VlRN8aWGb2wc0U93xU76mq8sz9eSCn+evA6PWv3rr0bkHXNr01jl6O1XZv5vxckkElbWk5yQNDJklV9oZAw9CGcLr2tUiUfbZbtGFZpbrNCllTI+DTdKvLhHid++erQhP6LQ0ka5T+9IHkdDir61RU796+rxPeE/YvKOjYjH5T67Rl7bYTnTahVascTX9WUxMdZzNdq3VyPd21PhQJIG219Sz86HVDbrlozX0LbnofFrswnt3Pj/CAg4q/1430Zt7k/Oz/JkU+FAknrio/r+ng366/krgyovcCHHyGjshNdH2w5rSWmxbmuYGkRZkqSwYzRjSu6YgFCZ69+T6VfH6R1wJL3Wh1GTZ3R+mX8hyvWUFg4kKZ4I9kHolnVxbVuXkOtJ9dNDWnJZhHBwlrDWyg4nFP/hbtmuZBBOHByRHCl8aZUvNbibDqY3DMUVai5XaP5U2c4BOdPKZPLO/LzMSRsQEvc9Jm/nPkmS29ouOzKqyLXc6GWTke7t2vPYXyg+2CbjjP3hjw+1Z7yG3s4N6u1cd8LzbnzwhOeAbDaYiGlV1wFt7jt80tf1xUd9qig7lUUjWlZeolWdPWPO7RwIfgWzNzeV6ZmOfu0eSN683NZQoplF/vWMV+UZbTyaKRVxpL9cmKuvbBjRiCu1DVv9zUsj+odlmV+3XZIK84zmNoRSex9I0sLmsB56KaYDnZ4aKh1ddl5E0Yg/N+g9hz1tfuno0kn7d7gqr3HUNHfS3n6dNbx9g4r/fJ9sz9h5Pe6WPt8Cgu0ce58Rf3G3TMxVaGFdRsKBNEkDgo0nUuHgNd6WXRIBIascXP01xQeTaxJbL70L2zhhFTdelfEaXPfkN0d1M27PeA3AmbZrsEcfXvOAese5+Q/JyD3mefnNtbP8LC2rHBoZ1Z+++LIODie/Tsf3JFxUEfwk7tKcsP77kuna2DOskmhYDQX+rvH/rtk52tDl6vCIlWOk987J0bY+TyPHdCw8257Qxu6EFpT6c8txy0URNVc76uyzmlnr6MVtCW1rTfZptHW7Gh6V3rjcn69TX9fY+SC947Qhu1hrFf/F/nHDgSSZcn+GrNm4Kw2PrcFuPiQrydvZKYVDCs2pPuPXnpQBwWttkxxH8o7+kppif8ZrYuJi/fvGtE1puELGGJXP/UPlls7IeA2llUtUWDxLA73bkg0mrMZZd8pNDKqs+iJV1V+Z8RqAM+0HezaMCQfJFfyMrquerqm5hWofHdTVVc26qLwumCKzwE/2tKbCgZQMBzOnFMizVrfVT9UlldmxAp5jjM4v9ecJ/fEaC0P66yV5+tGOUZVEHV1SHdaPdoydkzAQ92+YTyhktHhG8vbGWqu7n0i/Id/a6t/yohW1jowj2WNKqKx11HPYUzgiFbKrc3aKebLdJ5hbUxxW5Mozf0M+HhMJydQWyx7oPeFr3M1tBIQzwQ6PKPHz36WFA0UjCl+9IriiMK4p9Zera8tPUsc5xdMVH2iVO9qnwbY1yq9akvFxnMaEdMGV/64Du+5TbLRbNY3XqbB4ZkavCWRa/zg9B1bJXXEfaNuhry6+TsvKglmxI5t0jY59cveh2U1aVFqs7ph/y4kerz/u6sXDg5qaH9G84rzA6pCk7b2uPvH8kJLD/F09cTCut82Ips1DqMt3tNSnScqjR4JIzpEhRJ5N7oXQc8wojbIp/o3/zy90dNE1UW1eG1ciLjXODGnbywn1dCTrbJwV0pLLgtnZGSfmHTjx8rwmJyxT7N/3LHLbIiUe2CjvYJ9UGJWOm3N0/IpGZ8qkCAjW8+StfVXegXYpGh2zlaJpqpdT689YMkzMoQ3/pZ4d90lOWOGcYhVUL1Xf3sdkveT37tDL31akcKpKZ9yU8VrCkUI1zn5rxq8D+OWNtbP1TOeJ183e3HeYgCDp+qmVeritI63tkYMd+vzLW9QbT2h6Yb6+tHi+an1cDW9734g+smqPBhLJ2+/bG0v1sflnfhfVibp3V0zHzgEecaX/3BrTssqQ6vIdFUeNbm2KKpLhFV+stfrdmrjW7nAlIy2dEVI4JK3e5kpWCoekhCsV5krXLfX3hry6IaTqhuQiKNvWx1PhQJL2bnPVONtVeQ2LpGQT99mOk5z1edK7kcyUXDmeTU6aVnpAcDK0YtmkCAiJh5+Vt/bVE553aip8rAanMtD2kg6t+4/UcWK4U5GCqalw8JrBtpd8CQjAuWZlZaO+uvg6PdK+SzHr6qG2nWnne2IjAVWWXZoLxw7befDgodQOyzsHhvTNbbv094vm+VbTD3Z2psKBJP1ib7fe2lyu6nE2Tsu0rb2uHm4dvyflxQ5XH70yT3UF/gyh2X7A00vbjwwdskeCwXGqS4wK84xiPq5qFI9Zte505R1ZwWh4cOy1x2tDwE4yTSR8caVvZVjPKv6j1bLdJ1gQwRg5szLzgHtSDH7zNmxNbzBGyokmv7BzmhVqmdybAWWb4c5NY9qSqwWlP4HKK5vjU0XAuWdZWa0+Pe8SXVXVPObcva2valt/VwBVZZeCcFjR4558u8fdy+3yeSWj/nj6ja+V9O9bDunrm9vVOnTy/QjOtAf2xsYsKXqs0eO/WBl0qPfUE3/be6x2HPR0z5MxHerJ/ETheMzqiV+Oav0zcW14Lq7HfjGiyrr0noJIVKqqp/cg24SWl4/bHr61XqFF/i1OYNv7xoQDU5YvU5ovUzVFkdsXy8nQ/KNJERCUe1z3r7VSWbGiH32nIrddIxOZFB0pZ42CqiVj2np3/1Zls9+kULRIMiGVTL9RpbNZQQj4fTXmF41pc63VT/efuNd1siiMhPXuGY2pRxOF4ZBq89L/nqyo8Hei8rW16d8vI+mRtj7ds7tLH3hut3piiXE/LhPyT7JU6OLykKYX+Xfj21w98Wt5VtqyP/MTlQ/scjXYfzQkjQxJg32eLro2quJyo2iuVFTmaKCXHoRsE5pVpPCbG6Xw0Z9x5/xihRf7u3KZmZIrHfeQwmkoU86HLlPOn1yi0NzMTZaeFHfG4SsvVOK+x9IbD3ZII6PJngRkFTfWJ+NE05Y29WL96t3zqGbf/ksZ48gJswM2cCY05hfrttrZ+uWB9J7WhGUpRkl6R3ODLq+q0N6hYS0uLVJvLKH/u3WXdg8OaUVFmf5k1jRf6hhxPX3h5QN6qr1fjpId4UWRkLpjR290e+Ounmjr162N/tzE3DYtqt/tj6tjJHmDO7fY0aLykGryQ7qhwd8hT7Xljm5ZHtHDa+KaSEdKcUEwG5VZm5wv0duZ/Jp1tnl69oFRXfPmXOXmsXlatrDWyu4fSm7skWsUXlau8GX+z1U1hTkKXzlbice2Sp6VKctXeOXRFRztcFy2e0imeorMGZ6sPCkCQmj+TLkvb5Hdc+BoYyQs5Qa79TrGd2DVv4zZ90CS3NEeuaO9ihYGt3MpcC76izkrtGWgS68e2TQt6ji6o35uwFVlj4aCPDUUJFcLKgiH9YXF/s05eM1P93TpyfZ+SUemSFqlhYPXFEX9e2pfmefou1cU6oVDCRVGjJZWhOQEuEPwgmlh5eUY/eSJkyeEGVMdLWjM/NeptjmkresSGho4srJSnlRW7ejZB9LrcxNSR6urhpmT4pbsrOCu6Zb77NFNJBOPt8vbMyinuUCh5RUyYf8G4ISXNyt0Xq1s/4hMTVFq9Ub3lQOK3/+KlPCkwhxF33KBnJqxPcKv+7pn7DNlufBVyxW/5wFpcFhyjELLF8n2D8rQg5BVrPVSm6MdL6e4mXAAZMgf1M/Tgwe3qySaq3c3LdK0gpKgS8IxdvSNv2ljyBydF3F+SZ4uzdCKJieSFza6ojaiA4OeOkasqgN+Cj69JqQbl0X00raEBkashkalsCMtnxfW/MaQrJXKi/y5uYtEjS6/LUf7d7iynlQ/I6RXX4rLG2d0U4FPNWFivD3HLSVqJW/ngLydA7JdMUVuqfe1HlOYI1OYfKhtrVX80S3ynt999AUDo0o8ukXRty07Y9cMPCAYY0KSVktqtdbenKnrOFXlin7grbJtHXJf2Sb36TVyn3pJpq5akT+8gaCQJYxxVNR4hfr2PJJqcyIFKqheqpLmG9S7+2EV1l6kUJSN7YAz5Zs7XtIP976SOp5RUKq7CAhZpaWiQI+09Y1pv6iiULc2lirqGC0py/f9CX7cs/rbl4b1bHty7sN19RF9alFuoD0Ji6aHtWh68vYmnrAyRgqHgqknmmM0ff7RW63Y6Nj5BrXNjsqqCAjZxKnLk7ehZ9xz7ss9vgeEY3kbDqSHgyNsz4n3bng9suEn8qOSfJkNZ8IhKRyWt35zciCgJNvaLveljX5cHhNUt+Jzqpj/duVXLVHlee/S3D+4X+G8Cu176nPa99Rfaesv7tBI9/agywTOCZ61+tn+zWlt9zJBOevcWFesP55VqYqcoysrLSjO08fm12hFZaEuKC8I5Kb8kdZ4KhxI0m/3x7XqkH8TpU8lEjaBhYPxNM5Kfy5bUCS1XMkDymwTailXaGlpsovuuB8fMyXYZ+ve3u5x2515Z3Y/lED/lcaYekk3SfoHSR/345q2Z+wTmPHaEJxQJF81F/xZ6jjW36rubb9IHbuxPnW88j01rPy7IMqT5yU02LdLeQW1CkcKAqkBOFOMpIjjaOTYzeUdll3MNsYYvXNGhd45o0LWWg27nvLDwX+fDgyNnczeOk4bkqrqHVXWOepo9SST3ETNBNjbgvGZkFHkjfUK31Qnb3u/4vfulRJWChuFrw92qLOpK5bWp290GbqoSeHLZ57R6wQ9xOirkj4p6YTjRYwx75P0PklqbGz8vS/oNNVJ0YgUO2ZzF5c3s2zmxgfGtA0fHrtXgh8Gendo/dMf18hQm0KhPM1t+axqGq8LpBbgTDDG6F1Ni/Rv219MHkt6d9OiYIvCSRljsiIcSNIl1RH9cFssta9UxJFWVPm/advZYv8ONxkOJMlKOze6qm12VX4ay7TCP8YxCs0ukvPnc+UdHJYzNU8mP9hb59CietmOAblr90k5YYWvmK3w4jM/5Cmwf6Ux5mZJh6y1LxljrjjR66y135L0LUlqaWn5vRcLNrk5chbMkrf26A2mt2m7vBWL5VT4u74tJia3dLZCOcVyR3tTbbGBVsUG2xUtyNwawOPZ/vK/aWQoOYnadYe1Ze0/q6ruCjkhuohx9rqzcYEWlVRrU1+HFpfUaEYh74WYmDklIX1hWZ5+viumkGN05/Soan3aPfls1Nc19jZm96sJAkKWM/lhhWZkx9xH4xhFrpun8LVzM9r7FGQMukTSLcaYGyXlSioyxvzAWvuOjF85NnYJNNvVIxEQspIxRrmlszTYtvqYVqvEcKevAcHzEurv2ZbWloj1KR7vV05o/F0XgbPFvKIKzSuqCLoMnIVWVEe0oppeg4moqne045X0tv07PJVUJjRjQdCDOnA2yfTQtMBivrX2M9baemttk6Q7JT3qSziQ5Mw4bqhSNCKnsdaPS+N1Kp1xU9pxTtE05ZX7s067tZ4621dp1e/uUmzkcNq54vKFysklHAAATq2qLqR5F4wNAvt3ZM/EbkAKfg6CL+xoTO7TL8k72CGncapCK5YofN0lcjdslcnLVeiSpTJsmpbVSqa/QcYJq2f3w4oW1KhiwV0yJvP51vMSWvvkn6mnY82YcwVFM3T+xf874zUAmdAXH9V3dq3T5r7DWlJao3c3LVJOaFL8SQACVTc9pM1rEzp2s/LcfCYqI7tkxV8Da+3jkh7P1OdP3P+4vG17JElua7vs0IgiN6xUaMn8TF0SGVDcdK2Km6719ZqHDz49bjiQpNLKJfQe4Kz1hU1P6ZnO5EoYr/R1qC8+qk/OvTjgqs4Oo66n3YNDmlaQp9yQf2PHnznUr6fa+1WXH9Xt00rVG3P1873dirme3thQqllFub7Vgtdn46q4dmxMDwc5udK8pQzRQnbJioCQSdZ1U+HgNd6WXdINKwOqCGeTRGz8JXCNCWugb6e2b/iGps25S5HomdveHMi0mOfq2c70ZfKe6NhDQJiA9d29+ty6V9UbT6gwHNLnF87VhT7MX3uwtUdf3HAwdfxcx4D2DcbUG09uy/ub1l59++JmNRXSG56tug552r4hfShR8/yQFiyLKBSmBwHZ5dxfasBxpKLCtCZTkh0z0ZH9KmovUyRanDo2JqSisvNkbUI9HWu0Z/P39PKznwqwQuD0RYyjqpz0PTxq83hfnIivbt6p3njyJm8g4eorr+7w5bq/3p++q+srPcOpcCBJMc/qdwd6j/8wZJH+nvGXVCccIBud8wHBGKPwdZck9z6QpPxcha/mKRkmJppTopar/1MNs96quhl3aNk1/y3PS18Fq6djjUaG2gOqEDh9xhj9xZwVKgwn3xfLorn62KyLAq7q7HBgaCTtuG1kRJ79vVfgPqUpkVMPZZrIaxCcqrqQjt+DsKaB7xmy0zk/xEiSQjMa5Xz47bJdvTIVpTJZssEMzg75hfWavfhjqeOc3HIdu3WbE8pVOFI49gOBLHZxRb1+ecmbtW+oT80FJYqwe/KEXFFdrt8cOJQ6XllVLseHnXBvbyzVM4fSN42MOkYxLxlOphVEdVN9ScbrwOuXV2C0/Lqotq5LKJGwap4bVlU9v3fITpMiIEiSiUZkaljjG7+/6ed9QH1dmxSP9cqYkGac/0GFIwWn/kAgy+SFIpo9hYn2p+PP581QaU5U67t7Nb94it5z/LLZGbKkrEDFEUe98aPDVBaX5unt0ys06lm1lBco7DBUJdtV1oZUWUsoQPYz1oeu0TOlpaXFrl69+tQvBDLMTQyrt/MV5U9pVG6+v7s5A5icnmrv15c2HNBAwtPUvIj+8YIGJiWfZVp3JrR1fXIVoxnnhzVt9qR5TovsdMKnCvxkAq9DKJynsuplQZcBYBJZWT1FyypmqX04roaCqC9Dm3Dm9HZ5Wv1YPHW87qm4CouMymvoUUD2ISAAp8laq57D6+S5oyqtapHj8GsEwB+5IUfTsqTXYMS1+s7mUa05nNDM4pDePzdHZbnn/Nonr1tHqzum7VCrR0BAVuLOBjgNnpfQuqc+pu5DL0pK7qZ8wZX/oUiUJSIBTC7f3DSi/9mTfCK+s99T25Cnr13MfKwTKSobG56Kx2kDsgE/mcBp6Dz4TCocSNJg3w4d3P3rACsCgGA8256+6dfLXa4G4mfPvEa/VdWFNPP8sJyQZBypaV5IU5u4DUN2ogcBOA3x2NiNiOKjPeO8EgDObY2Fjg6PHB02U5FrlM9dxUktuDCiOUuSX6RwhDkkyF5EV+A0HL+zsuNEVd14fYAVAUAw/nR+rmrzkze5RRGjv1yYx8TpCQhHDOEAWY9lToHTNDSwT/u33yvXHVXd9NtUVDov6JIAIBCetTo4ZFWZaxQNcdMLZIq1Vt6eLmlgVM7MSpncyJn4tCxzCpwp+YUNmr3440GXAQCBc4xRXQHBAMi0+C/Wy3u1LXmQF1H0j5bLKc/cogAMMQIAYBwJz+qhA4f0za27tKaLuUYAguG19R0NB5I0HJf7wu6MXpMeBAAAxvHPm7bpNwcOSZJ+tLtVn14wUzfV1QRcFYBJJ5YY02THaTuT6EEAAOA4/fGEHjwSDl7z070HA6oGwGRm6ktlqgqPaTAKLanP6DXpQQAA4DghI4WMkXfMQh4RVugBEADjGEXfcZHcNXtlB0YVOq9WTl1JRq9JQAAA4Dj54bDePK1OP9y9X1IyMNw1vSHgqgBMViYvovAlM3y7HgEBAIBxfGB2k5ZXlGrnwKBaykvUWJAfdEkA4AsCAgAAJ7C4rFiLy4rT2kZdVxt7+1Wfn6eq3JyAKgOAzCEgAAAwQdv6BvTxlzaqJx5XyEgfmt2sN0+rC7osADijWMUIAIAJ+vb2PeqJxyVJrpW+tW2PBhOZXW4QAPxGQAAAYIIOj8bSjkc9T/1xAgKAcwsBAQCACbp2alXa8fklRarJyw2oGgDIDOYgAAAwQXc21akgHNIzHV2aVpCntzdndrMiAAgCAQEAgNPwxvoavbG+JugyACBjGGIEAAAAIIWAAAAAACCFgAAAAAAghYAAAAAAIIWAAAAAACCFgAAAAAAghWVOAWCS642P6N59r6pjdEhXVzfrwrLaoEsCAASIgAAAk5hnrT6y9iFtH+iWJP364Db97/Ov0srKxoArAwAEhSFGADCJbe4/nAoHr/n1wW0BVQMAyAaTogfBxuLydu6TyYnKNNXJGBN0SQCQFaaEc8a0FUfGtgEAJo9zPiDY/kHFvv8rqX9QkmSa6hR58xsICQAgqSG/SLfWztavDmyVJJVEcvX2xvMCrgoAEKRzPiC4azalwoEk2d2tsnsPyEyrC7AqAMgen5x7sd5YO1sdo8qYiwUAACAASURBVIO6oLRWBeFI0CUBAAJ0zgcEG4uNbRuNB1AJAGSveUUVmqeKoMsAAGSBc36Scuj8OVLomH9mcaGc5vrgCgIAAACy2Dnfg+DUVChy163yNmyVcnMUWjxPJnLO/7PPOfGhwxo6vEF5ZXMULWSNdgAAgEwJ7E7ZGNMg6XuSqiVZSd+y1n4tE9dyqivkVNN1frbq2/eU9j35WVkvLhlHdcs/o9KZbwy6LAAAgHNSkEOMEpI+Ya2dL2m5pA8bY+YHWA+yVPvabybDgSRZT21rviFrvWCLAgAAOEcFFhCstQettWuO/H+/pFclsbQQxnBjvWnHXnxA1nMDqgYAAODclhWTlI0xTZKWSHoh2EqQjUpn3Jx2XNx8g5wQyzACAABkQuCzdY0xhZJ+Julj1tq+cc6/T9L7JKmxsdHn6pANqha/X5HCWg22r1Fe+XyVz7kj6JIAAADOWcZaG9zFjYlI+rWkh6y1XznV61taWuzq1aszXxgAAABwbjMnOhHYECNjjJH0HUmvTiQcAAAAAMi8IOcgXCLpLklXGWPWHfnvxgDrAQAAACa9wOYgWGuf1km6NgAAAAD4LytWMQIAAACQHQgIAAAAAFIICAAAAABSCAgAAAAAUggIAAAAAFIICAAAAABSCAgAAAAAUggIAAAAAFIICAAAAABSCAgAAAAAUggIAAAAAFIICAAAAABSCAgAAAAAUggIAAAAAFIICAAAAABSCAgAAAAAUggIAAAAAFIICAAAAABSCAgAAAAAUggIAAAAAFIICAAAAABSCAgAAAAAUggIAAAAAFIICAAAAABSCAgAAAAAUggIAAAAAFIICAAAAABSCAgAAAAAUggIAAAAAFIICAAAAABSCAgAAAAAUggIAAAAAFIICAAAAABSCAgAAAAAUggIAAAAAFIICAAAAABSCAgAAAAAUggIAAAAAFIICAAAAABSCAgAAAAAUggIAAAAAFIICAAAAABSCAgAAAAAUgINCMaYG4wxW4wx240xnw6yFgAAAAABBgRjTEjSNyS9QdJ8SW81xswPqh4AAAAAwfYgXChpu7V2p7U2JunHkm4NsB4AAABg0gsyINRJ2nfM8f4jbWmMMe8zxqw2xqzu6OjwrTgAAABgMsr6ScrW2m9Za1ustS2VlZVBlwMAAACc04IMCK2SGo45rj/SBgAAACAgQQaEFyXNMsY0G2Oiku6U9D8B1gMAAABMeuGgLmytTRhj/lTSQ5JCkv7TWrsxqHoAAAAABBgQJMla+xtJvwmyBgAAAABHnXKIkTGm2hjzHWPMA0eO5xtj3pv50gAAAAD4bSJzEP5byWFAtUeOt0r6WKYKAgAAABCciQSECmvtPZI8KTl3QJKb0aoAAAAABGIiAWHQGFMuyUqSMWa5pN6MVgUAAAAgEBOZpPxxJZcfnWGMeUZSpaQ/yGhVAAAAAAJxyoBgrV1jjLlc0hxJRtIWa20845UBAAAA8N0pA4Ix5p3HNS01xsha+70M1QQAAAAgIBMZYrTsmP/PlXS1pDWSCAgAAADAOWYiQ4z+7NhjY0yJpB9nrCIAAAAAgZnIKkbHG5TUfKYLAQAAABC8icxBuE9HljhVMlDMl3RPJosCAAAAEIyJzEH48jH/n5C0x1q7P0P1AAAAAAjQROYgPOFHIQAAAACCd8KAYIzp19GhRWmnJFlrbVHGqgIAAAAQiBMGBGvtFD8LAQAAABC8icxBkCQZY6qU3AdBkmSt3ZuRigAAAAAE5pTLnBpjbjHGbJO0S9ITknZLeiDDdQEAAAAIwET2Qfh7ScslbbXWNiu5k/LzGa0KAAAAQCAmEhDi1tpOSY4xxrHWPiapJcN1AQAAAAjAROYg9BhjCiU9JemHxphDSu6mDAAAAOAcM5EehMckFUv6qKQHJe2Q9MZMFgUAAAAgGBMJCGFJv5X0uKQpkn5yZMgRAAAAgHPMKQOCtfbz1toFkj4saaqkJ4wxD2e8MgAAAAC+m0gPwmsOSWqT1CmpKjPlAAAAAAjSRPZB+JAx5nFJj0gql/Qn1tqFmS4MAAAAgP8msopRg6SPWWvXZboYAAAAAME6ZUCw1n7Gj0IAAAAABO905iAAAAAAOMcREAAAAACkEBAAAAAApBAQAAAAAKQQEAAAAACkEBAAAAAApBAQAAAAAKQQEAAAAACkEBAAAAAApBAQAAAAAKQQEAAAAACkEBAAAAAApBAQAAAAAKQQEAAAAACkEBAAAAAApAQSEIwx/2yM2WyMedkY8wtjTEkQdQAAAABIF1QPwu8knWetXShpq6TPBFQHAAAAgGMEEhCstb+11iaOHD4vqT6IOgAAAACky4Y5CO+R9MCJThpj3meMWW2MWd3R0eFjWQAAAMDkY6y1mfnExjwsqWacU5+z1v7qyGs+J6lF0u12AoW0tLTY1atXn9lCAQAAgMnHnOhEOFNXtNZec7Lzxph3SbpZ0tUTCQcAAAAAMi9jAeFkjDE3SPqkpMuttUNB1AAAAABgrKDmIHxd0hRJvzPGrDPG/HtAdQAAAAA4RiA9CNbamUFcFwAAAMDJZcMqRgAAAACyBAEBAAAAQAoBAQAAAEAKAQEAAABACgEBAAAAQAoBAQAAAEAKAQEAAABACgEBAAAAQAoBAQAAAEAKAQEAAABACgEBAAAAQAoBAQAAAEAKAQEAAABACgEBAAAAQAoBAQAAAEAKAQEAAABACgEBAAAAQAoBAQAAAEAKAQEAAABACgEBAAAAQAoBAQAAAEAKAQEAAABACgEBAAAAQAoBAQAAAEAKAQEAAABACgEBAAAAQAoBAQAAAEAKAQEAAABACgEBAAAAQAoBAQAAAEAKAQEAAABACgEBAAAAQAoBAQAAAEAKAQEAAABACgEBAAAAQAoBAQAAAEAKAQEAAABACgEBAAAAQAoBAQAAAEAKAQEAAABACgEBAAAAQAoBAQAAAEAKAQEAAABASqABwRjzCWOMNcZUBFkHAAAAgKTAAoIxpkHSdZL2BlUDAAAAgHRB9iD8H0mflGQDrAEAAADAMQIJCMaYWyW1WmvXT+C17zPGrDbGrO7o6PChOgAAAGDyCmfqExtjHpZUM86pz0n6rJLDi07JWvstSd+SpJaWFnobAAAAgAzKWECw1l4zXrsx5nxJzZLWG2MkqV7SGmPMhdbatkzVAwAAAODUMhYQTsRau0FS1WvHxpjdklqstYf9rgUAkN2eO9yuzX29WlJarsWl5UGXAwCTgu8BAQCAifiP7a/q+7u3p44/Pud83d7QFFxBADBJBL5RmrW2id4DAMCxEp6ne/buTGu7e8/2E7waAHAmBR4QAAAYj5Ocp5ZijjsGAGQGQ4wAnDHWenph70+09dCTqiho0soZ79WUnGA3Sl/ds1bf3vN99buDuq3mRv1h7a2B1oOJCTuO7qhv1g+O6TV4R9PMACsCgMmDgADgjHlm1/f02PZvSpL2dK/Rgb5N+uPl3w2sns5Ylz6+8a806sUkSV/e8XVVRst1RcWlgdWEiWkdGtRvDu5LHS8rq9QtddMCrAgAJg+GGAE4Y15tfzTt+GDfZnUPtQZUjbS6Z10qHLzm2e5VAVWD03H33h3qio2mjl/s6tC2/t4AKwKAyYOAAOCMKc6bmnYcCeUpP1oSUDXS9IKmMW3N+WPbkH0G4okxbf3xeACVAMDkQ0AAcMZcNeuDKs5NbqAecqK6bs5HlRMuCKyehtxavaX2TYqaiIyMrii/RLdPvTmwesYzlIjpuc4d2jfUFXQpWeXmuoa0P1BNBYVaxD4IAOALY60NuoYJa2lpsatXrw66DAAn4XkJtfVvU0lerfKjxYHV8UjHk/ritq9owB1UfW6t/mb2J7WweEFg9Yxn20C7/mzdj9SbGJYk/XHTSr23aWXAVWWP9d2d+m1bq8qiObq9oUml0ZygSwKAc8kJl4YjIADniFd33K1d++5Xbk6ZFs59vypKs+tm2E8xL6abXrhTfYn+VNvKshX68oK/C7CqsT79yk/1xOGtqeOwcXTfio+oJJofYFUAgEnihAGBVYwgt3O74nuelVPapEjTpTKGkWdnm51779eajV9NHXf2bNJt1/5KkQCH9wSpJ96XFg4kad/w/oCqOaorNqhRN66pecl5Gd2xobTzCeupLzFMQADgq95trgZbXRXUhlQ0y/F9zxE7ahX/XUzeLldOvaPI9Tky+ex7EiQCwiQX2/2UBh/4tGRdSVJ03htVcOXnAq4Kp6u1/em041i8X4c616uu+uKAKgpWVU6F5hbO0uaBbam2y8qD/Vr86/aHdc/+F+XKannZdH1pwR26seZ8vdx3NLicV1SnxnzG2QPwT/tzcR16PrkoQOdaV5XLwqq5NOJrDfH7RuWuSdbgHvRke61y3pXnaw1IR0CY5EbX/jAVDiQptvl+5V30QTn5ZQFWhdNVPKVZOnj02MhRcWFTYPVkg3+a/3l9c/d/aufgbl1cdqHe2/iOwGrZ2Nequ/cfXV71+a6d+tXBtXpL/YXKD0X1xOGtasgr1VsbLgqsRgCTU+e6xJhjvwOCuym9Bm+rq9jvRhW5MCJTzKiGIBAQJr3j5qBYm2qz1srt2CwTyVeolA2Kstm8GW9TR9cGtR1epVAoR4vmfkCFBbWB1WOtVVv/ZoWdHFUWTg+khuqcSv3tnE8Fcu3j7R1nhaLX2q6tXqBrqyfvfBEAwTLhkx/7UkOlI7vPS2tzH4vLXZVQ7kfzZQoZbuQ3AkKArLVSPCYT4MocOYvepkTbBskmfzHD9S2ywz2yoRz13/cRuYc2SZKic25UwdX/K7A6cXKRSKGuvvjfNDTSoUg4P7C5BxsOPqiHt/yrBmJdei1ozq++Rrcv/PtJPbflwtJm5ToRjXhH1/G/rGJ2gBUBQFL18ohaHz763lS9wt/eA0mK3BRV7IcjUv9xJwat3FcSCi/3v6bJjlWMAuLt36n4gz+Wejql6npFbvkjOcXBDOtJdGxRbMsDGt18vxRL/naGahbKbXs57XWFt31TkdolQZSIs0DvSLv+7ak3yR4zZO01b1v6Nc2oWB5AVdljfe8+fXfPsxpyR/Wm2qW6vvq8oEsCAEnSyGFPg62uwnlGBfUhhX2cIOzuchX78XAyHBiNGdgQuSNH4QsICBnCKkbZxFpP8ft/JPV3Jxva9yv+319W5B0flVNe7Xs94co5GtlwbyocSBoTDiTJG2j3syycZQ70bho3HEhS38ghn6tJuvfAr/SLg79WQThf7228S8tLWwKpQ5IWFTfoKwvfEtj1AeBEQnlGXRtcjXRYyYmrenlYVRf5c1Me/9kxPQfHhQNT7Sh0PreqQZi8ff5BGho8Gg5eEx9V/NffD6YeSV7H5pO/IFqoSOMKf4rBhPT279bQcHaEtrg7oq7BfRrvYUQ0lK9ZlZf4XtMTnc/qyzu+rh1Du/Vy3yb9xcb/pfbRDt/rOJHnOnfoO7uf0pruPUGXAmCSO7w6kQwHkuRJ7c8mFOv1Tv5BZ4CNW9mu8UeyRN6So5wP5clEmX8QBGJZAEzBFKm8Suo87qlqx0HFf/MjRW58m/9FhaInPZ134Z/IyQ1uV1wcFU8M6vEXPqFDnWslGc1u/gMtO/8vAq3px2s+rt3dL6WOw05URTlVmlq8QBc3vUOFOf4v3flc16q047iNa3XPWt1UfZ3vtRzv27ue1Hf2HF2a9o+bLtV7my4LsKLstWewXx2jI1pYUqaoEwq6HOCcNNg6tvc31mcVzfCffRMxcqaH5O1Mv74pNgovYlhRkOhBCEj45rvGbfc2vSSvy//hGJH6ZSc9H566yKdKcCrbdv/8SDiQJKutu+7V4e5XAqunvX9bWjiQpBkVK/ThlT/T7Qv/TjVFwUzGnVHQPKbtm7v/U2t61mvEHQmgoiTP2rQlTyXp27uf1lPH7KiMpH/d8ore/tzj+tia53XnM4/qwPDQqT8IwGlLjPOWmFPqz5P76Fty5MwMHe2AjkjhG6NytyXkdWa+FwPjowchIE5FjZRfKA0NjD0ZG/W9nsjMazSy9vup1YyOFZ3/JoUr5/peE8Y3MHhgTFv/wH4dOPS8du77tXKjZVo870OqqfRnvH3IGfuUJ2SCf/JzW82NWtf7ih4+/HiqrSPWqQ9v+KQ8ebq8/BL97ZxPKT/k72Y8RhozzlaS/mPXE1rJykYp+4cGdc++XanjQ6Mj+uHu7frLeQsDrAo4N4VzpXjvMQ1GMmF/AoKZ4ijnPXmyMSuvLXkPEvv+iDRoJSOFr4oqcvXJRzngzKMHISDGOApffbt0XJe5qWmUqa73vZ7EvhfGhIOc8/5QRW+7RwVXZMda8khqqL0y7TgcztdIrFsbtvw/DQ4dVGfPRj2+6hMajfWe4DOcWRUFTZpXfXXqOOLkanlTAMPkjhNxIvqHeZ9TRSR9eJOn5M/5E53P6Lv77va9ricOb5Fjxv7h7YsH16uRjbrGeVDSFeNrBGRCxQXpz4tLF4QUzvV37L+JGoUaQ3KfiSfDgSRZKfFYTHaAngS/0YMQoNCcRXIaZ8rdvFa246BMSYVCi1bIjHPzkGkmr2RsfVVzFSpp9L0WnNzUygt1acsXtX33LxSJFGjBrHdp0/YfpL3GdUfU0bVB9TWX+lLTHQu/oO2Hb1TvSLtmV16qolz/V+M6kYvKLtD97b8d99zmgW2+1tIdG9TfbPqVYuOs9nRr7WJfa8l2C4pLVZ9foP1Dg6m262v8f3gCTAYlc8KKFjvq3+0qt9xR0czgnh/b/uPCgCfZQckUBlPPZEVACFpunpyqOnmeJ/elJ+U+86Cc+UsVvuYOmZB/357ozGs0uvFXcts3SJJC1ecrOvPqU3wUgjKt9mpNqz36/SktnqW9Bx5OHRs5Kima4Vs9xjiaVZkMI64X147Dzys3MkV1xcHuEOxZT1NzqlUUnqIRd0QJ66Z6ECSppdjfm/ItA+1jwkF1TpHe33y5bmBfhDQhY/T1Cy7W3Xt26NDIsK6rqdfKqpqgywLOWfk1jvJrgh1YYl0rU+1Iu4++T5s6R041A178RkAIkB0eUvye/yvbcTCt3duwSm5JpcIXXeVbLSacqym3/4cSB9ZJksK1SwLpycDrM3f6neru2aJ9Bx9XJFKgxfM+rML8qb7XMTDaqf9e9T51D++XJM2rvkpXzfqw9ve8rLri81Re4G+P1E8O/FLf3nt0+eCySKlqcqp0KNah6yqv1Fvr/8DXeuZNqVHUCSvmJVJtTfkV2jvcpdaRHtXnlfpaT7aryMnVn80ONmQC56rEsFXfjuTmaFOaHRkn+L/5se+NyNt25CGKkUIXhBW5PifYoiYpAkKA3HXPjAkHr7EHd/tai5R8ChypW+r7dfH7C4dydWnLF5Vwh+Q40XEnDvvhxb33psKBJL3a/qg2tz8mKyvJ6Ob5n9WS+lt8q+fJzmfTjrvi3frKgr/XvClzfKvhWMWRfN3VsFzf2/uc4tZVyDh6oXunXujeqXv3r9b3Wt6r2nGG+wHAmTTa42nHj0flDiePCxocNd8RDfTBoLfPPRoOJMlK7paEIm8gIASBPpsA2YETTyI1dWOXaATGs33P/+jH91+uH923XL9+9E4NDrUFVstQvHtMm00t2WP12PZv+lrPtLz0MesRE9HU3OCGqfTGh/SDfc8rfmSYkXvMwgCD7qgeODLEDwAyqXNdIhUOJGlwn6fWh+Nqfz6uWF8wE4JtYpzGfin+uP8rO4KAEKjQ3CVK23nWGCkSlbPoYoWWsmkSTm1ktEcvrP+i3CPr+g+NHNLjqz4eWD0Lp94oY45dmSv9adRoYlDWjr9r5pk27A7rsvKLNatguiQp18nVx2d8UCWR4Db829R3UKPeeH8Fk3IC6vnJJn79fACT2Xg3492vuDr0XELbfziq+ID/v4fONEeqGNuD4b4ydlEHZB5DjALkNMxQ5I4/lrvhBSknT6FlV8gpqwq6LJxFOrrW6/hF9QcG94//Yh80lC7SXRd8Q+sP3KfcSJFcL67V+36aOn9B/Zt86cJ+vvtFfe7Vf9CAO6ii8BR9fvandWn5chWGCzJ+7ZOZVVitsHGUGGe/kam5xbq5ZvKu8b93cEBf2LhWm/p6dH5xqf5qwRLV5Qf7/QLORX07XA0dPHEvgTsi9byaUOUyfx9YGMco5315Gv3icRsi9lrZhPVtXwYkERAC5jTPldOcXZuQDb/0XcVevU8mt0h5F71fkYaLgi4JJ1BWPHYsfX7AS4xOK1uiaWVLJCWfBtcVL9C+nvWqKz5Pi2pv8qWGf97+dQ24yeUx+xL9+v7+n+iG6uBX5arIKdRfzb1Z/7r9EfUmhnV15VxdWj5LnqSV5bOUH568mwH9w6Z12tTXI0na0NutL21ar6+3XBxwVcC5ZaTT0577YmnPlQoaHA3uSw8MJhTMzbhT6MiUGtnuYwrkTjUQfNmRZnTLAxp54cg48T5p4DefVPFdv5CTXxZsYRhXQX6Nigub1TtwdMfZcDhPPX07VDxleuArURljtLD2Ri2svdG3a8bduFpH0if/HxgNbl7G8a6vPk/XVS2Qaz2Fj9socTLb1Js+f2VT39j5LAB+P/273TE7uedXGyUGjUa7kieixUYl84N7bwpfE1X8p6OpOsMro/QeBICAgDSJfavSG9xRJQ6uU3SGf0uu4vQMj3amHff279T9j79NxVOadeXyr6kgL3s2LfPDjw787JiJ0UmXl18SUDXjM8YobAgHx1pYUqb1PV2p40UlPJQAzrTcsrFTT3OrHFUtj6hvhyvrSUUzQwpFg7shDy+JyKlx5O10ZWpDCjXzXhkEJilDkuT1t2ngwU8rvueZ484YhSpmBVITJqa8ZP647b39u7Rhy3d8riZ4q3vWjmm7qeraACrB6fjcgiW6oLRCuU5IF5ZV6tPz2VkaONMKmxyVLQol148wUsm8kIpnheREjErmhlU6PxxoOHiNMzWk8CVRwkGA6EHwme3rlrt5rUw0V868pTI5uUGXJEkaeOizcg9tSm+MFijvog8oVNwQTFGYkAsXfkrPrPlfOtw9donM/oE9AVQUrDmFM7WqZ03qOMfJ0ZzC7Ay57SN96k+MaGYhixPU5uXraxesCLoM4JxmjFHdVVFVr7CSlcL5wYcBZCcCgo9sz2HFvv9VafTI4sMvPCJn/gUKNcyQ0xTMxk2S5I32jw0HMorOuFo5828NpCZMXGFBra5f+W395ok/Unfv5rRzlWWLfKmhrW+rfrr+M+odOaj8aKluPe9vNL38Ql+ufbx3NbxNu4f26umuF1QaKdEnZnxYG/o36Setv5BjHL2j/s1qKQn+6fRXt/9O9+x/UVbSgim1+j8L79SUSHY8MABwbgvnJYNB3w5Xg62e8mscFc1yAp+3huxhzqY1p1taWuzq1auDLuN1Szx5v9xVj457LnTZzQpfeKXPFSVZ66nvB7fL6x87kTP3wvcrr+XdAVSF07V+87/rla3/lTqORKbojusfVMjJ7HMA10voK49fr5HEQKotZCL6+BUPKDcyJaPXPplhd1hRJ6rtg7v07rUflivvSG0hfXfxNzSrcEZgtb3YtUsfefnutLY/abpM72m6NKCKAEw2h16Iq/3ZoxsiVC4Lq+ZS9mKZZE6YCJmD4KeTJHN39eP+1XEcYxzlX/XXUu7YDaQSB9cHUBFej/NmvUdzp79VBflTVVNxoa69+JsZDweSdGhgR1o4kCTXxrW/J9hdgfNCeQqZkJ7sfDYVDiTJta4+tOEv1RM/8U7mmfZfY+b6SK3DrNpzYHhI9+7dqac72uSdRQ+vgLNR57r03dIOr0mob1dC1gvmd89rc+VuTcjG+d3PBgwx8lFo4XK5Lz8nDQ+NPWmCzWqRugtU9NafqO/7t0qJo9uah6sXBFgVTkcoFNWsptslSda6x+1onDmleXVyTFhe2tacRlVTgntCf6y63Klj2voS/bqv7UHd1fCWACqS2kfHhpOW0ib/C8kiL/d06WNrnlPMS4a5q6tr9fnzLwi4KuDcldzr4OjNuHWlPb+MK7cioelvyfF1snLsvlG5z8WTB0XJDdOccVZcgn/46vvIFJcp+q5PKnzVbTKz08eGhy8KfhnRUF6JCq77gpwpUyUnrOjsG5S79K6gy8IEDY906qGn3qvNO+/Wll336KGn3qP+wdaMXzc3Uqib539Gjkk+bzBydNn096oo4A3bXnNd5ZWaUzBzTPuQOxxANUmLixvTjksj+bqh+ryAqskOd+/ZkQoHkvRI+wG1Dg0GWBFwbqtaPv4z4pHDVj2vur7V4R32joYDSeqzSjwR8+36GB89CD4zBVMUWrpSoaUr5e3fKe/AHqmoVLZ9vxJP/UahhctlioNb/zvatFLRppWy1jJZ6Syz9+CjisX7UscJd1h7Wn+r82Znfg7Jorqb5Xqu7n/1S7Ly9Mzu72lq0VzNrlqZ8WufStgJ6xsL/1lvXv1udcWTO/Xmh/J0Y3VwS59+ZObVGnRH9VzXTs0oqNSnZr9h0v++udYbp42hBkCmlJ0XVl61o47/396dB9lVlnkc/z536yWdztqBJumsphNilJA0iKyBgEUACSCMWjA1iMsMwyaOTIk4lpZjFe7WzNSgIO7siiIyakBl0ZmELCSQQCBEspKtsxCTkF7ufeaPc/rSSXqhpPu8J/TvU3Wrz11y7y/n3vve85zzvu9Z2MZrqw/+/hUPJPfd832Hv5bv1Xc/NBUIAWXGTMQGD6X1h1+HtqhbT3HpU+Q/ejOZmtpgubxUpLR7PZnaeiynWVWOFIX84Z+ZQkKDhN2dx9fcRsfh6mKplT+s/u9UFAgAg3M1/Pj423hoy29oKbXw/qPOpaFqdLA8Q/LV3Dr90mCvn0aXNkxg4Y7t5aJgck0td/7lRY6prKYim2XRzu2Mra7hqomN1FVWBU4r8vZQVZeh/swCe9cfoBj3Ls7kYciU5M4/kGnIYHWGb3+jKMhOytJ6zwF8T4nsjDy592jwdNJUhRzEtwAAD/5JREFUIARWXPVMuTgAoK2V4vwHyFzy0SB52ptXs+9/bqK0dwtWUUv1nM9T0MwqqbJh8+MsX/Vd2tr3MXncJUxvvBKAsfWzeWnYdJp3rQBgaO07mNAwN5FMJS/yeqejFwD7WtM16LauYiQfG5eeLnMHim2s3ruV8dUjNb0pcOKIUdxxwmk8uX0z21sO8MirG1i99+DP1PLdO3lhz25+eNIZgVKKvP3kBxmTPlzBjuXt4DD8XTkqhibXA90yRsXHqmj/Uxv+mpOZnqXt4Vb4a1QwlNa1QCE6w7IkRwVCaF0MTi5tXh8gSOT1P3+b0t5oulNv2cP+J75Cfux7sYzOZpgGe/e/ylOLP4t71D90+arbGDxoDONGn002W8E5p97O1uYllErt1NedSCaBWYwAspkc76o/l+WvPlK+7bjRFyTy2keiZ1/byE3PPcCe9tepzOT5t6kXcNaoY0PHCq6xdgiNtUO4csET3T7m5b172Lh/H2OqByWYTOTty93B4OhT8mTyYbo62uAM+bkVABTXtJeLgw7Fle0qEBKmAiGwzPQmik89Ap362tqQYcHyFHcfXJz4vu3QfgAK+jFOg207lpWLgw5bmhcxbvTZAGQsS31d8icoW9O8gJIXGT+8iYrsICaNPImZYy5KPMeR4j9efow97dEg6QOlNr758nxm100lM8DHIXRoKx0+HqFDZSbL8EJFgmlE3r5adpVY91ArLbucTAWMOafAkMlhdwjasEw0O3+nGiEzQnPqJC3YGjez68xslZmtNLOvhsoRWqZ6MNnZF1I+V0UuT+6084LlKUw4/aDruWNmYioOUmP4kKlv6rYkvbT9T9y99Aae2/xb1u5czI796zh+zDws8NS9abblkGlOd7buo7XU3s2jB555Y8Z1eXshk+H6Ke+kOqd9WyJ9YctTbbTsirvytMD6R1p59Y+tFFvDDRLODM+QO6cAcZ1iozPkTisEyzNQBWllzexMYB5wnLu3mNmoEDnSIjfrdLKT3kmpeTOZ0ROwqnAb5FUnXw/5Kto3LiI7spGq91wdLIscbmjtRGZN/xTPvXgH7e2vM3Hs+Uwa+/6gmZZtevig68371rJp9woahh3Xzb+QOXXTuH/TovL1U0ZMpjKrw+cdLmuYwKb9+3ho0zqyZnxo3CTOqKvn6KoqavPaUBDpKx3FQZnDjmVFigegYW6471p+doFcUx7f72RGaWdTCKF2w1wN3OruLQDuvi1QjtSwoSPIDh0ROgaWq6D6vdeGjiE9mDrxgzSO/wDuJbLZ8BtL1fnDz8BdVRgaIMmR47pJZzGsUM3iXWuZOriej4w7JXSk4Jbv3sEjmzZQk8tx2diJ3Dj1XVzTOI1SyfnFpnXcsWYVU2qHcPn4d1CV1REEkb5QOzHL9p2HH73csya58yB0x2oMq1G3y1DMA8wzbWbLgIeAc4EDwKfdfVE3j/0E8AmAsWPHzlq3bl1iOUWkdzv3b+BHT/8je1t3ADBzzMWcP+0zgVPJkWTFa7u4ZvGfy1OcjihUcM/JZ1Gdy/FfL63k3vV/KT929qh6/v3dTaGiirytlIrOtgXtNC9pp/Pwtso6Y/IVml1tAOi2Auu33TBm9hhwdBd33RK/7nDgJOAE4H4zm+hdVCvufjtwO0BTU5POnCGSMsOrG7j2tAd5ZediBlfUUV87JXQkOcL8bvPGg06KtqO1hYU7tnHmUcfw6JaDzwb+5LbNtBSLVGQ1s5rIW5XJGkefkqemIcP6R1opHoBcNRxzpro8DnT9ViC4+9nd3WdmVwMPxgXB02ZWAkYC2/srj4j0n3y2ksY6nS9D/jZDuxhXMCyeqaiuspIdrW+cK2ZooYJ8Rn2SRfpSzdgsUz9eSetupzDMyGTVtWegC9XK/hI4E8DMGoEC0Bwoi4iIBHRJw/iDzmtwet3RzBgWjcm6ZvI0BsVjDgqZDNc3vlPTwYr0g0zOqByZUXEgQLgxCAXg+8AMoJVoDMIfevt3TU1Nvnjx4v6OJyIiCWsvlVi6q5maXJ5ph5wLZn97O6v27GZCzeDykQUREXnLkh+D0BN3bwWuCPHaIiKSPrlMhhNHdD3jdXUux8zhIxNOJCIycKkjp4iIiIiIlKlAEBERERGRMhUIIiIiIiJSpgJBRERERETKVCCIiIiIiEiZCgQRERERESlTgSAiIiIiImUqEEREREREpEwFgoiIiIiIlKlAEBERERGRMnP30BneNDPbDqzrg6caCTT3wfP0lbTlgfRlUp7epS2T8vQubZmUp3dpy6Q8vUtbJuXpXdoyvV3zNLv7uV3dcUQVCH3FzBa7e1PoHB3SlgfSl0l5epe2TMrTu7RlUp7epS2T8vQubZmUp3dpyzQQ86iLkYiIiIiIlKlAEBERERGRsoFaINweOsAh0pYH0pdJeXqXtkzK07u0ZVKe3qUtk/L0Lm2ZlKd3acs04PIMyDEIIiIiIiLStYF6BEFERERERLqgAkFERERERMoGVIFgZt83s21mtiJ0FgAzazCzP5rZ82a20sxuCJyn0syeNrPlcZ4vhszTwcyyZvaMmf06dBYAM1trZs+Z2TIzW5yCPEPN7GdmtsrMXjCz9wbOMyVeNx2XPWb2ycCZbow/0yvM7B4zqwyc54Y4y8pQ66ar9tDMhpvZo2a2Ov47LHCey+J1VDKzRKcY7CbP1+Lv2bNm9gszG5qCTF+K8ywzs/lmdkzIPJ3u+xczczMbGTKPmX3BzDZ1ao/OSypPd5ni26+LP0srzeyrIfOY2X2d1s9aM1sWOM8MM1vQ8RtrZicmlaeHTMeZ2f/Fv/0Pm1ltgnm63Fbs9/ba3QfMBTgdmAmsCJ0lzlMPzIyXBwMvAdMC5jGgJl7OAwuBk1Kwnj4F3A38OnSWOM9aYGToHJ3y/Aj4WLxcAIaGztQpWxbYAowLmGE08ApQFV+/H7gyYJ7pwAqgGsgBjwHvCJDjsPYQ+CrwmXj5M8BXAuc5FpgCPA40pWD9vA/IxctfSXL99JCpttPy9cB3QuaJb28Afkd0YtPE2spu1s8XgE8n+T69iUxnxt/7ivj6qNDvWaf7vwF8PvD6mQ/MjZfPAx5PwXu2CDgjXr4K+FKCebrcVuzv9npAHUFw9yeBnaFzdHD3ze6+NF7+K/AC0cZMqDzu7nvjq/n4EnQUu5mNAc4HvhcyR1qZ2RCixuxOAHdvdffdYVMdZA6wxt374gzob0UOqDKzHNGG+asBsxwLLHT3/e7eDjwBXJJ0iG7aw3lEBSfx34tC5nH3F9z9xaQyvIk88+P3DGABMCYFmfZ0ujqIBNvsHn5TvwX8a5JZeskTTDeZrgZudfeW+DHbAucBwMwM+DvgnsB5HOjYQz+EhNvrbjI1Ak/Gy48CH0gwT3fbiv3aXg+oAiHNzGw8cDzRXvuQObLx4cVtwKPuHjQP8G2iH5pS4BydOTDfzJaY2ScCZ5kAbAd+EHfD+p6ZDQqcqbMPkeCPTVfcfRPwdWA9sBl4zd3nB4y0AjjNzEaYWTXRHrKGgHk6O8rdN8fLW4CjQoZJuauA34QOAWBmXzazDcDlwOcDZ5kHbHL35SFzHOLauBvW95PsNteDRqI2YKGZPWFmJ4QOFDsN2OruqwPn+CTwtfgz/XXg5sB5AFYSbZADXEagNvuQbcV+ba9VIKSAmdUAPwc+ecjeoMS5e9HdZxDtGTvRzKaHymJmFwDb3H1JqAzdONXdZwJzgWvM7PSAWXJEh0Jvc/fjgX1EhxqDM7MCcCHwQOAcw4ga9gnAMcAgM7siVB53f4Goe8p84LfAMqAYKk93PDpurXmwu2BmtwDtwF2hswC4+y3u3kCU59pQOeKC97MELlIOcRswCZhBtIPgG2HjAFG7PRw4CbgJuD/eex/ahwm8Qyd2NXBj/Jm+kfgIeWBXAf9sZkuIuvm0Jh2gp23F/mivVSAEZmZ5ojf8Lnd/MHSeDnE3lT8C5waMcQpwoZmtBe4FzjKznwbMA5T3SHccFv4FkOgAqkNsBDZ2OtLzM6KCIQ3mAkvdfWvgHGcDr7j7dndvAx4ETg4ZyN3vdPdZ7n46sIuoT2kabDWzeoD4b2JdH44UZnYlcAFwefyjnCZ3kWDXhy5MIirEl8ft9hhgqZkdHSqQu2+Nd3yVgDsI21532Ag8GHfrfZroCHlig7m7Ene/vAS4L2SO2D8QtdMQ7WAK/p65+yp3f5+7zyIqotYk+frdbCv2a3utAiGgeI/BncAL7v7NFOSp65iVw8yqgHOAVaHyuPvN7j7G3ccTdVX5g7sH2/MLYGaDzGxwxzLRoMVgs2K5+xZgg5lNiW+aAzwfKs8h0rI3aj1wkplVx9+5OUR9OIMxs1Hx37FEP8p3h8zTya+IfpyJ/z4UMEvqmNm5RF0eL3T3/aHzAJjZ5E5X5xG2zX7O3Ue5+/i43d5INLhyS6hMHRtQsYsJ2F538kuigcqYWSPR5BLNQRNFO1JWufvGwDkgGnNwRrx8FhC6y1PnNjsDfA74ToKv3d22Yv+213054jntF6KNlc1AG1HD9dHAeU4lOiT0LFE3g2XAeQHzvBt4Js6zggRnMngT2WaTglmMgInA8viyErglBZlmAIvj9+2XwLAUZBoE7ACGhM4S5/ki0YbTCuAnxLOHBMzzFFEhtxyYEyjDYe0hMAL4PdEP8mPA8MB5Lo6XW4CtwO8C53kZ2NCpvU5sxqAeMv08/lw/CzwMjA6Z55D715LsLEZdrZ+fAM/F6+dXQH0K3rMC8NP4fVsKnBX6PQN+CPxTkuumh/VzKrAkbh8XArNSkOkGoiO9LwG3ApZgni63Ffu7vbb4xUVERERERNTFSERERERE3qACQUREREREylQgiIiIiIhImQoEEREREREpU4EgIiIiIiJlKhBERKTPmNlFZjYtdA4REfnbqUAQEZG+dBHQZYEQn61VRERSTgWCiIj0yMyuMLOnzWyZmX3XzLJmttfMvmxmy81sgZkdZWYnAxcCX4sfO8nMHjezb5vZYuAWM3vFzPLx89Z2vi4iIumgAkFERLplZscCHwROcfcZQBG4nOhs2Qvc/TjgSeDj7v6/RGervcndZ7j7mvhpCu7e5O5fBB4Hzo9v/xDwoLu3Jfc/EhGR3qhAEBGRnswBZgGLzGxZfH0i0Ar8On7MEmB8D89xX6fl7wEfiZc/AvygL8OKiMhbp/6gIiLSEwN+5O43H3Sj2afd3eOrRXr+PdnXseDufzaz8WY2G8i6+4q+DiwiIm+NjiCIiEhPfg9camajAMxsuJmN6+HxfwUG9/KcPwbuRkcPRERSSQWCiIh0y92fBz4HzDezZ4FHgfoe/sm9wE1m9oyZTermMXcBw4B7+jSsiIj0CXvjCLGIiEj/M7NLgXnu/vehs4iIyOE0BkFERBJjZv8JzAXOC51FRES6piMIIiIiIiJSpjEIIiIiIiJSpgJBRERERETKVCCIiIiIiEiZCgQRERERESlTgSAiIiIiImX/D+CXpl/X+Y1OAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 864x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot = sns.catplot(x=\"entry\", y=\"value\", data=pre_sigmoid_df)\n",
    "plot.fig.set_figwidth(12)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot = sns.catplot(x=\"entry\", y=\"value\", data=post_sigmoid_df)\n",
    "plot.fig.set_figwidth(12)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Flat line test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "predictions tensor([0.9873, 0.9888])\n",
      "uncertainties tensor([0.0126, 0.0111])\n"
     ]
    }
   ],
   "source": [
    "new_x_test = minmax_scale(np.full((2, 100), 0., dtype=np.float32), axis=1)\n",
    "new_x_test = np.expand_dims(new_x_test, axis=-1)\n",
    "new_x_test = new_x_test.swapaxes(-2,-3)\n",
    "new_x_test = torch.tensor(new_x_test)\n",
    "\n",
    "net.eval()\n",
    "with torch.no_grad():\n",
    "    outputs = net.sample(10, new_x_test)\n",
    "    outputs = torch.sigmoid(outputs)\n",
    "    uncertainties = uncertainty_avg(outputs)\n",
    "    scores = score_avg(outputs)\n",
    "\n",
    "plt.plot(new_x_test[:,0])\n",
    "plt.show()\n",
    "print(\"predictions\", scores)\n",
    "print(\"uncertainties\", uncertainties)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Noise test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "predictions tensor([0.9804, 0.8770])\n",
      "uncertainties tensor([0.0192, 0.1079])\n"
     ]
    }
   ],
   "source": [
    "new_x_test = minmax_scale(np.random.randn(2, 100).astype(np.float32), axis=1)\n",
    "new_x_test = np.expand_dims(new_x_test, axis=-1)\n",
    "new_x_test = new_x_test.swapaxes(-2,-3)\n",
    "new_x_test = torch.tensor(new_x_test)\n",
    "\n",
    "net.eval()\n",
    "with torch.no_grad():\n",
    "    outputs = net.sample(10, new_x_test)\n",
    "    outputs = torch.sigmoid(outputs)\n",
    "    uncertainties = uncertainty_avg(outputs)\n",
    "    scores = score_avg(outputs)\n",
    "\n",
    "plt.plot(new_x_test[:,0])\n",
    "plt.show()\n",
    "print(\"predictions\", scores)\n",
    "print(\"uncertainties\", uncertainties)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Note\n",
    "In binary classification, the maximum uncertainty is 0.25 (need to verify this), and the minimum is 0."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}