7.model_trackster_graph_net.link.prediction.PU.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "6c07744a",
   "metadata": {},
   "source": [
    "# Graph net for link prediction\n",
    "\n",
    "Question: does including the neighborhood provide a benefit over the pairwise classification?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "e3962d51",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import torch\n",
    "import sys\n",
    "\n",
    "import torch.nn as nn\n",
    "from torch.optim import Adam\n",
    "from torch.optim.lr_scheduler import CosineAnnealingLR\n",
    "\n",
    "from torch.utils.data import random_split\n",
    "from torch_geometric.loader import DataLoader\n",
    "\n",
    "import torch_geometric.nn as gnn\n",
    "\n",
    "from reco.model import DynamicEdgeConvBlock\n",
    "\n",
    "from reco.training import precision_recall_curve, roc_auc\n",
    "from reco.loss import QualityFocalLoss\n",
    "from reco.datasetPU import TracksterGraph\n",
    "\n",
    "\n",
    "ds_name = \"CloseByPion200P\"\n",
    "\n",
    "data_root = \"data\"\n",
    "raw_dir = f\"/Users/ecuba/data/{ds_name}\"\n",
    "\n",
    "data_root = \"/mnt/ceph/users/ecuba/processed\"\n",
    "raw_dir = f\"/mnt/ceph/users/ecuba/{ds_name}\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "02f59304-76e4-4132-a379-978a46a042fd",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Using device: cuda\n"
     ]
    }
   ],
   "source": [
    "# CUDA Setup\n",
    "device = torch.device('cuda' if torch.cuda.is_available() else \"cpu\")\n",
    "print(f\"Using device: {device}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "ba68cd4a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "TracksterGraphPU(graphs=7569, nodes=341008, radius=15, bigT_e_th=5, lp)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ds = TracksterGraph(\n",
    "    ds_name,\n",
    "    data_root,\n",
    "    raw_dir,\n",
    "    N_FILES=400,\n",
    "    radius=15,\n",
    "    bigT_e_th=5,\n",
    "    link_prediction=True,\n",
    "    pileup=True,\n",
    ")\n",
    "ds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "f81c0a6c-7d1a-4be2-96df-7663a261e1b8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "729995"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(ds.data.y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "604298f4",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train graphs: 7191, Test graphs: 378, total: 7569\n"
     ]
    }
   ],
   "source": [
    "ds_size = len(ds)\n",
    "test_set_size = ds_size // 20\n",
    "train_set_size = ds_size - test_set_size\n",
    "train_set, test_set = random_split(ds, [train_set_size, test_set_size])\n",
    "print(f\"Train graphs: {len(train_set)}, Test graphs: {len(test_set)}, total: {ds_size}\")\n",
    "\n",
    "train_dl = DataLoader(train_set, batch_size=8, shuffle=True, drop_last=True)\n",
    "test_dl = DataLoader(test_set, batch_size=8, shuffle=True, drop_last=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "3f4a0b99-bef3-4c82-a0ac-d88e1521382b",
   "metadata": {},
   "outputs": [],
   "source": [
    "class LinkPredictionNet(nn.Module):\n",
    "    def __init__(self, input_dim, output_dim=1, dropout=0.2):\n",
    "        super(LinkPredictionNet, self).__init__()\n",
    "        self.norm = gnn.LayerNorm(input_dim)\n",
    "        \n",
    "        hdim1 = 32\n",
    "        hdim2 = 64\n",
    "        hdim_fc = 128\n",
    "\n",
    "        self.edgeconv1 = DynamicEdgeConvBlock(input_dim, hdim1, k=8)\n",
    "        self.edgeconv2 = DynamicEdgeConvBlock(hdim1, hdim2, k=8)\n",
    "        \n",
    "        # Edge features from node embeddings for classification        \n",
    "        self.nodenetwork = nn.Sequential(\n",
    "            nn.Linear(2 * hdim2, hdim_fc),\n",
    "            nn.Sigmoid(),\n",
    "            nn.Dropout(dropout),\n",
    "            nn.Linear(hdim_fc, output_dim),\n",
    "        )\n",
    "\n",
    "    def forward(self, X, edge_index, batch=None):\n",
    "        H0 = X #self.norm(X, batch)\n",
    "        H1 = self.edgeconv1(H0, batch=batch)\n",
    "        H2 = self.edgeconv2(H1, batch=batch)\n",
    "        \n",
    "        H3 = H2\n",
    "\n",
    "        src, dst = edge_index\n",
    "        edges = torch.cat([H3[src], H3[dst]], dim=-1)\n",
    "        return self.nodenetwork(edges).squeeze(-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "2539cf65",
   "metadata": {},
   "outputs": [],
   "source": [
    "def train_link_prediction(model, device, optimizer, loss_func, train_dl):\n",
    "    train_loss = []\n",
    "    model.train()\n",
    "\n",
    "    for data in train_dl:\n",
    "        data = data.to(device)\n",
    "        optimizer.zero_grad()\n",
    "\n",
    "        seg_pred = model(data.x, data.edge_index, batch=data.batch)        \n",
    "        loss = loss_func(seg_pred, data.y)\n",
    "        \n",
    "        loss.backward()\n",
    "        \n",
    "        optimizer.step()\n",
    "        train_loss.append(loss.item())\n",
    "\n",
    "    return np.mean(train_loss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "259ee8b3",
   "metadata": {},
   "outputs": [],
   "source": [
    "@torch.no_grad()\n",
    "def test_link_prediction(model, device, loss_func, test_dl):\n",
    "    test_loss = []\n",
    "    model.eval()\n",
    "    for data in test_dl:\n",
    "        data = data.to(device)\n",
    "        seg_pred = model(data.x, data.edge_index, data.batch)\n",
    "        loss = loss_func(seg_pred, data.y)\n",
    "        test_loss.append(loss.item())\n",
    "\n",
    "    return np.mean(test_loss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "b85638de",
   "metadata": {},
   "outputs": [],
   "source": [
    "model = LinkPredictionNet(input_dim=ds.data.x.shape[1])\n",
    "epochs = 101\n",
    "model_path = f\"models/LinkPredictionNet.PnetL.ns.{epochs}e-{ds_name}.r{ds.RADIUS}.e{ds.bigT_e_th}.f{ds.N_FILES}.pt\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "2d3f1e82-74c5-4629-9b20-f05d7662aaf1",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 0:\t train loss: 64.004\t validation loss: 52.243\n",
      "Epoch 2:\t train loss: 44.874\t validation loss: 46.791\n",
      "Epoch 4:\t train loss: 40.543\t validation loss: 45.314\n",
      "Epoch 6:\t train loss: 38.907\t validation loss: 43.985\n",
      "Epoch 8:\t train loss: 37.668\t validation loss: 51.030\n",
      "Epoch 10:\t train loss: 36.810\t validation loss: 50.124\n",
      "Epoch 12:\t train loss: 36.358\t validation loss: 49.663\n",
      "Epoch 14:\t train loss: 35.639\t validation loss: 39.058\n",
      "Epoch 16:\t train loss: 35.273\t validation loss: 39.405\n",
      "Epoch 18:\t train loss: 34.547\t validation loss: 39.565\n",
      "Epoch 20:\t train loss: 34.229\t validation loss: 39.191\n",
      "Epoch 22:\t train loss: 33.837\t validation loss: 81.096\n",
      "Epoch 24:\t train loss: 33.530\t validation loss: 40.683\n",
      "Epoch 26:\t train loss: 33.269\t validation loss: 37.758\n",
      "Epoch 28:\t train loss: 33.004\t validation loss: 47.886\n",
      "Epoch 30:\t train loss: 32.742\t validation loss: 37.735\n",
      "Epoch 32:\t train loss: 32.618\t validation loss: 37.632\n",
      "Epoch 34:\t train loss: 32.269\t validation loss: 37.850\n",
      "Epoch 36:\t train loss: 32.281\t validation loss: 38.510\n",
      "Epoch 38:\t train loss: 32.096\t validation loss: 37.961\n",
      "Epoch 40:\t train loss: 31.665\t validation loss: 38.306\n",
      "Epoch 42:\t train loss: 31.625\t validation loss: 43.113\n",
      "Epoch 44:\t train loss: 31.457\t validation loss: 36.321\n",
      "Epoch 46:\t train loss: 31.128\t validation loss: 36.797\n",
      "Epoch 48:\t train loss: 30.916\t validation loss: 37.203\n",
      "Epoch 50:\t train loss: 30.891\t validation loss: 36.413\n",
      "Epoch 52:\t train loss: 30.864\t validation loss: 36.556\n",
      "Epoch 54:\t train loss: 30.555\t validation loss: 35.925\n",
      "Epoch 56:\t train loss: 30.423\t validation loss: 36.861\n",
      "Epoch 58:\t train loss: 30.283\t validation loss: 36.806\n",
      "Epoch 60:\t train loss: 30.137\t validation loss: 36.783\n",
      "Epoch 62:\t train loss: 30.034\t validation loss: 36.358\n",
      "Epoch 64:\t train loss: 29.937\t validation loss: 36.870\n",
      "Epoch 66:\t train loss: 29.784\t validation loss: 35.694\n",
      "Epoch 68:\t train loss: 29.718\t validation loss: 37.642\n",
      "Epoch 70:\t train loss: 29.530\t validation loss: 36.530\n",
      "Epoch 72:\t train loss: 29.434\t validation loss: 36.249\n",
      "Epoch 74:\t train loss: 29.427\t validation loss: 35.569\n",
      "Epoch 76:\t train loss: 29.170\t validation loss: 35.619\n",
      "Epoch 78:\t train loss: 29.165\t validation loss: 35.587\n",
      "Epoch 80:\t train loss: 29.006\t validation loss: 35.898\n",
      "Epoch 82:\t train loss: 29.123\t validation loss: 35.775\n",
      "Epoch 84:\t train loss: 29.044\t validation loss: 35.887\n",
      "Epoch 86:\t train loss: 28.905\t validation loss: 35.661\n",
      "Epoch 88:\t train loss: 28.829\t validation loss: 35.387\n",
      "Epoch 90:\t train loss: 28.780\t validation loss: 35.912\n",
      "Epoch 92:\t train loss: 28.729\t validation loss: 35.684\n",
      "Epoch 94:\t train loss: 28.732\t validation loss: 35.772\n",
      "Epoch 96:\t train loss: 28.726\t validation loss: 35.438\n",
      "Epoch 98:\t train loss: 28.587\t validation loss: 35.358\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "models/LinkPredictionNet.PnetL.ns.101e-CloseByPion200P.r15.e5.f400.pt\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 100:\t train loss: 28.649\t validation loss: 35.964\n"
     ]
    }
   ],
   "source": [
    "#%%script echo skipping\n",
    "# alpha - percentage of negative edges\n",
    "loss_func = QualityFocalLoss()\n",
    "\n",
    "model = model.to(device)\n",
    "optimizer = Adam(model.parameters(), lr=1e-3)\n",
    "scheduler = CosineAnnealingLR(optimizer, epochs, eta_min=1e-5)\n",
    "\n",
    "for epoch in range(epochs):\n",
    "\n",
    "    train_loss = train_link_prediction(\n",
    "        model,\n",
    "        device,\n",
    "        optimizer,\n",
    "        loss_func,\n",
    "        train_dl\n",
    "    )\n",
    "    \n",
    "    scheduler.step()\n",
    "\n",
    "    if epoch % 2 == 0:\n",
    "        test_loss = test_link_prediction(model, device, loss_func, test_dl)\n",
    "        print(\n",
    "            f\"Epoch {epoch}:\\t train loss: {train_loss:.3f}\\t validation loss: {test_loss:.3f}\",\n",
    "            file=sys.stderr\n",
    "        )\n",
    "        \n",
    "torch.save(model.state_dict(), model_path)\n",
    "print(model_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "0b9cbaa6",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "skipping\n"
     ]
    }
   ],
   "source": [
    "%%script echo skipping\n",
    "model.load_state_dict(torch.load(\n",
    "    model_path,\n",
    "    map_location=device\n",
    "))\n",
    "model = model.to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "c86c98db-0ac0-4a0c-84fd-4c50df507528",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.932794292424757\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TP: 11683, TN: 21609, FP: 1533, FN: 4503\n",
      "TH: 0.52 precision: 0.884 recall: 0.722 fbeta: 0.846 b_acc: 0.828\n"
     ]
    }
   ],
   "source": [
    "#%%script echo skipping\n",
    "print(roc_auc(model, device, test_dl))\n",
    "precision_recall_curve(model, device, test_dl, focus_metric=\"fbeta\", beta=0.5, step=3, truth_threshold=0.5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "eb8ce4e2",
   "metadata": {},
   "source": [
    "## Eval"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "fb3f12d6",
   "metadata": {},
   "outputs": [],
   "source": [
    "import uproot\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "from reco.data import get_event_data\n",
    "from reco.evaluation import model_evaluation\n",
    "from reco.dummy import DummyPleaser\n",
    "\n",
    "file_name = f\"{raw_dir}/new_ntuples_15643817_2.root\"\n",
    "cluster_data, trackster_data, simtrackster_data, assoc_data = get_event_data(file_name, pileup=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "6b5ffc46",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " --- Radius threshold: 3 ---\n",
      "Event 0:\n",
      "\tclue3d_to_sim:\tP: 0.836 R: 0.298 F: 0.614\n",
      "\ttarget_to_sim:\tP: 0.818 R: 0.347 F: 0.643\n",
      "\treco_to_sim:\tP: 0.825 R: 0.347 F: 0.647\n",
      "\t|S| = 1 |T| = 2 |R| = 2\n",
      "Event 1:\n",
      "\tNo data\n",
      "Event 2:\n",
      "\tclue3d_to_sim:\tP: 0.970 R: 0.215 F: 0.569\n",
      "\ttarget_to_sim:\tP: 0.971 R: 0.260 F: 0.627\n",
      "\treco_to_sim:\tP: 0.971 R: 0.260 F: 0.627\n",
      "\t|S| = 1 |T| = 2 |R| = 2\n",
      "Event 3:\n",
      "\tNo data\n",
      "Event 4:\n",
      "\tNo data\n",
      "Event 5:\n",
      "\tNo data\n",
      "Event 6:\n",
      "\tclue3d_to_sim:\tP: 0.971 R: 0.499 F: 0.816\n",
      "\ttarget_to_sim:\tP: 0.964 R: 0.524 F: 0.825\n",
      "\treco_to_sim:\tP: 0.971 R: 0.501 F: 0.818\n",
      "\t|S| = 1 |T| = 4 |R| = 4\n",
      "Event 7:\n",
      "\tclue3d_to_sim:\tP: 0.890 R: 0.022 F: 0.100\n",
      "\ttarget_to_sim:\tP: 0.902 R: 0.023 F: 0.105\n",
      "\treco_to_sim:\tP: 0.902 R: 0.023 F: 0.105\n",
      "\t|S| = 1 |T| = 2 |R| = 2\n",
      "Event 8:\n",
      "\tclue3d_to_sim:\tP: 0.881 R: 0.306 F: 0.640\n",
      "\ttarget_to_sim:\tP: 0.867 R: 0.408 F: 0.708\n",
      "\treco_to_sim:\tP: 0.881 R: 0.399 F: 0.710\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 9:\n",
      "\tclue3d_to_sim:\tP: 0.994 R: 0.095 F: 0.343\n",
      "\ttarget_to_sim:\tP: 0.995 R: 0.138 F: 0.444\n",
      "\treco_to_sim:\tP: 0.994 R: 0.095 F: 0.343\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 10:\n",
      "\tNo data\n",
      "Event 11:\n",
      "\tclue3d_to_sim:\tP: 0.984 R: 0.677 F: 0.903\n",
      "\ttarget_to_sim:\tP: 0.984 R: 0.721 F: 0.917\n",
      "\treco_to_sim:\tP: 0.984 R: 0.677 F: 0.903\n",
      "\t|S| = 1 |T| = 2 |R| = 3\n",
      "Event 12:\n",
      "\tclue3d_to_sim:\tP: 0.744 R: 0.001 F: 0.005\n",
      "\ttarget_to_sim:\tP: 0.794 R: 0.003 F: 0.016\n",
      "\treco_to_sim:\tP: 0.799 R: 0.003 F: 0.015\n",
      "\t|S| = 1 |T| = 2 |R| = 2\n",
      "Event 13:\n",
      "\tNo data\n",
      "Event 14:\n",
      "\tNo data\n",
      "Event 15:\n",
      "\tNo data\n",
      "Event 16:\n",
      "\tclue3d_to_sim:\tP: 0.986 R: 0.679 F: 0.904\n",
      "\ttarget_to_sim:\tP: 0.984 R: 0.701 F: 0.910\n",
      "\treco_to_sim:\tP: 0.984 R: 0.701 F: 0.910\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 17:\n",
      "\tNo data\n",
      "Event 18:\n",
      "\tNo data\n",
      "Event 19:\n",
      "\tclue3d_to_sim:\tP: 0.768 R: 0.153 F: 0.426\n",
      "\ttarget_to_sim:\tP: 0.777 R: 0.401 F: 0.654\n",
      "\treco_to_sim:\tP: 0.766 R: 0.299 F: 0.584\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "-----\n",
      "mean clue3d_to_sim:\tP: 0.451 R: 0.147 F: 0.266\n",
      "mean target_to_sim:\tP: 0.453 R: 0.176 F: 0.293\n",
      "mean reco_to_sim:\tP: 0.454 R: 0.165 F: 0.283\n",
      "mean n_tracksters:\tP: 0.500 R: 0.900 F: 0.950\n",
      " --- Radius threshold: 5 ---\n",
      "Event 0:\n",
      "\tclue3d_to_sim:\tP: 0.822 R: 0.307 F: 0.615\n",
      "\ttarget_to_sim:\tP: 0.799 R: 0.576 F: 0.742\n",
      "\treco_to_sim:\tP: 0.818 R: 0.314 F: 0.619\n",
      "\t|S| = 1 |T| = 2 |R| = 5\n",
      "Event 1:\n",
      "\tclue3d_to_sim:\tP: 0.999 R: 0.334 F: 0.715\n",
      "\ttarget_to_sim:\tP: 0.996 R: 0.348 F: 0.726\n",
      "\treco_to_sim:\tP: 0.996 R: 0.348 F: 0.726\n",
      "\t|S| = 1 |T| = 2 |R| = 2\n",
      "Event 2:\n",
      "\tclue3d_to_sim:\tP: 0.955 R: 0.221 F: 0.573\n",
      "\ttarget_to_sim:\tP: 0.940 R: 0.375 F: 0.723\n",
      "\treco_to_sim:\tP: 0.953 R: 0.357 F: 0.715\n",
      "\t|S| = 1 |T| = 2 |R| = 2\n",
      "Event 3:\n",
      "\tclue3d_to_sim:\tP: 0.741 R: 0.001 F: 0.007\n",
      "\ttarget_to_sim:\tP: 0.741 R: 0.001 F: 0.007\n",
      "\treco_to_sim:\tP: 0.741 R: 0.001 F: 0.007\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 4:\n",
      "\tclue3d_to_sim:\tP: 0.727 R: 0.595 F: 0.696\n",
      "\ttarget_to_sim:\tP: 0.727 R: 0.595 F: 0.696\n",
      "\treco_to_sim:\tP: 0.727 R: 0.595 F: 0.696\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 5:\n",
      "\tNo data\n",
      "Event 6:\n",
      "\tclue3d_to_sim:\tP: 0.971 R: 0.499 F: 0.816\n",
      "\ttarget_to_sim:\tP: 0.951 R: 0.656 F: 0.873\n",
      "\treco_to_sim:\tP: 0.959 R: 0.610 F: 0.861\n",
      "\t|S| = 1 |T| = 2 |R| = 2\n",
      "Event 7:\n",
      "\tclue3d_to_sim:\tP: 0.974 R: 0.353 F: 0.720\n",
      "\ttarget_to_sim:\tP: 0.966 R: 0.449 F: 0.785\n",
      "\treco_to_sim:\tP: 0.975 R: 0.445 F: 0.787\n",
      "\t|S| = 1 |T| = 2 |R| = 2\n",
      "Event 8:\n",
      "\tclue3d_to_sim:\tP: 0.881 R: 0.306 F: 0.640\n",
      "\ttarget_to_sim:\tP: 0.866 R: 0.434 F: 0.722\n",
      "\treco_to_sim:\tP: 0.881 R: 0.399 F: 0.710\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 9:\n",
      "\tclue3d_to_sim:\tP: 0.994 R: 0.095 F: 0.343\n",
      "\ttarget_to_sim:\tP: 0.994 R: 0.177 F: 0.517\n",
      "\treco_to_sim:\tP: 0.994 R: 0.177 F: 0.517\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 10:\n",
      "\tclue3d_to_sim:\tP: 0.976 R: 0.576 F: 0.857\n",
      "\ttarget_to_sim:\tP: 0.972 R: 0.634 F: 0.878\n",
      "\treco_to_sim:\tP: 0.972 R: 0.634 F: 0.878\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 11:\n",
      "\tclue3d_to_sim:\tP: 0.957 R: 0.679 F: 0.885\n",
      "\ttarget_to_sim:\tP: 0.950 R: 0.823 F: 0.921\n",
      "\treco_to_sim:\tP: 0.923 R: 0.840 F: 0.905\n",
      "\t|S| = 1 |T| = 2 |R| = 1\n",
      "Event 12:\n",
      "\tclue3d_to_sim:\tP: 0.927 R: 0.207 F: 0.546\n",
      "\ttarget_to_sim:\tP: 0.922 R: 0.574 F: 0.822\n",
      "\treco_to_sim:\tP: 0.924 R: 0.620 F: 0.842\n",
      "\t|S| = 1 |T| = 2 |R| = 1\n",
      "Event 13:\n",
      "\tNo data\n",
      "Event 14:\n",
      "\tclue3d_to_sim:\tP: 0.997 R: 0.718 F: 0.925\n",
      "\ttarget_to_sim:\tP: 0.997 R: 0.756 F: 0.937\n",
      "\treco_to_sim:\tP: 0.997 R: 0.756 F: 0.937\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 15:\n",
      "\tclue3d_to_sim:\tP: 0.993 R: 0.487 F: 0.822\n",
      "\ttarget_to_sim:\tP: 0.994 R: 0.617 F: 0.886\n",
      "\treco_to_sim:\tP: 0.994 R: 0.617 F: 0.886\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 16:\n",
      "\tclue3d_to_sim:\tP: 0.973 R: 0.680 F: 0.896\n",
      "\ttarget_to_sim:\tP: 0.969 R: 0.772 F: 0.922\n",
      "\treco_to_sim:\tP: 0.970 R: 0.727 F: 0.909\n",
      "\t|S| = 1 |T| = 2 |R| = 3\n",
      "Event 17:\n",
      "\tNo data\n",
      "Event 18:\n",
      "\tclue3d_to_sim:\tP: 0.901 R: 0.454 F: 0.753\n",
      "\ttarget_to_sim:\tP: 0.898 R: 0.572 F: 0.806\n",
      "\treco_to_sim:\tP: 0.901 R: 0.454 F: 0.753\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 19:\n",
      "\tclue3d_to_sim:\tP: 0.768 R: 0.153 F: 0.426\n",
      "\ttarget_to_sim:\tP: 0.735 R: 0.501 F: 0.672\n",
      "\treco_to_sim:\tP: 0.757 R: 0.305 F: 0.584\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "-----\n",
      "mean clue3d_to_sim:\tP: 0.778 R: 0.333 F: 0.562\n",
      "mean target_to_sim:\tP: 0.771 R: 0.443 F: 0.632\n",
      "mean reco_to_sim:\tP: 0.774 R: 0.410 F: 0.617\n",
      "mean n_tracksters:\tP: 0.850 R: 1.250 F: 1.350\n",
      " --- Radius threshold: 10 ---\n",
      "Event 0:\n",
      "\tclue3d_to_sim:\tP: 0.802 R: 0.307 F: 0.607\n",
      "\ttarget_to_sim:\tP: 0.758 R: 0.690 F: 0.744\n",
      "\treco_to_sim:\tP: 0.780 R: 0.546 F: 0.718\n",
      "\t|S| = 1 |T| = 2 |R| = 3\n",
      "Event 1:\n",
      "\tclue3d_to_sim:\tP: 0.993 R: 0.335 F: 0.713\n",
      "\ttarget_to_sim:\tP: 0.987 R: 0.775 F: 0.936\n",
      "\treco_to_sim:\tP: 0.987 R: 0.771 F: 0.935\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 2:\n",
      "\tclue3d_to_sim:\tP: 0.955 R: 0.221 F: 0.573\n",
      "\ttarget_to_sim:\tP: 0.936 R: 0.423 F: 0.754\n",
      "\treco_to_sim:\tP: 0.945 R: 0.413 F: 0.751\n",
      "\t|S| = 1 |T| = 2 |R| = 2\n",
      "Event 3:\n",
      "\tclue3d_to_sim:\tP: 0.976 R: 0.793 F: 0.933\n",
      "\ttarget_to_sim:\tP: 0.960 R: 0.823 F: 0.929\n",
      "\treco_to_sim:\tP: 0.976 R: 0.793 F: 0.933\n",
      "\t|S| = 1 |T| = 2 |R| = 2\n",
      "Event 4:\n",
      "\tclue3d_to_sim:\tP: 0.727 R: 0.595 F: 0.696\n",
      "\ttarget_to_sim:\tP: 0.727 R: 0.595 F: 0.696\n",
      "\treco_to_sim:\tP: 0.727 R: 0.595 F: 0.696\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 5:\n",
      "\tclue3d_to_sim:\tP: 1.000 R: 0.511 F: 0.839\n",
      "\ttarget_to_sim:\tP: 1.000 R: 0.538 F: 0.853\n",
      "\treco_to_sim:\tP: 1.000 R: 0.511 F: 0.839\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 6:\n",
      "\tclue3d_to_sim:\tP: 0.971 R: 0.500 F: 0.817\n",
      "\ttarget_to_sim:\tP: 0.934 R: 0.681 F: 0.870\n",
      "\treco_to_sim:\tP: 0.932 R: 0.656 F: 0.860\n",
      "\t|S| = 1 |T| = 2 |R| = 2\n",
      "Event 7:\n",
      "\tclue3d_to_sim:\tP: 0.974 R: 0.353 F: 0.721\n",
      "\ttarget_to_sim:\tP: 0.966 R: 0.835 F: 0.937\n",
      "\treco_to_sim:\tP: 0.975 R: 0.813 F: 0.938\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 8:\n",
      "\tclue3d_to_sim:\tP: 0.881 R: 0.306 F: 0.640\n",
      "\ttarget_to_sim:\tP: 0.816 R: 0.442 F: 0.698\n",
      "\treco_to_sim:\tP: 0.881 R: 0.399 F: 0.710\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 9:\n",
      "\tclue3d_to_sim:\tP: 0.994 R: 0.095 F: 0.343\n",
      "\ttarget_to_sim:\tP: 0.996 R: 0.371 F: 0.745\n",
      "\treco_to_sim:\tP: 0.996 R: 0.371 F: 0.745\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 10:\n",
      "\tclue3d_to_sim:\tP: 0.976 R: 0.576 F: 0.857\n",
      "\ttarget_to_sim:\tP: 0.972 R: 0.660 F: 0.888\n",
      "\treco_to_sim:\tP: 0.972 R: 0.660 F: 0.888\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 11:\n",
      "\tclue3d_to_sim:\tP: 0.957 R: 0.679 F: 0.885\n",
      "\ttarget_to_sim:\tP: 0.943 R: 0.888 F: 0.932\n",
      "\treco_to_sim:\tP: 0.918 R: 0.880 F: 0.910\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 12:\n",
      "\tclue3d_to_sim:\tP: 0.928 R: 0.207 F: 0.547\n",
      "\ttarget_to_sim:\tP: 0.875 R: 0.634 F: 0.813\n",
      "\treco_to_sim:\tP: 0.912 R: 0.663 F: 0.848\n",
      "\t|S| = 1 |T| = 2 |R| = 1\n",
      "Event 13:\n",
      "\tNo data\n",
      "Event 14:\n",
      "\tclue3d_to_sim:\tP: 0.997 R: 0.718 F: 0.925\n",
      "\ttarget_to_sim:\tP: 0.992 R: 0.862 F: 0.963\n",
      "\treco_to_sim:\tP: 0.992 R: 0.862 F: 0.963\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 15:\n",
      "\tclue3d_to_sim:\tP: 0.993 R: 0.487 F: 0.822\n",
      "\ttarget_to_sim:\tP: 0.993 R: 0.727 F: 0.926\n",
      "\treco_to_sim:\tP: 0.993 R: 0.727 F: 0.926\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 16:\n",
      "\tclue3d_to_sim:\tP: 0.973 R: 0.680 F: 0.896\n",
      "\ttarget_to_sim:\tP: 0.956 R: 0.832 F: 0.929\n",
      "\treco_to_sim:\tP: 0.955 R: 0.824 F: 0.925\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 17:\n",
      "\tNo data\n",
      "Event 18:\n",
      "\tclue3d_to_sim:\tP: 0.901 R: 0.454 F: 0.753\n",
      "\ttarget_to_sim:\tP: 0.800 R: 0.680 F: 0.773\n",
      "\treco_to_sim:\tP: 0.789 R: 0.702 F: 0.770\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 19:\n",
      "\tclue3d_to_sim:\tP: 0.779 R: 0.160 F: 0.439\n",
      "\ttarget_to_sim:\tP: 0.687 R: 0.647 F: 0.678\n",
      "\treco_to_sim:\tP: 0.769 R: 0.416 F: 0.657\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "-----\n",
      "mean clue3d_to_sim:\tP: 0.839 R: 0.399 F: 0.650\n",
      "mean target_to_sim:\tP: 0.815 R: 0.605 F: 0.753\n",
      "mean reco_to_sim:\tP: 0.825 R: 0.580 F: 0.751\n",
      "mean n_tracksters:\tP: 0.900 R: 1.150 F: 1.150\n",
      " --- Radius threshold: 15 ---\n",
      "Event 0:\n",
      "\tclue3d_to_sim:\tP: 0.802 R: 0.307 F: 0.607\n",
      "\ttarget_to_sim:\tP: 0.718 R: 0.727 F: 0.720\n",
      "\treco_to_sim:\tP: 0.761 R: 0.590 F: 0.719\n",
      "\t|S| = 1 |T| = 2 |R| = 2\n",
      "Event 1:\n",
      "\tclue3d_to_sim:\tP: 0.993 R: 0.335 F: 0.713\n",
      "\ttarget_to_sim:\tP: 0.992 R: 0.797 F: 0.946\n",
      "\treco_to_sim:\tP: 0.992 R: 0.797 F: 0.946\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 2:\n",
      "\tclue3d_to_sim:\tP: 0.955 R: 0.221 F: 0.573\n",
      "\ttarget_to_sim:\tP: 0.925 R: 0.430 F: 0.752\n",
      "\treco_to_sim:\tP: 0.939 R: 0.417 F: 0.751\n",
      "\t|S| = 1 |T| = 2 |R| = 2\n",
      "Event 3:\n",
      "\tclue3d_to_sim:\tP: 0.976 R: 0.793 F: 0.933\n",
      "\ttarget_to_sim:\tP: 0.960 R: 0.823 F: 0.929\n",
      "\treco_to_sim:\tP: 0.976 R: 0.793 F: 0.933\n",
      "\t|S| = 1 |T| = 2 |R| = 2\n",
      "Event 4:\n",
      "\tclue3d_to_sim:\tP: 0.727 R: 0.595 F: 0.696\n",
      "\ttarget_to_sim:\tP: 0.711 R: 0.633 F: 0.694\n",
      "\treco_to_sim:\tP: 0.727 R: 0.595 F: 0.696\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 5:\n",
      "\tclue3d_to_sim:\tP: 1.000 R: 0.511 F: 0.839\n",
      "\ttarget_to_sim:\tP: 1.000 R: 0.566 F: 0.867\n",
      "\treco_to_sim:\tP: 1.000 R: 0.566 F: 0.867\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 6:\n",
      "\tclue3d_to_sim:\tP: 0.971 R: 0.500 F: 0.817\n",
      "\ttarget_to_sim:\tP: 0.926 R: 0.691 F: 0.867\n",
      "\treco_to_sim:\tP: 0.896 R: 0.688 F: 0.845\n",
      "\t|S| = 1 |T| = 2 |R| = 2\n",
      "Event 7:\n",
      "\tclue3d_to_sim:\tP: 0.974 R: 0.353 F: 0.721\n",
      "\ttarget_to_sim:\tP: 0.967 R: 0.835 F: 0.938\n",
      "\treco_to_sim:\tP: 0.974 R: 0.818 F: 0.938\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 8:\n",
      "\tclue3d_to_sim:\tP: 0.881 R: 0.306 F: 0.640\n",
      "\ttarget_to_sim:\tP: 0.816 R: 0.442 F: 0.698\n",
      "\treco_to_sim:\tP: 0.881 R: 0.399 F: 0.710\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 9:\n",
      "\tclue3d_to_sim:\tP: 0.994 R: 0.095 F: 0.343\n",
      "\ttarget_to_sim:\tP: 0.996 R: 0.371 F: 0.745\n",
      "\treco_to_sim:\tP: 0.996 R: 0.371 F: 0.745\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 10:\n",
      "\tclue3d_to_sim:\tP: 0.976 R: 0.576 F: 0.857\n",
      "\ttarget_to_sim:\tP: 0.972 R: 0.660 F: 0.888\n",
      "\treco_to_sim:\tP: 0.972 R: 0.660 F: 0.888\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 11:\n",
      "\tclue3d_to_sim:\tP: 0.957 R: 0.679 F: 0.885\n",
      "\ttarget_to_sim:\tP: 0.943 R: 0.888 F: 0.932\n",
      "\treco_to_sim:\tP: 0.901 R: 0.885 F: 0.898\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 12:\n",
      "\tclue3d_to_sim:\tP: 0.928 R: 0.207 F: 0.547\n",
      "\ttarget_to_sim:\tP: 0.871 R: 0.637 F: 0.811\n",
      "\treco_to_sim:\tP: 0.877 R: 0.686 F: 0.831\n",
      "\t|S| = 1 |T| = 2 |R| = 1\n",
      "Event 13:\n",
      "\tclue3d_to_sim:\tP: 0.999 R: 0.761 F: 0.940\n",
      "\ttarget_to_sim:\tP: 1.000 R: 0.834 F: 0.962\n",
      "\treco_to_sim:\tP: 1.000 R: 0.834 F: 0.962\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 14:\n",
      "\tclue3d_to_sim:\tP: 0.997 R: 0.718 F: 0.925\n",
      "\ttarget_to_sim:\tP: 0.995 R: 0.902 F: 0.975\n",
      "\treco_to_sim:\tP: 0.995 R: 0.902 F: 0.975\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 15:\n",
      "\tclue3d_to_sim:\tP: 0.993 R: 0.489 F: 0.823\n",
      "\ttarget_to_sim:\tP: 0.993 R: 0.731 F: 0.927\n",
      "\treco_to_sim:\tP: 0.961 R: 0.737 F: 0.906\n",
      "\t|S| = 1 |T| = 2 |R| = 2\n",
      "Event 16:\n",
      "\tclue3d_to_sim:\tP: 0.973 R: 0.680 F: 0.896\n",
      "\ttarget_to_sim:\tP: 0.956 R: 0.832 F: 0.929\n",
      "\treco_to_sim:\tP: 0.956 R: 0.826 F: 0.927\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 17:\n",
      "\tNo data\n",
      "Event 18:\n",
      "\tclue3d_to_sim:\tP: 0.901 R: 0.454 F: 0.753\n",
      "\ttarget_to_sim:\tP: 0.800 R: 0.680 F: 0.773\n",
      "\treco_to_sim:\tP: 0.651 R: 0.803 F: 0.677\n",
      "\t|S| = 1 |T| = 1 |R| = 1\n",
      "Event 19:\n",
      "\tclue3d_to_sim:\tP: 0.782 R: 0.161 F: 0.441\n",
      "\ttarget_to_sim:\tP: 0.673 R: 0.760 F: 0.689\n",
      "\treco_to_sim:\tP: 0.761 R: 0.488 F: 0.685\n",
      "\t|S| = 1 |T| = 1 |R| = 2\n",
      "-----\n",
      "mean clue3d_to_sim:\tP: 0.889 R: 0.437 F: 0.697\n",
      "mean target_to_sim:\tP: 0.861 R: 0.662 F: 0.802\n",
      "mean reco_to_sim:\tP: 0.861 R: 0.643 F: 0.795\n",
      "mean n_tracksters:\tP: 0.950 R: 1.250 F: 1.250\n"
     ]
    }
   ],
   "source": [
    "r_ranges = [3, 5, 10, 15]\n",
    "results = []\n",
    "\n",
    "max_events = 20\n",
    "clue3D_F = []\n",
    "target_F = []\n",
    "model_reco_F = []\n",
    "for r in r_ranges:\n",
    "    print(f\" --- Radius threshold: {r} ---\")\n",
    "    result = model_evaluation(\n",
    "        cluster_data,\n",
    "        trackster_data,\n",
    "        simtrackster_data,\n",
    "        assoc_data,\n",
    "        model.to(\"cpu\"),\n",
    "        decision_th=0.5,\n",
    "        radius=r,\n",
    "        max_events=max_events,\n",
    "        bigT_e_th=5,\n",
    "        pileup=True,\n",
    "        graph=True,\n",
    "        link_prediction=True,\n",
    "    )\n",
    "    clue3D_F.append(np.sum(np.array(result[\"clue3d_to_sim\"])[:,2]) / max_events)\n",
    "    target_F.append(np.sum(np.array(result[\"target_to_sim\"])[:,2]) / max_events)\n",
    "    model_reco_F.append(np.sum(np.array(result[\"reco_to_sim\"])[:,2]) / max_events)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "0992efe0-fa35-494e-aeb8-860e060a531e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CLUE: [0.2660786195101392, 0.5618487610549607, 0.6503284712982198, 0.6974920132483364]\n",
      "Target: [0.2925245062036259, 0.631796889255891, 0.7530917443918212, 0.8020145508567055]\n",
      "Reco: [0.28307734475674334, 0.616571273678623, 0.7505698827151652, 0.7948727273697329]\n"
     ]
    }
   ],
   "source": [
    "print(\"CLUE:\", clue3D_F)\n",
    "print(\"Target:\", target_F)\n",
    "print(\"Reco:\", model_reco_F)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "bc9b49b1",
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'naive_reco_F' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "Cell \u001b[0;32mIn [35], line 9\u001b[0m\n\u001b[1;32m      6\u001b[0m ax\u001b[38;5;241m.\u001b[39mplot(r_ranges[:\u001b[38;5;241m6\u001b[39m], (np\u001b[38;5;241m.\u001b[39marray(model_reco_F) \u001b[38;5;241m-\u001b[39m np\u001b[38;5;241m.\u001b[39marray(clue3D_F))[:\u001b[38;5;241m6\u001b[39m], \u001b[38;5;124m'\u001b[39m\u001b[38;5;124m-o\u001b[39m\u001b[38;5;124m'\u001b[39m, label\u001b[38;5;241m=\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mGNN LP $\u001b[39m\u001b[38;5;124m\\\u001b[39m\u001b[38;5;124mDelta F_\u001b[39m\u001b[38;5;132;01m{0.5}\u001b[39;00m\u001b[38;5;124m$\u001b[39m\u001b[38;5;124m\"\u001b[39m, c\u001b[38;5;241m=\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m#56B4E9\u001b[39m\u001b[38;5;124m\"\u001b[39m)\n\u001b[1;32m      7\u001b[0m \u001b[38;5;66;03m# ax.plot(r_ranges[:6], (np.array(naive_reco_F) - np.array(clue3D_F))[:6], '-v', label=\"naive $\\Delta F_{0.5}$\", c=\"#E69F00\")\u001b[39;00m\n\u001b[0;32m----> 9\u001b[0m ax\u001b[38;5;241m.\u001b[39maxhline(\u001b[38;5;28mmax\u001b[39m(np\u001b[38;5;241m.\u001b[39marray(naive_reco_F) \u001b[38;5;241m-\u001b[39m np\u001b[38;5;241m.\u001b[39marray(clue3D_F)), label\u001b[38;5;241m=\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mbaseline\u001b[39m\u001b[38;5;124m\"\u001b[39m, c\u001b[38;5;241m=\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mlightgray\u001b[39m\u001b[38;5;124m\"\u001b[39m, linestyle\u001b[38;5;241m=\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m--\u001b[39m\u001b[38;5;124m\"\u001b[39m)\n\u001b[1;32m     11\u001b[0m ax\u001b[38;5;241m.\u001b[39mlegend()\n\u001b[1;32m     12\u001b[0m ax\u001b[38;5;241m.\u001b[39mset_xlabel(\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mNeighborhood radius (cm)\u001b[39m\u001b[38;5;124m\"\u001b[39m)\n",
      "\u001b[0;31mNameError\u001b[0m: name 'naive_reco_F' is not defined"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAqgAAAGsCAYAAAARwVXXAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy89olMNAAAACXBIWXMAAA9hAAAPYQGoP6dpAABbVklEQVR4nO3deXhU5f3+8ffsk2UmIQlJCCTssu8gglas0uJSFTeQWqXUpYt1w1px/7ZWEStIXX5Sba3dFJcqdSsu1K2VihIQkE2UHbIRkkkmyWzn/P4IBCIBEkhmJpP7dV25rsmZZ87zmTEmN88553MspmmaiIiIiIjECWusCxAREREROZgCqoiIiIjEFQVUEREREYkrCqgiIiIiElcUUEVEREQkriigioiIiEhcUUAVERERkbhij3UBrcEwDHbt2oXH48FiscS6HBERERH5BtM0qaqqIi8vD6v1yGukCRFQd+3aRX5+fqzLEBEREZGj2L59O926dTvimIQIqB6PB6h/w16vN8bViIiIiMg3+Xw+8vPzG3LbkSREQN1/WN/r9SqgioiIiMSx5pyOqYukRERERCSuKKCKiIiISFxRQBURERGRuKKAKiIiIiJxRQFVREREROKKAqqIiIiIxBUFVBERERGJKwqoIiIiIhJXFFBFREREJK4ooIqIiIhIXEmIW52KiIiIyNGF9mwjXFV22OftniwcmQVRrOgwdcS6ABERERFpe6E929h0az/MUN1hx1gcbvrM2RDzkKpD/CIiIiIdQLiq7IjhFMAM1R1xhTVaFFBFREREEly4qozQ3l2xLqPZdIhfREREpA2FKnZjBmsxQ3UYoTrMcKD+cbAOqzOJlAGnNYyt+PBPhKtKMIN1GPvGmfteY0vJIOfS3zaM3fnkdAK71u4bE6jf9/6xns70/e2mhrHbH/4etV99Es23fVwUUEVERCQhmaZJuHw7Rqhx0Nsf5GypmST3Gdcwfs9b8zGCNQ2B7+BA6czpS+fz7mgYu/W3k4hUle7b14H9G6E63N1H0PPO/zSM3fx/Ywjv3dlkja5ug+l93+qG78venENw94YmxzqyejQKqIGdX1C3ZXmTYy12V+Pvnclgd0E4cIRPLH4ooIqIiEirME0Ti8VS/zgSJlS+40BwO3g1MBTAnplPUvcRABiBGsqXPN7kSqAZqiOpzzgyzvhZ/dhgLVvun9Do+YNXJT3Dz6Pbz19oqOnLmd0PW2/q0LMouPnNhu9LXroDM1jT5NikE05pFFDrtn9OpLK46c/hG/uwulOxulOx2F1YHG4sDjfWfY+dOX0ajfWMnEy4shjrvnEWR/04q8ONLTWz0dicSx/CCPj3PX9g3xa7C6szqdHYHrP+Te2WQjbfM+qwn0c8UUAVEZHj1l5a1yQq0zTBNLFY6y8tMYJ1hPfuPCTo7V/tc+X1x5U3AIBwZTEVH/3pkJXA/SuNnuHnkjZuGgDBsq3seOTCRquQB+8744yfk3vZw/v2W8SmX/Q8bM3pE64m6UdP1tcfDlLy/C8P//4i4YaAarHaqdv86WHHGgddBGSxWLAmeesf7wt5B4dEZ07fxjWdfDlmJNwQDK0HBT5HVuOg2/XqP9d/5t8IkRa7C6srpdHYPg+sP2y935Qz5YFmjz341IBEo4AqIiLHpT21rmkLh4TDumpCFbswg3WHrO6ZoTrcPUfj7Fwf3ILFm6j85PnGQS944HH6qT8idcgkAGq/WsauZ65pfDj5oP3mTP0tmWfdDEDdtpVsuXdc0wUDnS/4FZ0n3w1AuKqUkhdvO+xYR0a3hoCKaVC3tfDwn8VBh48tDjcWZ/IhQW//Y0fnHgfGOpNIO/mKAyHP0Xil0Zk38MAkNjv5N7564Pn9odNZ/9jq9jSqqf+CysPW+01dfrig2WP3/3eRtqGAKiIix6UlrWtaO6CaRgSwNITDiH8v4Yrdh1yIsv9xSv/TsKfnAlC7pZDqla83Xg086LzDrO/NIqnnaACqVr5OyYu3Nxk4zXCQrj97nrSxUwCoXvUvdjw+5bA1d7nyj40Cauk/7jzs2KS+JzcEISMcILDt88OONQ4Kh1ZnEla350DQszcOifb0vIax9tQs0r81o/HhYYcbi7P+8f7PAMCelkv+zDexOg+sRB68Krl/tRLqV80HPOU/bL0HszpcdL3mz80aa7FY8Iw4t1ljpTG7JwuLw33Uf0zaPVlRrKppCqgiIhIVRqC64bF/3Xv4173fKOg1XMgSDpBz6UM4s3sBsPf9P1D+1sONDinvf4wRoced/yW573gAKj56huLnZh62hoJfvEXqvoBat6WQ0lfuOezYtHHfbwhnRm0VgR2rDzu20cqhKwVrclqTh5Stdhf21AN//B1Z3UmfcNWhIXLfqmNyv1Mbxrq7DabgF28dcjh5/+ODw6G7YBj9f+87bL0Hs6fnknfV080aa3Um4Rl2VrPGSvxxZBbQZ86GdnE6jgKqiIhERcRX2vDYv+59yv7568OOzTpnFuwLqJGavQR2rT3s2IPDodXtwZaS0fjw78GHiw8Kca6uA+n07R83hMNvjnfnD2sYmzLg2xT88t1DL0RpYr+eYWfT/4mKZn0mrrwB5P3oqWaNtaV0InXId5s1VuRwHJkFcRFAj0YBVUREosN64E9OUq8T6XTGz5oMhlaHu9EFKd4Tp5DUc/QhVz/vX2m0JaU1jO102lV0Ou2qZpWT3Hd8w8rr0djTcxtWXkWk7SmgiojIMTNNs9ljHZn5DY89w8/BM/ycZr3OmdUdZ9bhWwWJSOLRrU5FRKTFjLpqSl6++4gXA4mIHCutoIqISLOZRoSKj56h9B93Eq4sAiB16NkxrkpEEo0CqoiINEv1mncoXvgLAttXAeDI7k3OlDm4950f2h5a14hI+6CAKiIiRxSuLGbXH2ZQvepfAFiT0+l8/t10OuNnWB319/ve37rGMGFDjYvKsJU0u0G/5ABWS/y0rhGR9kEBVUREjsianE5g9waw2ck441qyzr8L+zfuCe7ILKAwmM0za6sprzMatme4rfxwYCpjM93RLltE2rFjukjq8ccfp0ePHrjdbsaOHcuyZcsOO/aLL77goosuokePHlgsFubPn3/c+xQRkbZjBGspf+cxzHAI2H+Xn7/Q+/615F42/5BwCvDJ7jrmFfoahVOA8jqDeYU+Ptl95DtNiYgcrMUB9fnnn2fmzJncc889FBYWMmzYMCZNmkRJSUmT42tqaujVqxcPPPAAublN95Br6T5FRKT1mYZBxX//xqZb+1H0t+vY+/6TDc8ln3Ayrty+Tb7OME2eWVvd5HP7/XltNUYLWlKJSMdmMVvSxA4YO3YsY8aM4bHHHgPAMAzy8/O57rrrmDVr1hFf26NHD2688UZuvPHGVtsngM/nIy0tjcrKSrxe71HHi4hIY/71H1K88GbqNn8GgD0jn9zL5uMdfeFRX/vFniC//l/FUcelOCx4HFbcdgtuu4Ukm6XhsdtmIemgxwe2WQ8dY7PgsFmO9y2LSJS1JK+16BzUYDDI8uXLue222xq2Wa1WJk6cyNKlS4+p2GPZZyAQIBA4cGs7n6959xsWEZHGAkVfUvL8L6kqXATU3yo069zbyfjuDVidSc3aR8U3Dusfjj9k4g9FjrXURmwW6kPuIYF2/+PmB+EkuwWXzYLVotArEi9aFFDLysqIRCLk5OQ02p6Tk8P69euPqYBj2efs2bP51a9+dUzziYjIAUV//Tn+NW+D1Uan066h8wX/h92b3aJ9pLubd7bYNYNT6eqxUxs2qYuY1IX3fUXM+m0HbT8wxqj/ft/20L4sHDH3B97WO23AZWs60DYv5FoPBOV94+1WsCj0ihyTdnkV/2233cbMmTMbvvf5fOTn5x/hFSIiAmCEAhAJYXWnApA95QFK7U5ypszB1XVgi/cXipj8d2ftUcdluq18uyDpuFcpw4ZJ4IiBdv9245Btjb4/aB/7I24gUr/vyuOq8ID9q7wHr9Q25/SFhpD7jfFuu1Z5peNoUUDNysrCZrNRXFzcaHtxcfFhL4Bqi326XC5cLtcxzSci0hGZpknVpy9R/MIsPCPOI/eyhwFI6j6CgpteO6Z9ltZEmFdYydeV4aOOnT4wtVXCld1qwW61kOI47l0B9Z9L0GBfgDVasLL7jef3vba2zVd52RdWrYdf2T1syLUesvLr0CqvxKkWBVSn08moUaNYsmQJkydPBuovaFqyZAk///nPj6mAttiniIgcULPpfxQ/dzO1mz4GoKrwFbIvmY3Veey9SVeWBnh0hY/qkEmqw8J1w70EIuYhfVAz3VamD0xlbJf47INqsVhw2eoP76e5jqnz4iEiB6/ytuD0hW+G3v37qG20ylu/0lsZbMVzeZtx+oLroO37Q3GjbVrllVbW4kP8M2fOZPr06YwePZoTTzyR+fPn4/f7mTFjBgBXXHEFXbt2Zfbs2UD9RVBr165teLxz505WrlxJamoqffr0adY+RUSk5YKlWyh58TZ8nywEwOJMJvPsW8g66xfHHE4N0+TlTTW8tNGPCfRKszNzZBqdk20AjMl1sa48REWdQbrbyoAMR4cLLDarhWSrheRWXOUN7V/lbQi5RrNOXzg4KDcE4nD9qjHsW+UNm/jDrbfK67RyIOTuOzc36ZuB1mbB9c0V3m+c+pCkVd4OrcUBderUqZSWlnL33XdTVFTE8OHDWbx4ccNFTtu2bcNqPfCv0F27djFixIiG7x966CEeeughJkyYwPvvv9+sfYqISMv4li9i5xOXYoYCYLGQfsoP6XzhvTgyuh7zPquDBo+u9LGyNAjAGQVufjjQg/Oglk9Wi4VBmc7jrl8OsFgsOG3gtFlorUaKhnlomK0NH2bl9+AwfJhzfmsjJsa+jBs0IBg0qQyaQPM6PByJdd8q7yEh9+DvbYcG4qbO/93/GptVgdcwzbj+x2SL+6DGI/VBFRFpLFxZzKZf9iGp11hyLn0Id/fhx7W/rytDzFteSWmtgcMKVw32cFp+89pQSeI7eJW3UeiNGASavIit6ZB7cCAOtM5ZDE1yWjko5B56bu6RQu6B83sPhOH2tsr7ye66w9+WuA1Px2lJXlNAFRFp50zTpHrl61SveZsulz/asD1YuhlHVo/j+sNpmib/3l7H019UETYgJ9nGzFFeenhb6fi1yGEcvMrbdKeGJkLuN1Z7D6wCN17lbW3Wg8/lbSrkNrqI7aAevUcIxW21yrv/tsSHM3Okt81Caps16hcRkfhSu3UFxc/dTM269wDwjDiP1MHfAcDZuedx7TsYMfnjmire31EHwKhsJ9cO95LiaJ2LiUSOxGqxkOxog3N5Dwm5xlFPXzhcKA5E6hOvYUJN2KSmFc/ldexf5W3m6QuHuzGF216/Quy0ggnNui3xmFxXzA/3K6CKiLRDofIdlPzjTir/+xcwTSwOFxnfvYmkXie2yv6LayLMW17JFl8YC3BpvxTO650c8z9aIseq0bm8rXSatGE2ry9v45BrHDL+4NXefZmXkAGhoElVQw+H4zvnwUJ96A0e5bTgPXUG68pDMT+XXAFVRKQdMYK1lL0+mz3/eggzWN8g3zvu+2RffD/OrO6tMsfy4gCPr/ThD5t4nRauH5HGkCxd+CTyTVZL/QpmUiumqVCk5acvNATjJlZ796/ymhw9nO7X3NsXtyUFVBGR9sRipfLjv2MGa0nqezK50+aR1Lt1Vk0N0+SFDX5e+aoGgL7pdm4amUZmkq1V9i8iR+ewWXDYLHhaeZW3LmyyuizI459XHfU1zb19cVtSQBURiXP+de+TfMIpWGx2rA4XuZc/hhmswTP6wla7ctgXMHhkZSWry0IAnNkjicsHpGJXOx6Rdu3gVd5Turp5boO/0dX735S5r+VUrMU+IouISJPqdnzB1ofOYusD32bv+081bPcMOwvvmItaLZxu3Bti1n/KWV0WwmWD64d7mTHIo3AqkmCsFgs/HJh6xDGtdVvi46UVVBGROBOuLKbk5bup+OAPYBpgcxCpLmv1eUzT5O2ttfx5bTURE7qk2Lh5VBr5Hv1pEElUY7u4mTmSuL8tsX4LiYjECSNQw563HmbPGw9g1NW3gvGMvpCcKXNw5vRp1bnqwiZPrfbxn10BAMbmuvjJUA/JaiElkvDGdnHH/W2JFVBFROLErj9eie+ThQC4e44hZ9pcUvp9q/XnqQ4zr7CS7VURrBa4rH8q5/RMald3whGR4xPvtyVWQBURiSHTiGCx1l8ln3n2LdR+tZTsi+7De9I0LNbWX838ZHcdT6yqojZsku6ycuNILwMy4vePlIh0TAqoIiIxECjaSMnzv8SR1ZPcyx4GIKnHSPo8uAmLrfV/NUcMk+c2+Hnt6/oWUv0zHNw4wksnt1pIiUj8UUAVEYmicPUeyhb9ivJ/PwGRMBZnElnn34k9NROgTcJpRV2E+St8rCuvbyF1bq9kLu2Xoqv0RSRuKaCKiESBEQpQ/s6jlL32G4yaSgBSh51DztQHG8JpW1hfHmR+oY+9AYMku4WfDvXEzVW6IiKHo4AqItLGar/+lB3/byqh0s0AuPKHkjNtLqmDJrbZnKZp8ubmWv62vhrDhG6p9S2k8lL1a19E4p9+U4mItDFHRj6RqlLs6V3ofNFvSD9lesOFUW2hNmywYFUV/9td30Lq5DwX1wzx4LarhZSItA8KqCIirSxYuhnfpy+RdfYtANjTc8m/6Q2SeozE6j7yXVyO146qMHOXV7LLH8FmgSsGpjKpu1pIiUj7ooAqItJKIv4Kyl6/n/K3f4cZDuLuPpLUQWcAkNL/1Daf/7+76vj9qioCEZMMt5WbRqZxQqfY31NbRKSlFFBFRI6TGQ6x973fU7ro/4hU7wEgZeAZ2NNyojJ/2DD527pq/rWlFoDBmQ6uH5FGmkuH9EWkfVJAFRE5RqZpUr3iNYpf+CXB3RsAcOYNIGfqb0kddnZUDquX10V4uNDHxr31LaQu6J3MlH4pcXXLQhGRllJAFRE5VpEQRX+/gVDZFmyeznS+8Nd0mnBVm/QybcqasiCPrKikMmiSbLdw7XAvo3NcUZlbRKQtKaCKiLRAqHwn9rQcLDY7FruTnGlzqf36U7LOvQ1bkjcqNZimyatf1/Dcej8m0N1rZ+ZIL7kp+pUuIolBv81ERJrBqKum7I057Fk8l5xp88g4/ScAeEdfiHf0hVGroyZk8PjnPj4rDgIwoZubKwd7cNl0SF9EEocCqojIEZhGhIoP/0TJy3cSqSwGwL/m7YaAGk1bfWHmLa+kqCaC3QozBnk4I9+tFlIiknAUUEVEDqN69dsUL/wFgR2rAXBk9yZnyhw8UVwx3e/DHbU8tbqKoAGdk+pbSPVOVwspEUlMCqgiIk0oeekOyl67HwBrSic6n3cXGROvxWJ3RrWOUMTkz2ureWdbfQupYZ2dXDfci8epFlIikrgUUEVEmuAZdSF7Fs+l0+k/o/N5d2JLzYh6DWW1EeYtr+SryjAW4KK+KVzUN1ktpEQk4SmgikiHZwRq2PPWw5jhANkX/hqApJ6j6DtvO3Zv55jUtLI0wGMrfFSFTFIdFq4b7mV4tlpIiUjHoIAqIh2WaRhULv07JS/dTrh8B9gcpH/rRzg79wCISTg1TJOXN9Xw0sb6FlK90uzcNDKN7GRb1GsREYkVBVQR6ZD86z+g+LmbqduyHABHZgHZlzyAI7MgZjVVBw0eXeljZWl9C6kzCtz8cKAHp1pIiUgHo4AqIh1KaM92iv52HVWF/wTA6vaQde7tZHz3BqzOpJjV9XVliHnLKymtNXBY4arBHk7Lj109IiKxpIAqIh2Kxeag+ot3wWqj02nX0PmC/8PuzY5pTf/eVsvTX1QRMiAn2crMkWn0SFMLKRHpuBRQRSShGaEA1StfxzvmIgDs6bnkXfk07vwhuPIGxLS2YMTk6TVVvLejDoBR2U6uHe4lxaEWUiLSsSmgikhCMk0T37IXKXlxFqHSzXSf9W9SBnwbgLSxU2JcHRTX1LeQ2uKrbyE1tV8K5/dWCykREVBAFZEEVLNpKcXP3UztpqUA2NO7YNRVx7iqA5YXB3h8pQ9/2MTrtHD9iDSGZEX3BgAiIvFMAVVEEkawdDMlL8zCt+wFACzOZLLO/iWZZ/8CqyslxtXVt5B6YaOfVzbVANA3vb6FVGaSWkiJiBxMAVVEEoJpmmybezbB3evBYiH9WzPofOG9ODrlxbo0AHwBg0dWVrK6LATAmT2SuHxAKnarDumLiHyTAqqItFtmOAQWCxabHYvFQucL/o+K958iZ9pc3AXDYl1egy/3hni4sJI9dQYuG1wzxMspXd2xLktEJG4poIpIu2OaJtUrXqX4+V+SeeZMOn37xwB4T5yC98QpWOLkQiPTNHl7ay1/XltNxIQuKTZmjkqjwKNfvSIiR6LfkiLSrtRuKaT4uZupWf8+AOXvPkb6addgsVjiJpgC1IVN/rDGx0c7AwCcmOvip0M9JKuFlIjIUSmgiki7ECrfQclLd1D58V/BNLE4XGR89yayzr0troIpwK7qMPMKK9leFcFqge/3T+V7PZPirk4RkXilgCoica/iP39m959/ihmsBcA77vtkX3w/zqzuMa7sUMuKAvy/z33Uhk3SXVZuGOFlYKZaSImItIQCqojEPVfXQZjBWpJOOIXcS+eS1PvEWJd0iIhh8twGP699Xd9Cqn8nBzeO9NLJrRZSIiItpYAqInGnevVbBHZvIPO71wOQ1HM0Pe9Zhrvn6Lg8TF5RF+F3K3ysLa9vIfW9nklM668WUiIix0oBVUTiRt2ONRQv/AX+1W9hsTvxDP8ezuxeACT1GhPj6pq2vjzI/EIfewMGbpuFnw7zcFIXtZASETkeCqgiEnPhiiJKXr6big//CKYBNgedzrgWW0qnWJd2WKZp8uaWWv6+rr6FVLdUGzePSiMvVb9WRUSOl36TikjMGIEa9iyex54352DUVQPgGX0ROVMewJnTJ8bVHV5t2OD3q6pYuru+hdT4PBc/HuLBbVcLKRGR1qCAKiIxE6mpoOyNBzADfty9TiR32lySTzgl1mUd0Y6qMHOXV7LLH8FmgcsHpHJmD7WQEhFpTQqoIhJVddtW4S4YCoCjUx45U3+LLTkd79ipWKzxvQL58a46FqyqIhAxyXBbuWlkGid0csS6LBGRhKOAKiJREdi9geLnf0n1ilfpPus9UgacBkDGGT+NbWHNEDZM/raumn9tqe/DOijTwQ0j0khzxXegFhFprxRQRaRNhavKKPvnryn/9xMQCYPVRt2W5Q0BNd6V10V4uNDHxr31LaQm905mygkp2NRCSkSkzSigikibMEIByt95lLLXfoNRUwlA6vDvkTP1QVx5A2JcXfOsKQvyyIpKKoMmyXYLPxvmZUyuK9ZliYgkPAVUEWkT2+aeRc269wBwFQwj59K5pA46I8ZVNY9pmrz6dQ3PrfdjAgUeOzeP8pKbol+ZIiLRoN+2ItImOk24muDuDWRffB9pJ1+Oxdo+bvlZEzL4f5/7+LQ4CMCpXd1cNcSDy6ZD+iIi0aKAKiLHLVi6mZIXZpEyaCKdTrsaAO9Jl+IZeR5WV0qMq2u+rb4w85ZXUlQTwW6FGYM8nJHvVgspEZEoU0AVkWMW8VdQ9tp9lL/zCGY4SM3Gj0g/ZToWuxOLxYKlHYXTD3fU8tTqKoIGZCXVt5Dqk64WUiIisaCAKiItZoZD7H1vAaWLfkWkeg8AKYO+Q86lD2GxO2NcXcuEIiZ/XlvNO9vqW0gN6+zkuuFePE61kBIRiRUFVBFpEf+Gj9j99FUEizYC4MobSPalD5E69Mx2dyi8rDbCw4WVbKoIYwEu6pvMRX1TsLaz9yEikmgUUEWkRawON8Gijdi82WRf+GvST70Si639/Sr5vDTAoyt8VIVMUhwWrhvuZUS2WkiJiMSD9vdXRUSiKrRnOzWblpI2dgoASb3G0PVnC0kdeha2JG+Mq2s5wzR5ZVMNL26sbyHVK83OTSPTyE5uH10GREQ6AgVUEWlSpLaKPW/MYc/iuWAaJPUag7NzTwDSxk6NcXXHpjpo8NhKHytK61tInVHg5ocDPTjVQkpEJK4ooIpII2YkTMWHT1Pyyt1EKosBSD7hW5ihQIwrOz5fV4aYt7yS0loDhxWuGuzhtPykWJclIiJNUEAVkQbVq9+ieOEvCOxYA4Azpw/ZUx7EM2pyu7sA6mD/3lbL019UETIgJ9nKzJFp9EhTCykRkXilgCoiAISr97D9kQsxgzVYUzrRefI9ZJz+03bXNupgwYjJ02uqeG9HHQAjs51cO9xLqkMtpERE4pkCqkgHFvFXYEtJB8Cemknn8+4kXFVK5/PuxJaaEdvijlNxTYR5yyvZ4qtvITW1Xwrn905WCykRkXZAAVWkAzICNexZPI89b84h/6bXSek/AYCsc2+LcWWto7A4wGMrffjDJh6nhetHpDE0q/2uBIuIdDQKqCIdiGkYVH78N0peup3w3p0AVP73rw0Btb0zTJOXNvr5x6YaAPqk17eQykpSCykRkfZEAVWkg/Cve5/i526mbmshAI6s7mRfPBtvO20Z9U2+oMGjKypZVRYCYFL3JK4YmIrdqkP6IiLtjQKqSAew+y/XsnfJ/wPAmuQl69zbyfjODVid7hhX1jo2VdS3kNpTZ+CywdVDvHyra2K8NxGRjuiYLmV9/PHH6dGjB263m7Fjx7Js2bIjjn/xxRfp378/brebIUOG8OabbzZ6vrq6mp///Od069aNpKQkBg4cyIIFC46lNBFpQnKf8WC10emMn9HnwU1knXNrQoRT0zR5e2sNd3+8lz11Bl1SbPzm5AyFUxGRdq7FAfX5559n5syZ3HPPPRQWFjJs2DAmTZpESUlJk+M//vhjpk2bxpVXXsmKFSuYPHkykydPZs2aNQ1jZs6cyeLFi/nb3/7GunXruPHGG/n5z3/Oq6++euzvTKSDMoJ1lL35Wyo//nvDNu9J0+jzwHq6XPE4dm/nGFbXegIRk8c/9/HHNdVETDgx18X9J3eiwKMDQyIi7Z3FNE2zJS8YO3YsY8aM4bHHHgPAMAzy8/O57rrrmDVr1iHjp06dit/v5/XXX2/YdtJJJzF8+PCGVdLBgwczdepU7rrrroYxo0aN4qyzzuI3v/nNUWvy+XykpaVRWVmJ19v+7g0u0hpM08S37AVKXphFqGwL9rRc+jz4JVZ3aqxLa3W7/WHmLa9kW1UEqwW+3y+V7/VKatc3ExARSXQtyWstWkENBoMsX76ciRMnHtiB1crEiRNZunRpk69ZunRpo/EAkyZNajR+/PjxvPrqq+zcuRPTNHnvvffYuHEj3/3ud5vcZyAQwOfzNfoS6chqNi1ly73j2fn/Lq0Pp+l5ZF8yG4sz8W7l+WlRgNv+s5dtVRHSXFbuGpvOub2TFU5FRBJIi46FlZWVEYlEyMnJabQ9JyeH9evXN/maoqKiJscXFRU1fP/oo49yzTXX0K1bN+x2O1arlaeeeopTTz21yX3Onj2bX/3qVy0pXSQhBUu3UPL8L/F9+iIAFlcKWWf/ksyzbsbqSolxda0rYpgs3ODn1a/rW0j17+TghpFeMtxqISUikmji4mStRx99lP/973+8+uqrdO/enQ8//JBrr72WvLy8Q1ZfAW677TZmzpzZ8L3P5yM/Pz+aJYvEhbCvuD6cWiykf+tHdL7oXhzpXWJdVqurCBj8rrCSteX1LaTO6ZnE9/urhZSISKJqUUDNysrCZrNRXFzcaHtxcTG5ublNviY3N/eI42tra7n99tt55ZVXOOeccwAYOnQoK1eu5KGHHmoyoLpcLlwuV0tKF0kIZjhE7dfLSD7hZACSe48le+qDpA6ehLtgaIyraxvry4PML/SxN2Dgtln4yTAP47roKn0RkUTWonNQnU4no0aNYsmSJQ3bDMNgyZIljBs3rsnXjBs3rtF4gHfeeadhfCgUIhQKYbU2LsVms2EYRkvKE0lYpmlSVfhPvrpjMFvnnE6wdHPDc1ln35KQ4dQ0Td7cXMOv/1fB3oBB11Qb95/SSeFURKQDaPEh/pkzZzJ9+nRGjx7NiSeeyPz58/H7/cyYMQOAK664gq5duzJ79mwAbrjhBiZMmMDcuXM555xzWLhwIZ999hlPPvkkAF6vlwkTJnDLLbeQlJRE9+7d+eCDD/jLX/7CvHnzWvGtirRPtZuXU7zwZmrWfwCAzdOZYPEmnJ17xriytlMbNvj9qiqW7g4AMK6Li58M9eC2H1PrZhERaWdaHFCnTp1KaWkpd999N0VFRQwfPpzFixc3XAi1bdu2Rquh48eP59lnn+XOO+/k9ttvp2/fvixatIjBgwc3jFm4cCG33XYbl112GeXl5XTv3p377ruPn/zkJ63wFkXap9Ce7ZS8dAeVH/8VAIvDTcakm8j63ixsSYnbTm1HVZh5hZXsrI5gs8DlA1I5s4daSImIdCQt7oMaj9QHVRKNEazly5vyiVTvASBt3GVkX3I/jsyCGFfWtpbuqmPBqirqIiadXFZuGplGvwxHrMsSEZFW0JK8FhdX8YsImIaBZd/RB6sziU5nXEvNuvfImTaXpF5jYlxd2wobJn9fX82bm2sBGJTp4IYRaaS5dEhfRKQj0gqqSByoXrWY4udvIffyx0jpPwGov2Ifmz3hD22X10WYX+hjw976FlLn905m6gkp2NRCSkQkoWgFVaSdqNu+muKFv8C/5m0Ayl69ryGgWuyJf2j7iz1BfldYSWXQJMlu4dphXsbkqoWciEhHp4AqEgPhiiJKXr6Lig+fBtMAm4OM71xP5/PuiHVpUWGaJq99XcNzG/wYJhR47Nw8yktuin4liYiIAqpI1JUveYLi52/BDPgB8I65hOwpD+DM7hXjyqKjJmTw/z738WlxEIBTu7q5aogHl02H9EVEpJ4CqkiUWd0ezICfpN5jybl0bsNdoTqCbb76FlK7/RHsVvjhQA8TC9wJf56tiIi0jAKqSBvzr3sfo9aHZ+R5AKSN+z62lE6kDju7QwWzj3bU8eRqH0EDspLqW0j1SU/882xFRKTlFFBF2khg9waKF95C9crXsKflkjLwdKzuVCxWK57h58S6vKgJRUz+sq6at7fWt5AamuXkuhFevE61kBIRkaYpoIq0snBVGaWLfsXe9xZAJAxWG57RF2FGwrEuLerKaiM8XFjJpor6935R32Qu7puCtQOtHIuISMspoIq0EiNYR/m7j1L22n0YNZUApA4/l5ypD+LK6x/j6qJvVWmQR1ZUUhUySXFYuG64lxHZaiElIiJHp4Aq0koCO1ZT8vwvAXAXDCdn2lxSBp4e46qizzBNFm2q4YWNfkygp9fOzFFpZCfbYl2aiIi0EwqoIschVL4DR0Y3AJJ6jSHjuzfiLhhG2smXY7F2vEBWHTJ4fKWPwpL6FlJn5Lv54SAPTrWQEhGRFlBAFTkGwZKvKXlhFlUrX6P37HU4O/cAIPeyh2NbWAxtrgwxb3klJbUGDitcOdjDt/OTYl2WiIi0QwqoIi0Q8e+l9NX72Pvuo5jhIFgs+L94F+dpV8W6tJh6b3stf1xTRciA7CQrM0el0TNNLaREROTYKKCKNIMZDlH+7ycoW/QrIv5yAFIGfYecSx/CXTA0xtXFTjBi8qcvqvj39joARmY7uXa4l1SHWkiJiMixU0AVOQrTiLD5VydSt20lAK68gWRf+hCpQ8/sUI32v6mkJsK85ZVs9oWxAFP7pXB+72S1kBIRkeOmgCpyFBarjdTh3yNUsYvsC39N+qlXYrF17P91VpQEeHSlD3/IxOO0cP2INIZmOWNdloiIJAiLaZpmrIs4Xj6fj7S0NCorK/F6vbEuR9q50J7tlLx0O+mnXUNKv28BYAT8mEYEW1LH/vkyTJOXNvr5x6YaAPqk27lpZBpZSR2vY4GIiLRMS/Jax14GEjlIpLaKPW88wJ7F8zBDdQSLv6THXUuxWCxYXSmxLi/mfEGDR1f4WFVW30JqUvckLh+QikMtpEREpJUpoEqHZ0bCVHz4NCUv30XEVwJAcr9TyZk2t0OfY3qwTRX1LaT21Bk4rXDNUC/f6uqOdVkiIpKgFFClQ/Ove4+iv15HYOcXADhz+pI99UE8I89XOAVM0+TdbXU8s7aKsAFdUmzMHJlGgVe/OkREpO3or4x0aKHyHQR2foEtJYOsyfeQcfpPsNh1sQ9AIGLyh9VVfLizvoXUmBwXPxvmIVktpEREpI0poEqHEq4oIliyieQTTgEgbdxlRHwlpJ/6I2wpnWJcXfzY7Q8zb3kl26oiWC3w/X6pfK9XklaVRUQkKhRQpUMwAjXsWTyXsjfmYEvy0mfORqzuVCxWK5ln3Rzr8uLKp0UBHv/cR23YJM1l5cYRXgZmalVZRESiRwFVEpppGFR+/FdKXrqD8N6dADi6DSZcVYrTnRrj6uJLxDB5fqOff35V30KqXycHN470kuFWCykREYkuBVRJWP5171H83M3UbV0BgCOrO9mXPIB37FQdqv6GioDBIysq+WJPCIBzeibx/f6p2K36nEREJPoUUCUhBXatZ+sDpwNgTfKSde4dZHzneqxOtUb6pg3lIR4urGRvwMBts/CTYR7GddHnJCIisaOAKgnDDAcbrsB35fUnbdxlWJPT6Tz5HuzezjGuLv6YpsniLbX8dV01ERO6pta3kOrm0a8FERGJLf0lkrgX2rONcFXZYZ+3ulKpKlzEnrfm0fPuT3BmdQcg78d/1aH8w6gLGyxYVcXS3QEAxnVx8ZOhHtx2tZASEZHYU0CVuBbas41Nt/bDDNU1a/ze935PziX3AyicHsbO6jBzl1eyszqCzQKXD0jlzB5qISUiIvFDAVXiWriqrFnh1ObpTM6lD5E2/gdRqKr9Wrq7jgWfV1EXMenksnLjSC/9M9RCSkRE4osCqiSE/OtfIfmEk2NdRtwKGybPrq/mjc21AAzMcHDDyDTSXTqkLyIi8UcBVRKCxZkU6xLiVnldhPmFPjbsrW8hdX7vZKaekIJNLaRERCROKaCKJLC1e4LMX+GjMmCQZLdw7TAvY3JdsS5LRETkiBRQRRKQaZq8/nUtz26oxjChwGNj5qg0uqTof3kREYl/+mslkmBqQgZPrKpiWVF9C6lvdXVz9RAPLpsO6YuISPuggCqSQLZVhZm3vJLd/gh2K/xwoIeJBW61kBIRkXZFAVXimsV69B9Ri8ON3ZMVhWri20c763hqtY9ABDLdVmaOSqNPuiPWZYmIiLSYAqrEtcqP/wqAI7sPXX/6LBar7ZAxdk8WjsyCaJcWN0IRk7+uq+atrfUtpIZmObhuRBpep1pIiYhI+6SAKnErtGcb5e8+CkDuDx4hudeYGFcUf8pqIzxcWMmmijAAF/VN5uK+KVh1SF9ERNoxBVSJWyUv340ZCpDc/zRSh54Z63LizqqyII+sqKQqaJLisHDdcC8jstVCSkRE2j8FVIlLpmFAJAwWCzlT5ugin4MYpsmiTTW8sNGPCfTw2rl5VBrZyYee/iAiItIeWUzTNGNdxPHy+XykpaVRWVmJ1+uNdTnSioIlX+PM7hXrMuJGdcjg8ZU+CkuCAJye72bGIA9OtZASEZE415K8phVUiWsKpwdsrgwxb3klJbUGDitcOdjDt/N1i1cREUk8usxX4oppmpS8fA/B0s2xLiWuvLe9lrs+3ktJrUF2kpV7x3dSOBURkYSlgCpxpWr5K5T989d8fc8ojEBNrMuJuWDE5MlVPhasqiJkwMhsJ7O/lUHPNPU3FRGRxKVD/BI3zHCIkhdvAyBj4nVYXckxrii2SmoizFteyWZfGAsw5YQUJvdJVgspERFJeAqoEjcqPnqaYNFGbJ7OZJ51c6zLiakVJQEeXenDHzLxOCxcPyKNoZ2dsS5LREQkKhRQJS4YAT+lr/wfAJ3PvwtbUsfsxmCYJi996eflL2swgT7pdm4amUZWklpIiYhIx6GAKnFhz+KHCVcW4ejci07f/nGsy4kJX9DgsZU+Pi+tbyH13e5JXDEgFYdaSImISAejgCoxF/aVsufNBwHIvvg+LPaOdyh7U0WIhwsrKas1cFrhmiFevtXNHeuyREREYkIBVWLO6kwi88yb8a9/D++JU2JdTlSZpsm72+p4Zm0VYQNyk23cPCqNAq/+1xQRkY5Ld5KSuGGaZoe6pWkgYvKH1VV8uLMOgDE5Ln42zEOyQ93fREQk8ehOUtJuHBxKO1I4LfKHmbvcx7aq+hZS3++fwrm9kjvUZyAiInI4WqqRmKnduoLNvz4J/9p/x7qUqPq0KMCs/+xlW1WYNKeFu05K57zeKQqnIiIi+2gFVWKm5IVZ1H29jL0f/pGUgafHupw2FzFMnt/o559f1d8hq18nBzeO9JLhVgspERGRgymgSkxUf/Eu/jVvg81B9oX3xrqcNlcZMHhkRSVr9oQAOLtnEpf1T8Vu1aqpiIjINymgStSZhkHJC7MAyDj9pzize8W4ora1oby+hdTegIHLZuGnQz2My1MLKRERkcNRQJWo8y17gboty7G6PWSdd2esy2kzpmmyeEstf11XTcSEvJT6FlLdPPrfTkRE5Ej0l1KiygwHKXnpDgAyz/4ldm/nGFfUNurCBr9fXcXHuwIAjOvi4sdDPSTZdV2iiIjI0SigSlRVfvICodKvsaXlkHnmTbEup03srA4zb3klO6oj2CzwgwGpnNUjSVfpi4iINJMCqkRV2rjvY7HawGrF6kqJdTmt7n+763ji8yrqIiadXFZuHOmlf0bHu3WriIjI8VBAlaiyWK2kjZsW6zJaXdgweXZ9NW9srgVgYIaDG0amke7SIX0REZGWUkCVqIj492KxObC6U2NdSqvbWxdh/gof68vrW0id3zuZqSekYFMLKRERkWOi5R2JipJ/3MmXv+yDb/miWJfSqtbuCXLrf/ayvjxEkt3CL0al8f3+qQqnIiIix0ErqNLmAkVfsvf9JyESxpbSKdbltArTNHn961qe3VCNYUK+p76FVJcU/S8lIiJyvPTXVNpc6Ut3QCRM6rCzSek/IdblHLeakMETq6pYVlTfQupbXV1cNdiL265VUxERkdaggCptqvarZfg+fREsFrIveSDW5Ry3bVX1LaR2++tbSP1wUCrfKVALKRERkdakgCptxjRNil+4FYC0k6/AnT8kxhUdn//srOPJ1T4CEch0W5k5Ko0+6Y5YlyUiIpJwjukiqccff5wePXrgdrsZO3Ysy5YtO+L4F198kf79++N2uxkyZAhvvvnmIWPWrVvHeeedR1paGikpKYwZM4Zt27YdS3kSJ/yr36Jm/ftYHC6yL/x1rMs5ZmHD5Ok1VTy6sj6cDs1y8MC3MhRORURE2kiLA+rzzz/PzJkzueeeeygsLGTYsGFMmjSJkpKSJsd//PHHTJs2jSuvvJIVK1YwefJkJk+ezJo1axrGfPXVV5xyyin079+f999/n1WrVnHXXXfhdruP/Z1JzNV8+V8AOp3xcxyZBTGu5tiU1Ub4v6V7eWtrfX/Ti/okc9uJ6XidaoAhIiLSViymaZotecHYsWMZM2YMjz32GACGYZCfn891113HrFmzDhk/depU/H4/r7/+esO2k046ieHDh7NgwQIALr30UhwOB3/961+P6U34fD7S0tKorKzE6/Ue0z6kbdRu/gxn517YUjNiXUqLrS4L8rsVlVQFTVLsFn4+3MvIHFesyxIREWmXWpLXWrQMFAwGWb58ORMnTjywA6uViRMnsnTp0iZfs3Tp0kbjASZNmtQw3jAM3njjDU444QQmTZpEdnY2Y8eOZdGiRYetIxAI4PP5Gn1JfErqObrdhVPDNHllk5/7PqmgKmjSw2vngW9lKJyKiIhESYsCallZGZFIhJycnEbbc3JyKCoqavI1RUVFRxxfUlJCdXU1DzzwAGeeeSZvv/02F1xwARdeeCEffPBBk/ucPXs2aWlpDV/5+fkteRvSxqpXLSZYujnWZRyT6pDBQ59VsnCDHxP4dr6be8d3IjvZFuvSREREOoyYX8VvGAYA559/PjfddBMAw4cP5+OPP2bBggVMmHBo38zbbruNmTNnNnzv8/kUUuNEpKaSnQsuwwhU0/22D0juc1KsS2q2LZUh5hVWUlxj4LDCjwZ5OL0gKdZliYiIdDgtCqhZWVnYbDaKi4sbbS8uLiY3N7fJ1+Tm5h5xfFZWFna7nYEDBzYaM2DAAP7zn/80uU+Xy4XLpcOt8WjPmw8S8ZfjzBtAUs/RsS6n2d7fXssf1lQRMqBzUn0LqV5pukpfREQkFlp0iN/pdDJq1CiWLFnSsM0wDJYsWcK4ceOafM24ceMajQd45513GsY7nU7GjBnDhg0bGo3ZuHEj3bt3b0l5EmOh8p3seethAHIumY3FFvMF+qMKRkyeXOXjiVX14XRktpMHTslQOBUREYmhFieImTNnMn36dEaPHs2JJ57I/Pnz8fv9zJgxA4ArrriCrl27Mnv2bABuuOEGJkyYwNy5cznnnHNYuHAhn332GU8++WTDPm+55RamTp3Kqaeeyre//W0WL17Ma6+9xvvvv98671KionTRrzCDtST1PZnUEefFupyjKqmJ8HBhJV9XhrEAl5yQwgV9krHqrlAiIiIx1eKAOnXqVEpLS7n77rspKipi+PDhLF68uOFCqG3btmG1HliYHT9+PM8++yx33nknt99+O3379mXRokUMHjy4YcwFF1zAggULmD17Ntdffz39+vXjH//4B6ecckorvEWJhsCu9VR8+EcAcqbMiftbf64oCfDoSh/+kInHYeG6EV6GddZpIyIiIvGgxX1Q45H6oMbe9kcupGr5K3hGnk/+DYtiXc5hGabJP770848vazCB3ml2Zo5KIytJV+mLiIi0pZbktfg/SVDinmkYuLoNwb92CdmXzI51OYdVFTR4bKWPlaVBAL5TkMT0gak4bPG92isiItLRaAVVWo1RV43VnRrrMpq0qSLEw4WVlNUaOK1w9RAPp3ZTCykREZFo0QqqxEQ8hlPTNFmyvY4/fVFF2IDcZBszR6XR3asffRERkXjVojZTIgczI2F2LvgB/nXvx7qUJgUiJk+squKp1fXhdEyOk9mndFI4FRERiXP6Sy3HrOK/f6Fy6d+pXvUv+s7bGlcrqEX+MPMKfWz11beQmtY/hfN6Jcd9dwERERFRQJVjZARrKX35bgCyzrsjrsLpZ8UBHl/poyZskua0cMPINAZlOmNdloiIiDSTAqock/J3HiG8dyeOzAI6nf6zWJcDQMQweWGjn0Vf1QBwQicHN430kuFWCykREZH2RAFVWixSXU7Z6/XtpDpfeC9WpzvGFUFlwOCRFZWs2RMC4OyeSVzWPxW7VYf0RURE2hsFVGmxstdnY9RU4uo2hLTxl8W6HDburW8hVV5n4LJZ+MlQD+PzYh+aRURE5NgooEqLhMp3UP7uowBkT5mDxRq9w+eGabKuPERFnUG620r/Tnbe3lrHX9dVEzEhL8XGzaPS6ObRj7WIiEh7pr/k0iL29Dzyrnya6jVvkzr0zKjN+8nuOp5ZW015ndGwzWmF4L5vx3Vx8eOhHpLs6pwmIiLS3imgSotYrFbSxn2ftHHfj9qcn+yuY16h75Dt+8PphG5ufjrUoxZSIiIiCULLTdJsRigQ/TlNk2fWVh9xzJqyIO3+fr0iIiLSQAFVmsW//gM2/aInez/4Y1TnXVceanRYvyl76gzWlYeiVJGIiIi0NQVUOSrTNCl54VbCFbup27I8qnNXHCWctnSciIiIxD8FVDmqqs9epvarT7A4k+l8/t1RnTvd3bwf0eaOExERkfinv+pyRGY4RMlLtwOQedbN2NNzozr/gAwHXueRL37KdFsZkOGIUkUiIiLS1hRQ5YgqPnqaYNFGbJ7OZJ71i6jPHzHAdpSr86cPTMWqK/hFREQShtpMyWEZAT+lr/wfAJ3PvwtbkjfqNbz6dQ17AwbJdgsum4W9gQPnmma6rUwfmMrYLrprlIiISCJRQJXDql61mHBlEY7Ovej07R9Hff4if5hXNvkBuGqwh3F5rkZ3khqQ4dDKqYiISAJSQJXD8o65iJ73LMOoq8Jid0Z1btM0+cOaKkIGDM1yMj7PhcViYVBmdOsQERGR6FNAlSNK6jUmJvP+d1eA1WUhHFa4cnCq7hIlIiLSgegiKTlEqHwnwdItMZu/OmTwl7VVAFzYJ4XcFP07SkREpCNRQJVDFL/wS76a1Y+97/8hJvM/u76ayqBJ11Qb5/VOjkkNIiIiEjtampJGareuwLf0WQDcPUZGff4N5SGWbKsD4OohHuxWHdoXERHpaLSCKo2UvDALAO9J00iKckANGyZPrfEB8O18NwMydEGUiIhIR6SAKg2qv3gX/5q3weYg+6LfRH3+NzbXsL0qgsdp4bL+qVGfX0REROKDAqoAYBpGw+ppxuk/xZndK6rzl9REeGljfc/Tywek4nHqR1NERKSjUgoQAHyfvkjdluVY3R6yzrszqnObpskf11QRNGBQpoNTu+rOUCIiIh2ZAqoAEKkqw+JMJvPsW7B7O0d17v8VBVhZGsRurb9jlHqeioiIdGy6il8AyJh4LZ7RF2JL8kZ13pqQwTNfVAMwuXcKean6kRQREenolAakgSO9S9TnXLjBT0XAoEuKjcnqeSoiIiLoEH+HV/Hhn/Cv/yAmc2+qCPH21lqg/tC+w6ZD+yIiIqKA2qGFK4rY/bfr2Dr7tKiH1Ihh8uTqKkzg1K5uBmep56mIiIjUU0DtwEpfvRcz4Mfd60SS+50a1bnf3FLLVl+YVIeFyweo56mIiIgcoIDaQQWKvmTv+08CkDP1waheOV9WG+HFjfUXRv1gQCpel34MRURE5AAlgw6q9KU7IBImddjZpPSfELV5TdPk6TVVBCLQP8PBad3U81REREQaU0DtgGq/Wobv0xfBYiH74tlRnfvT4iDLS4LYLHC1ep6KiIhIExRQOxjTNCl+4VYA0sZfjrtgaNTmrg0b/OmLKgDO651MN4+6nImIiMihlBA6oE4TriJcWUT2RfdGdd7nN/gprzPISbZxYZ+UqM4tIiIi7YcCagdjsVhIG38Z3nHfj+rh9a8rQyzeUt/z9MrBqTjV81REREQOQ4f4OxDTNBseRzOcGqbJU/t6np6c52JYZ1fU5hYREZH2RwG1gzCCdWz5zcnsff8pzEg4qnO/taWWryvDpNgtXKGepyIiInIUOsTfQez99xPUblpKaM820sb/AIstOv/p99RGeH6jH4Bp/VNJd9uiMq+IiIi0X1pB7QAiNZWUvfobADpf8CuszqSozf3M2mpqwyYnpNs5o0A9T0VEROToFFA7gD1vPkjEX44zbwDpp0yP2rzLiwMsKwrU9zwd4sWqnqciIiLSDAqoCS5UvpM9bz0MQM4lD0Tt0H5d2OTpfT1Pz+mZTIFXZ5OIiIhI8yigJrjSRb/CDNaS1PdkUkecG7V5X/rST1mtQeckKxf1Vc9TERERaT4F1AQW2ruLio+eBiBnypyotZba4gvxxuYaAK4c7MFt16F9ERERaT4dd01gjk559Ljzv1Sv+hfJJ5wclTn39zw1TDipi4sR2ep5KiIiIi2jgJrgknuPJbn32KjN9+62WjZVhEmyW5g+UD1PRUREpOV0iD8BmaZJaO+uqM+7ty7Cs+vre55e2i+FDPU8FRERkWOggJqAqle8xqZbelHyj7uiOu+f9/U87Z1m57vdo9drVURERBKLAmqCMSNhSl68DTMUwDQiUZt3ZUmApbsDWICrh3jU81RERESOmQJqgqn4718I7FqLLSWDrLN/GZU5AxGTP66p73l6ds8keqY5ojKviIiIJCYF1ARiBGspffluALLOuwNbSnpU5n35Sz8ltQaZbitTTlDPUxERETk+CqgJpPydRwjv3Ykjs4BOp/8sKnNuqwrz2tf1PU9/NNiD264fKRERETk+ShMJIlJdTtnrswHofOG9WJ3uNp/TME3+sLqKiAljcpyMzlHPUxERETl+CqgJovbrZZjhIK5uQ0gbf1lU5nxvex0b9oZw2yz8cJAnKnOKiIhI4lOj/gSROvRM+jy4iUhVKRZr2/cfrQwY/H19NQBT+qWQlaSepyIiItI6FFATiKNTHo5OeVGZ66/rqvCHTHp47ZypnqciIiLSinSIv50L7FqPf8NHUZ1zVVmQj3bW9zy9ZogHm1U9T0VERKT1KKC2c8ULf8HW+0+l7I0HozJfMGLyx9X1PU8n9Uiid7p6noqIiEjrUkBtx/zrP6D68zfAasMz6oKozLlok5+imgidXFamquepiIiItAEF1HbKNE1KXrgVgE6nXYMrt2+bz7mzOsyir+p7ns4YlEqyQz8+IiIi0vqUMNqpqs9epvarT7A4k+l8/t1tPp95UM/TkdlOTsxVz1MRERFpGwqo7ZAZDlHy0u0AZJ51M/b03Daf84MddawtD+GywYxBHiwWXRglIiIibUMBtR2q+OhpgkUbsXk6k3nWL9p8Pl/Q4G/r6nueXtw3hexk9TwVERGRtqM+qO2QzdMZR2YBmWf9AluSt83n+9u6aqpCJgUeO2f3TG7z+URERKRjO6YV1Mcff5wePXrgdrsZO3Ysy5YtO+L4F198kf79++N2uxkyZAhvvvnmYcf+5Cc/wWKxMH/+/GMprUPwjr6Q3g9soNO3f9zmc32xJ8gHO+qwAFcP8WBXz1MRERFpYy0OqM8//zwzZ87knnvuobCwkGHDhjFp0iRKSkqaHP/xxx8zbdo0rrzySlasWMHkyZOZPHkya9asOWTsK6+8wv/+9z/y8qJzN6T2zOp0Y7E723SOUKT+wiiAiQVJnNBJPU9FRESk7bU4oM6bN4+rr76aGTNmMHDgQBYsWEBycjJPP/10k+N/97vfceaZZ3LLLbcwYMAA7r33XkaOHMljjz3WaNzOnTu57rrr+Pvf/47DoSDUlNJX72Pv+3/AjISjMt+rX9ewyx8hzWVlWn/1PBUREZHoaFFADQaDLF++nIkTJx7YgdXKxIkTWbp0aZOvWbp0aaPxAJMmTWo03jAMLr/8cm655RYGDRp01DoCgQA+n6/RV6ILlm6mdNGv2P2nq6n58r9tPt9uf5hXNvkBmD4wlRT1PBUREZEoaVHqKCsrIxKJkJOT02h7Tk4ORUVFTb6mqKjoqOPnzJmD3W7n+uuvb1Yds2fPJi0treErPz+/JW+jXSr9x10QCZEy6Duk9J/QpnPt73kaMmBYZyfju6jnqYiIiERPzJfFli9fzu9+9zueeeaZZvfWvO2226isrGz42r59extXGVu1W1dQufTvAGRPeaDN5/vPrgBr9oRwWOHKwep5KiIiItHVooCalZWFzWajuLi40fbi4mJyc5tuFp+bm3vE8R999BElJSUUFBRgt9ux2+1s3bqVm2++mR49ejS5T5fLhdfrbfSVyEpemAWA96RpJPUY2aZzVQcN/rK2/sKoi/qmkKOepyIiIhJlLQqoTqeTUaNGsWTJkoZthmGwZMkSxo0b1+Rrxo0b12g8wDvvvNMw/vLLL2fVqlWsXLmy4SsvL49bbrmFt956q6XvJ+FUf/Eu/jVvg81B9kW/afP5nl1fjS9o0i3Vxrm91PNUREREoq/FjfpnzpzJ9OnTGT16NCeeeCLz58/H7/czY8YMAK644gq6du3K7NmzAbjhhhuYMGECc+fO5ZxzzmHhwoV89tlnPPnkkwBkZmaSmZnZaA6Hw0Fubi79+vU73vfXrpmm2bB6mnH6T3Fm92rT+daXB1myvQ5Qz1MRERGJnRYH1KlTp1JaWsrdd99NUVERw4cPZ/HixQ0XQm3btg2r9cDC7Pjx43n22We58847uf322+nbty+LFi1i8ODBrfcuEpTFYiH38scoe/Vess67s03nChsmT+3reXp6vpv+GW3bY1VERETkcCymaZqxLuJ4+Xw+0tLSqKysTPjzUdvKok1+ntvgx+u08PCETFKdMb9+TkRERBJIS/KaUkicMkKBqM1VXBPhpS/re55eMcCjcCoiIiIxpSQShyK1VXx16wkUL7wFI+Bv07lM0+SPa+p7ng7OdHBKV/U8FRERkdhSQI1De/71EKE926ha8SoWW9ueC7p0d4DPS4PY1fNURERE4oQCapwJVxSxZ/FcALIvvh+L3dFmc/lDBn9eWw3ABX1SyEtt8TVzIiIiIq1OATXOlL56L2bAj7vXiXhGX9imcy3c4KciYJCXYuN89TwVERGROKGAGkcCRV+y9/36/rA5Ux9s08PtX+4N8c7WWgCuGuLBYdOhfREREYkPCqhxpPSlOyASJnXYOaT0n9Bm84QNkydXV2ECE7q5GZSpnqciIiISPxRQ40S4ooiqz98Ai4XsS2a36Vz/2lzLtqowqQ4LPxiQ2qZziYiIiLSUroqJE/b0XPrM2Yh/zdu484e02TylNRFe/LL+wqgfDEjFq56nIiIiEmeUTuKII6Mr6afOaLP9m6bJ019UEYjAgAwHp3Vzt9lcIiIiIsdKATXGTMOg9qtlUZlrWVGAwpIgNgtcpZ6nIiIiEqcUUGOscunf2fzrsex8cnqbzlMTMnhmX8/T83sn082jsztEREQkPimgxpARrKP0H3cC4Mob0KZzvbDRT3mdQU6yjQv6pLTpXCIiIiLHQwE1hvb++wlCe7Zh79SVjO/e0GbzfFURYvGWfT1PB3twquepiIiIxDEF1BiJ1FRS9upvAOh8wa+wOpPaZh7D5Kl9PU9PznMxtLN6noqIiEh8U0CNkbI35hDxl+PMG0D6KW13/ulbW2vZ7AuTYrdwxUBPm80jIiIi0loUUGMgVL6T8rfnA5BzyQNYbG1zwdKe2gjPb/AD8P0BqaS79J9bRERE4p8u5Y6BUNkWbKlZODILSB1xbpvN86cvqqmLmJzQycHp+ep5KiIiIu2DAmoMJJ9wMn3mbCRSVdJmvUg/Kw7waXEAmwWuHuLBqp6nIiIi0k4ooMaI1enGmlnQJvuuCxs8vaYKgO/1SqZAPU9FRESkHdFJiVFU+9Uy9n7wR8xIuE3neXGjnz11Bp2TrFzUVz1PRUREpH3R0lqUmKZJ0cKbqd34H0Klm8m++DdtMs+WyhBv7ut5euVgDy71PBUREZF2RiuoUVK98nVqN/4Hi8NNp9N/2iZzGKbJU2uqMEwY18XFiGxXm8wjIiIi0pYUUKPAjIQpeWEWABnfvRFHRtc2meedrbVsqgiTZLcwfWBqm8whIiIi0tYUUKOg4r9/IbBrLbaUDLLOubVN5iivi/Dcvp6n0/ql0Mlta5N5RERERNqaAmobM4K1lL58NwBZ592BLSW9Teb5y9pqasMmfdLtfKd729w2VURERCQaFFDbWPk7jxLeuxNHZgGdTv9Zm8yxoiTA0t0BrBa4erB6noqIiEj7pqv421hyv1NJOuEUOp12DVZn69/NKRAx+eO+nqdn90iiR5qj1ecQERERiSYF1DaW3Ocketz+YZvt/x9f+imtNch0W7nkBPU8FRERkfZPATUK2up2ptuqwrz+dQ1Q3/PUbdcZGyIiItL+KdG0kd1//hnFz99KxL+3TfZvmCZPra4iYsKYHBejctTzVERERBKDVlDbQN321ex9bwGYJt7RF5HU+8RWn+Pf2+vYuDeE22ZhxiD1PBUREZHEoRXUNlDy0u1gmnjGXNwm4bQiYPDsumoApvZLITNJPU9FREQkcSigtjL/+g+pXvk6WG1kX3x/m8zxl7VV+MMmPb12zuyhnqciIiKSWBRQW5FpmpS88EsAOp12Da7cvq0+x6rSIP/dFcACXDNUPU9FREQk8SigtqKq5a9Q+9UnWFwpdD7/7lbff/Cgnqdn9kiil3qeioiISAJSQG0lpmlS9s97Acg882bs6bmtPscrm/wU1UTo5LIyRT1PRUREJEEpoLYSi8VC/k2vk/Gd68k86+ZW3/+OqjD//Kq+5+mMQakkO/SfTkRERBKT2ky1IkdGV3J/8LtW369pmvxhTX3P05HZTk7MVc9TERERSVxahmsF4YqiNt3/+zvqWFcewmWDHw3ytNmdqURERETigQLqcQr7Stl06wls/91kIv6KVt+/L2Dwt309Ty/pm0rnZPU8FRERkcSmgHqcyl67D6OuitCe7ViTvK2+/7+tr6Y6ZNLda+esnup5KiIiIolP56C2QGjPNsJVZQe+37uT8ncfByB9wlWE9+7AkVnQavN9sSfIBzvqsABXD/Zgt+rQvoiIiCQ+BdRmCu3ZxqZb+2GG6pp8vugvP6P4OTd95mxolZAaipg8tbq+5+nEgiT6dlLPUxEREekYdIi/mcJVZYcNp/uZobpGK6zH459f17DbHyHdZWVaf/U8FRERkY5DATUO7aoO88omPwDTB6aSop6nIiIi0oEo+cQZ06y/nWnYgOGdnYzrop6nIiIi0rEooMaZj3bWsWZPCIcVfjRYPU9FRESk41FAjSPVQYO/7ut5elHfFHLU81REREQ6IAXUOPL39dX4gibdUm2c2ys51uWIiIiIxIQCapxYVx7k39vruwRcPUQ9T0VERKTjUkBtJrsnC4vDfcQxFocbuyerxfsOGyZ/2Nfz9Ix8N/0znMdUo4iIiEgiUKP+ZnJkFtBnzoYj9jm1e7KOqUn/a1/XsKM6gtdp4fv9U4+nTBEREZF2TwG1BRyZBa16K1OAIn+Yf3xZ3/P0ioEeUp1a1BYREZGOTWkohkzT5OkvqgkZMCTLwSl56nkqIiIiooAaQx/vDvB5aRCHFa5Uz1MRERERQAE1Zvwhgz+vre95ekGfFLqk6GwLEREREVBAjZnn1vupDBjkpdg4Tz1PRURERBoooMbAxr0h3t1WC9T3PHXYdGhfREREZD8F1CgLGyZPrfZhAhO6uRmYqZ6nIiIiIgdTQI2yNzfXsK0qgsdh4QcD1PNURERE5JsUUKOopCbCS/t6nv5gQCpe9TwVEREROYQSUpSYpsmfvqgiEIGBGQ4mdDvybVNFREREOioF1ChZVhSgsCSIzQJXDVHPUxEREZHDUUCNgpqQwZ++qO95Orl3Ml1T1fNURERE5HAUUKPg+Y1+9gYMcpNtTO6TEutyREREROKaAmob21QR4q0t9T1PrxziwamepyIiIiJHpIDahiKGyVOrqzCBU/JcDM1Sz1MRERGRo1FAbUOLt9ayxRcmxWHhioGeWJcjIiIi0i4cU0B9/PHH6dGjB263m7Fjx7Js2bIjjn/xxRfp378/brebIUOG8OabbzY8FwqFuPXWWxkyZAgpKSnk5eVxxRVXsGvXrmMpLW6U1UZ4fkN9z9PL+qeS5tK/BURERESao8Wp6fnnn2fmzJncc889FBYWMmzYMCZNmkRJSUmT4z/++GOmTZvGlVdeyYoVK5g8eTKTJ09mzZo1ANTU1FBYWMhdd91FYWEhL7/8Mhs2bOC88847vncWY898UUUgYtKvk4Nv56vnqYiIiEhzWUzTNFvygrFjxzJmzBgee+wxAAzDID8/n+uuu45Zs2YdMn7q1Kn4/X5ef/31hm0nnXQSw4cPZ8GCBU3O8emnn3LiiSeydetWCgoKjlqTz+cjLS2NyspKvF5vS95Om/i0KMBDyyuxWWDOtzLI96itlIiIiHRsLclrLVpBDQaDLF++nIkTJx7YgdXKxIkTWbp0aZOvWbp0aaPxAJMmTTrseIDKykosFgvp6elNPh8IBPD5fI2+4kVd2OBPX1QB8L1eyQqnIiIiIi3UooBaVlZGJBIhJyen0facnByKioqafE1RUVGLxtfV1XHrrbcybdq0w6br2bNnk5aW1vCVn5/fkrfRpl7Y6GdPnUF2kpWL+qrnqYiIiEhLxdWVO6FQiClTpmCaJk888cRhx912221UVlY2fG3fvj2KVR7e5soQb27e1/N0sAeXep6KiIiItFiLjj9nZWVhs9koLi5utL24uJjc3NwmX5Obm9us8fvD6datW/n3v/99xHMTXC4XLperJaW3OcM80PN0XBcXw7Pjqz4RERGR9qJFK6hOp5NRo0axZMmShm2GYbBkyRLGjRvX5GvGjRvXaDzAO++802j8/nD65Zdf8u6775KZmdmSsuLC21tr+aoyTJLdwvSBqbEuR0RERKTdavEVPDNnzmT69OmMHj2aE088kfnz5+P3+5kxYwYAV1xxBV27dmX27NkA3HDDDUyYMIG5c+dyzjnnsHDhQj777DOefPJJoD6cXnzxxRQWFvL6668TiUQazk/NyMjA6Yz/uy+V10VYuK/n6ff7p9DJbYtxRSIiIiLtV4sD6tSpUyktLeXuu++mqKiI4cOHs3jx4oYLobZt24bVemBhdvz48Tz77LPceeed3H777fTt25dFixYxePBgAHbu3Mmrr74KwPDhwxvN9d5773Haaacd41uLnme+qKY2bNIn3c7EgqRYlyMiIiLSrrW4D2o8imUf1BUlAR74tBKrBWaf0okeXkdU5xcRERFpD1qS19Sk8xgYpsm68hAlNRGeW18NwDk9kxVORURERFqBAmoLfbK7jmfWVlNeZzRsswI9vPooRURERFqDUlULfLK7jnmFh961ygAeXenDYYWxXdzRL0xEREQkgcRVo/54Zpgmz6ytPuKYP6+txmj/p/SKiIiIxJQCajOtKw81OqzflD11BuvKQ1GqSERERCQxKaA2U8VRwmlLx4mIiIhI0xRQmynd3byPqrnjRERERKRpSlPNNCDDQcZRwmem28qADLWaEhERETkeCqjNZLVY+OHA1COOmT4wFavFEqWKRERERBKTAmoLjO3iZuZI7yErqZluKzNHetViSkRERKQVqA9qC43t4mZMrot15SEq6gzS9x3W18qpiIiISOtQQD0GVouFQZnOWJchIiIikpB0iF9ERERE4ooCqoiIiIjEFQVUEREREYkrCqgiIiIiElcUUEVEREQkriigioiIiEhcUUAVERERkbiigCoiIiIicUUBVURERETiSkLcSco0TQB8Pl+MKxERERGRpuzPaftz25EkRECtqqoCID8/P8aViIiIiMiRVFVVkZaWdsQxFrM5MTbOGYbBrl278Hg8WCyWWJcT93w+H/n5+Wzfvh2v1xvrcjoMfe6xoc89+vSZx4Y+99jQ5958pmlSVVVFXl4eVuuRzzJNiBVUq9VKt27dYl1Gu+P1evU/Uwzoc48Nfe7Rp888NvS5x4Y+9+Y52srpfrpISkRERETiigKqiIiIiMQVBdQOyOVycc899+ByuWJdSoeizz029LlHnz7z2NDnHhv63NtGQlwkJSIiIiKJQyuoIiIiIhJXFFBFREREJK4ooIqIiIhIXFFAFREREZG4ooAqIiIiInFFAbWDeuCBB7BYLNx4442xLiXh7dy5kx/84AdkZmaSlJTEkCFD+Oyzz2JdVkKLRCLcdddd9OzZk6SkJHr37s29996Lmpa0rg8//JBzzz2XvLw8LBYLixYtavS8aZrcfffddOnShaSkJCZOnMiXX34Zm2ITyJE+91AoxK233sqQIUNISUkhLy+PK664gl27dsWu4ARxtJ/3g/3kJz/BYrEwf/78qNWXaBRQO6BPP/2U3//+9wwdOjTWpSS8vXv3cvLJJ+NwOPjXv/7F2rVrmTt3Lp06dYp1aQltzpw5PPHEEzz22GOsW7eOOXPm8OCDD/Loo4/GurSE4vf7GTZsGI8//niTzz/44IM88sgjLFiwgE8++YSUlBQmTZpEXV1dlCtNLEf63GtqaigsLOSuu+6isLCQl19+mQ0bNnDeeefFoNLEcrSf9/1eeeUV/ve//5GXlxelyhKUKR1KVVWV2bdvX/Odd94xJ0yYYN5www2xLimh3XrrreYpp5wS6zI6nHPOOcf80Y9+1GjbhRdeaF522WUxqijxAeYrr7zS8L1hGGZubq7529/+tmFbRUWF6XK5zOeeey4GFSamb37uTVm2bJkJmFu3bo1OUR3A4T73HTt2mF27djXXrFljdu/e3Xz44YejXlui0ApqB3PttddyzjnnMHHixFiX0iG8+uqrjB49mksuuYTs7GxGjBjBU089FeuyEt748eNZsmQJGzduBODzzz/nP//5D2eddVaMK+s4Nm/eTFFRUaPfNWlpaYwdO5alS5fGsLKOp7KyEovFQnp6eqxLSWiGYXD55Zdzyy23MGjQoFiX0+7ZY12ARM/ChQspLCzk008/jXUpHcbXX3/NE088wcyZM7n99tv59NNPuf7663E6nUyfPj3W5SWsWbNm4fP56N+/PzabjUgkwn333cdll10W69I6jKKiIgBycnIabc/JyWl4TtpeXV0dt956K9OmTcPr9ca6nIQ2Z84c7HY7119/faxLSQgKqB3E9u3bueGGG3jnnXdwu92xLqfDMAyD0aNHc//99wMwYsQI1qxZw4IFCxRQ29ALL7zA3//+d5599lkGDRrEypUrufHGG8nLy9PnLh1GKBRiypQpmKbJE088EetyEtry5cv53e9+R2FhIRaLJdblJAQd4u8gli9fTklJCSNHjsRut2O32/nggw945JFHsNvtRCKRWJeYkLp06cLAgQMbbRswYADbtm2LUUUdwy233MKsWbO49NJLGTJkCJdffjk33XQTs2fPjnVpHUZubi4AxcXFjbYXFxc3PCdtZ3843bp1K++8845WT9vYRx99RElJCQUFBQ1/Y7du3crNN99Mjx49Yl1eu6QV1A7ijDPOYPXq1Y22zZgxg/79+3Prrbdis9liVFliO/nkk9mwYUOjbRs3bqR79+4xqqhjqKmpwWpt/O9vm82GYRgxqqjj6dmzJ7m5uSxZsoThw4cD4PP5+OSTT/jpT38a2+IS3P5w+uWXX/Lee++RmZkZ65IS3uWXX37ItR2TJk3i8ssvZ8aMGTGqqn1TQO0gPB4PgwcPbrQtJSWFzMzMQ7ZL67npppsYP348999/P1OmTGHZsmU8+eSTPPnkk7EuLaGde+653HfffRQUFDBo0CBWrFjBvHnz+NGPfhTr0hJKdXU1mzZtavh+8+bNrFy5koyMDAoKCrjxxhv5zW9+Q9++fenZsyd33XUXeXl5TJ48OXZFJ4Ajfe5dunTh4osvprCwkNdff51IJNJwzm9GRgZOpzNWZbd7R/t5/+Y/BBwOB7m5ufTr1y/apSaGWLcRkNhRm6noeO2118zBgwebLpfL7N+/v/nkk0/GuqSE5/P5zBtuuMEsKCgw3W632atXL/OOO+4wA4FArEtLKO+9954JHPI1ffp00zTrW03dddddZk5OjulyucwzzjjD3LBhQ2yLTgBH+tw3b97c5HOA+d5778W69HbtaD/v36Q2U8fHYpq6tYqIiIiIxA9dJCUiIiIicUUBVURERETiigKqiIiIiMQVBVQRERERiSsKqCIiIiISVxRQRURERCSuKKCKiIiISFxRQBURERGRuKKAKiIiIiJxRQFVREREROKKAqqIiIiIxJX/D+Q8iul4bKpcAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 800x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(8, 5))\n",
    "\n",
    "ax = fig.add_subplot(111)\n",
    "\n",
    "ax.plot(r_ranges[:6], (np.array(target_F) - np.array(clue3D_F))[:6], '--s', label=\"target $\\Delta F_{0.5}$\", c=\"#D55E00\")\n",
    "ax.plot(r_ranges[:6], (np.array(model_reco_F) - np.array(clue3D_F))[:6], '-o', label=\"GNN LP $\\Delta F_{0.5}$\", c=\"#56B4E9\")\n",
    "# ax.plot(r_ranges[:6], (np.array(naive_reco_F) - np.array(clue3D_F))[:6], '-v', label=\"naive $\\Delta F_{0.5}$\", c=\"#E69F00\")\n",
    "\n",
    "ax.axhline(max(np.array(naive_reco_F) - np.array(clue3D_F)), label=\"baseline\", c=\"lightgray\", linestyle=\"--\")\n",
    "\n",
    "ax.legend()\n",
    "ax.set_xlabel(\"Neighborhood radius (cm)\")\n",
    "ax.set_ylabel(\"$\\Delta F_{0.5}$\")\n",
    "ax.set_ylim((-0.1, 0.1))\n",
    "\n",
    "fig.savefig(f'export/{ds_name}-e5-graph-lp-de-radius-plot.png', dpi=300, bbox_inches='tight', transparent=True)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9da0d004-6534-4045-bb10-801dbf764267",
   "metadata": {},
   "outputs": [],
   "source": [
    "print(\"Reco\", max(np.array(model_reco_F) - np.array(clue3D_F)))\n",
    "print(\"Naive\", max(np.array(naive_reco_F) - np.array(clue3D_F)))\n",
    "print(\"Target\", max(np.array(target_F) - np.array(clue3D_F)))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9a309e4f",
   "metadata": {},
   "source": [
    "## Setup comparison:\n",
    "\n",
    "Pairwise:\n",
    "- mean clue3d_to_sim:\tP: 0.994 R: 0.301 F: 0.652\n",
    "- mean target_to_sim:\tP: 0.992 R: 0.689 F: 0.892\n",
    "- mean reco_to_sim:\t    P: 0.659 R: 0.699 F: 0.642\n",
    "\n",
    "Graph (node-pred):\n",
    "- mean clue3d_to_sim:\tP: 0.994 R: 0.301 F: 0.652\n",
    "- mean target_to_sim:\tP: 0.992 R: 0.688 F: 0.892\n",
    "- mean reco_to_sim:\t    P: 0.659 R: 0.699 F: 0.642\n",
    "\n",
    "Graph (link-pred):\n",
    "- mean clue3d_to_sim:\tP: 0.994 R: 0.301 F: 0.652\n",
    "- mean target_to_sim:\tP: 0.992 R: 0.689 F: 0.892\n",
    "- mean reco_to_sim:\tP: 0.915 R: 0.329 F: 0.637\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "328866f1-c862-4dfb-a201-91d2baa53d9f",
   "metadata": {},
   "source": [
    "## Results\n",
    "\n",
    "Using alpha=balance seems to converge faster (100 epochs rather than 200)\n",
    "- maybe just sticking to 0.25? (works just fine)\n",
    "\n",
    "Hyperparams\n",
    "- (64, 128, noskip), alpha=balance, roc_auc = 0.9733\n",
    "- (64, 128, 256fc, noskip, 200e), alpha=1-balance, roc_auc = 0.9771\n",
    "- (64, 128, 256fc, noskip, 200e), alpha=balance, roc_auc = 0.981\n",
    "- (64, 128, 256fc, noskip, 100e), alpha=0.25, roc_auc = 0.9796"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "ve",
   "language": "python",
   "name": "ve"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.12"
  },
  "vscode": {
   "interpreter": {
    "hash": "37c238d9819f69c2c770157eac01081978c120e64661e10d7fd52c4caf977dc9"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}