Pitfall2-Example.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-03T16:37:22.560920Z",
     "start_time": "2021-02-03T16:37:16.957448Z"
    }
   },
   "outputs": [],
   "source": [
    "import random\n",
    "\n",
    "import os.path as osp\n",
    "\n",
    "import networkx as nx\n",
    "import numpy as np\n",
    "import torch\n",
    "import torch.nn.functional as F\n",
    "from torch.nn import Sequential, Linear, ReLU\n",
    "from torch_geometric.datasets import TUDataset\n",
    "from torch_geometric.data import DataLoader\n",
    "from torch_geometric.utils import from_networkx, to_networkx\n",
    "from torch_geometric.nn import GINConv, global_add_pool, MessagePassing, GraphConv, GCNConv\n",
    "\n",
    "from matplotlib import pyplot as plt\n",
    "from captum.attr import Saliency, IntegratedGradients\n",
    "from tqdm.notebook import tqdm as tq\n",
    "from collections import defaultdict"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Creating a dataset\n",
    "We create a $G_{n,p}$ random graph with $n=20$ and $p=0.2$. Initially the color for all nodes are blue (constant feature=1). We randomly select 3 nodes and assign them the color yellow(feature=2) such that there is at most on edge between two yellow nodes, therefore the explanation is unique. The task is to classify whether a graph has an edge between two yellow nodes or not."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-03T16:37:22.578036Z",
     "start_time": "2021-02-03T16:37:22.564327Z"
    }
   },
   "outputs": [],
   "source": [
    "def create_graph():\n",
    "    g=nx.erdos_renyi_graph(20,0.2)\n",
    "    while True:\n",
    "        colored_nodes = random.sample(list(g.nodes()),3)\n",
    "        label = 0\n",
    "        colored_edges = 0\n",
    "        for u in colored_nodes:\n",
    "            for v in colored_nodes:\n",
    "                if u!=v and g.has_edge(u,v):\n",
    "                    label = 1\n",
    "                    colored_edges +=1\n",
    "        if colored_edges <= 2: # at most 1 edge between colored nodes\n",
    "            break\n",
    "    data = from_networkx(g)\n",
    "    data.y = torch.tensor([label])\n",
    "    \n",
    "    data.x = torch.ones((g.number_of_nodes(),1))\n",
    "    \n",
    "    for u in colored_nodes:\n",
    "        data.x[u,0]=2\n",
    "\n",
    "    return data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## We create 1000 graphs for train/test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-03T16:37:23.479334Z",
     "start_time": "2021-02-03T16:37:22.582336Z"
    }
   },
   "outputs": [],
   "source": [
    "GRAPH_NUM = 1000\n",
    "dataset = [create_graph() for i in range(GRAPH_NUM)]\n",
    "NUM_FEATURES = 1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# GNN Model\n",
    "We use a general GNN model where can configure via initialization parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-03T16:37:26.571438Z",
     "start_time": "2021-02-03T16:37:26.558923Z"
    }
   },
   "outputs": [],
   "source": [
    "class Net(torch.nn.Module):\n",
    "    def __init__(self, num_node_features, num_classes, num_layers, concat_features, pre_processing):\n",
    "        super(Net, self).__init__()\n",
    "        dim = 32\n",
    "        self.convs = torch.nn.ModuleList()\n",
    "        self.pre_processing = pre_processing\n",
    "\n",
    "        if self.pre_processing:\n",
    "            self.pre_fc1 = Linear(num_node_features, dim)\n",
    "            self.pre_fc2 = Linear(dim, dim)\n",
    "            before_conv_dim = dim\n",
    "        else:\n",
    "            before_conv_dim = num_node_features\n",
    "        self.convs.append(GraphConv(before_conv_dim, dim))\n",
    "        for i in range(num_layers - 1):\n",
    "            self.convs.append(GraphConv(dim, dim))\n",
    "        self.concat_features = concat_features\n",
    "        if concat_features:\n",
    "            self.fc1 = Linear(dim * num_layers + before_conv_dim, dim)\n",
    "        else:\n",
    "            self.fc1 = Linear(dim, dim)\n",
    "        self.fc2 = Linear(dim, num_classes)\n",
    "\n",
    "    def forward(self, x, edge_index, batch, edge_weight=None):\n",
    "        if self.pre_processing:\n",
    "            x = F.relu(self.pre_fc1(x))\n",
    "            x = F.relu(self.pre_fc2(x))\n",
    "        xs = [x]\n",
    "        for conv in self.convs:\n",
    "            x = conv(x, edge_index, edge_weight)\n",
    "            x = F.relu(x)\n",
    "            xs.append(x)\n",
    "        if self.concat_features:\n",
    "            x = torch.cat(xs, dim=1)\n",
    "        x = global_add_pool(x, batch)\n",
    "        x = F.relu(self.fc1(x))\n",
    "        x = self.fc2(x)\n",
    "        return F.log_softmax(x, dim=-1)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-03T16:38:43.734689Z",
     "start_time": "2021-02-03T16:38:43.729058Z"
    }
   },
   "outputs": [],
   "source": [
    "test_dataset = dataset[:len(dataset) // 10]\n",
    "train_dataset = dataset[len(dataset) // 10:]\n",
    "test_loader = DataLoader(test_dataset, batch_size=64)\n",
    "train_loader = DataLoader(train_dataset, batch_size=64)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-03T16:38:44.520885Z",
     "start_time": "2021-02-03T16:38:44.508222Z"
    },
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
    "\n",
    "def train(model, optimizer):\n",
    "    model.train()\n",
    "    loss_all = 0\n",
    "    for data in train_loader:\n",
    "        data = data.to(device)\n",
    "        optimizer.zero_grad()\n",
    "        output = model(data.x, data.edge_index, data.batch)\n",
    "        loss = F.nll_loss(output, data.y)\n",
    "        loss.backward()\n",
    "        loss_all += loss.item() * data.num_graphs\n",
    "        optimizer.step()\n",
    "    return loss_all / len(train_dataset)\n",
    "\n",
    "\n",
    "def test(model, loader):\n",
    "    model.eval()\n",
    "\n",
    "    correct = 0\n",
    "    for data in loader:\n",
    "        data = data.to(device)\n",
    "        output = model(data.x, data.edge_index, data.batch)\n",
    "        pred = output.max(dim=1)[1]\n",
    "        correct += pred.eq(data.y).sum().item()\n",
    "    return correct / len(loader.dataset)\n",
    "\n",
    "def train_and_test(model):\n",
    "    optimizer = torch.optim.Adam(model.parameters(), lr=0.0003, weight_decay=5e-4)\n",
    "    pbar = tq(range(150))\n",
    "    for epoch in pbar:\n",
    "        train_loss = train(model, optimizer)\n",
    "        train_acc = test(model, train_loader)\n",
    "        test_acc = test(model, test_loader)\n",
    "        pbar.set_postfix(train_loss = train_loss,train_acc=train_acc,test_acc=test_acc)\n",
    "#         print('Epoch: {:03d}, Train Loss: {:.7f}, '\n",
    "#               'Train Acc: {:.7f}, Test Acc: {:.7f}'.format(epoch, train_loss,\n",
    "#                                                            train_acc, test_acc))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-03T16:39:11.379891Z",
     "start_time": "2021-02-03T16:39:11.362317Z"
    }
   },
   "outputs": [],
   "source": [
    "model1 = Net(num_node_features=1, num_classes=2,num_layers=3, concat_features=True, pre_processing=False).to(device)\n",
    "model1.desc='without_preprocess'\n",
    "\n",
    "model2 = Net(num_node_features=1, num_classes=2,num_layers=3, concat_features=True, pre_processing=True).to(device)\n",
    "model2.desc='with_preprocess'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-03T16:42:00.060463Z",
     "start_time": "2021-02-03T16:39:11.707661Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "80f1f0b0156c4848b9fa0563a1c18100",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=150.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7175ad3a098443f587291f175e1a9f4d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=150.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "train_and_test(model1)\n",
    "train_and_test(model2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Defining the explanation method\n",
    "Here we use saliency and integrated gradients on edge weights, we use the same method throughout the whole experiment in order to show how much model architecture can affect the explanation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-03T16:45:13.493149Z",
     "start_time": "2021-02-03T16:45:13.481657Z"
    }
   },
   "outputs": [],
   "source": [
    "def explain(model, method, data, target):\n",
    "    def model_forward(edge_mask, data):\n",
    "        batch = torch.zeros(data.x.shape[0], dtype=int)\n",
    "        out = model(data.x, data.edge_index, batch, edge_mask)\n",
    "        return out\n",
    "    input_mask = torch.ones(data.edge_index.shape[1]).requires_grad_(True).to(device)\n",
    "    if method == 'ig':\n",
    "        ig = IntegratedGradients(model_forward)\n",
    "        mask = ig.attribute(input_mask, target=target,\n",
    "                            additional_forward_args=(data,),\n",
    "                            internal_batch_size=data.edge_index.shape[1])\n",
    "    elif method == 'gradients':\n",
    "        saliency = Saliency(model_forward)\n",
    "        mask = saliency.attribute(input_mask, target=target,\n",
    "                                  additional_forward_args=(data,), abs=False)\n",
    "    else:\n",
    "        raise Exception('Unknown explanation method')\n",
    "\n",
    "    edge_mask = mask.detach().numpy()\n",
    "    \n",
    "    # we use the absolute value since it does not matter\n",
    "    # that edge weight affect the result positvely or negatively\n",
    "    edge_mask = np.abs(edge_mask)\n",
    "    if edge_mask.max() > 0:  # avoid division by zero\n",
    "        edge_mask = edge_mask / edge_mask.max()\n",
    "    \n",
    "    edge_attribution = defaultdict(float)\n",
    "    for value, u, v in list(zip(edge_mask, *data.edge_index.numpy())):            \n",
    "        edge_attribution[(min(u,v),max(u,v))] += value\n",
    "    return edge_attribution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-03T16:42:02.041611Z",
     "start_time": "2021-02-03T16:42:02.036836Z"
    }
   },
   "outputs": [],
   "source": [
    "def make_colorblind_friendly(node_color):\n",
    "    colors = []\n",
    "    for value in node_color:\n",
    "        colors.append('#FFB000' if value == 0 else '#648FFF') # Based on IBM palette\n",
    "    return colors"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Visualizing the explanations\n",
    "Here we pick a single graph from the dataset, and try to explain the prediction for both models via the same explanation method (Integrated Gradients or Edge Gradients)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-03T16:45:15.764152Z",
     "start_time": "2021-02-03T16:45:15.146523Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x432 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "dds = random.choice([data for data in dataset if data.y.item()])\n",
    "METHOD = 'ig' # you can change this to gradients\n",
    "plt.figure(figsize=(15,6))\n",
    "g = to_networkx(dds).to_undirected()\n",
    "pos = nx.kamada_kawai_layout(g)\n",
    "for idx, model in enumerate([model1,model2]):\n",
    "    model.eval()\n",
    "    edge_attribution = explain(model, METHOD, dds,target=dds.y.item())\n",
    "\n",
    "    plt.subplot(1,2,idx+1)\n",
    "    plt.title(model.desc)\n",
    "    node_color = -dds.x.numpy()\n",
    "    edge_color = [edge_attribution[(u,v)] for u,v in g.edges()]\n",
    "    nx.draw(g, pos=pos, node_color=make_colorblind_friendly(node_color), edge_color=edge_color, edge_cmap=plt.get_cmap('Greys'))\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Save plots as pdf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-02T20:16:20.671372Z",
     "start_time": "2021-02-02T20:16:19.446412Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for idx, model in enumerate([model1,model2]):\n",
    "#     pp = PdfPages(model.desc+'.pdf')\n",
    "    f = plt.figure()\n",
    "    plt.clf()\n",
    "    model.eval()\n",
    "    edge_attribution = explain(model, METHOD, dds,target=dds.y.item())\n",
    "    node_color = -dds.x.numpy()\n",
    "    edge_color = [edge_attribution[(u,v)] for u,v in g.edges()]\n",
    "    nx.draw(g, pos=pos, node_color=make_colorblind_friendly(node_color), edge_color=edge_color, edge_cmap=plt.get_cmap('Greys'))\n",
    "    plt.show()\n",
    "    f.savefig(model.desc+\".pdf\")\n",
    "#     pp.savefig()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Quantifiying the differences\n",
    "We calculate the attribution value for all unimportant edges and we expect them to be close to 0, we plot the distribution of attribution value for unimportant edges on all graphs for both models. We also consider an explanation correct, if the most important edge in explanation is the edge between two yellow nodes and use that to calculate the explanation accuracy."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-03T16:45:33.620213Z",
     "start_time": "2021-02-03T16:45:33.611451Z"
    }
   },
   "outputs": [],
   "source": [
    "def calculate_avg_wrong(model):\n",
    "    METHOD = 'ig'\n",
    "    model.eval()\n",
    "    avg_wrong_masks = []\n",
    "    accuracy = []\n",
    "\n",
    "    for dds in tq(dataset):\n",
    "        if not dds.y.item():\n",
    "            continue\n",
    "        mask_sum = []\n",
    "        edge_attribution = explain(model, METHOD, dds,target=dds.y.item())\n",
    "        for (u, v), value in edge_attribution.items():\n",
    "            if dds.x[u].item() != 1.0 and dds.x[v].item() != 1.0: # skip the important edge\n",
    "                right_edge_value = value\n",
    "                continue\n",
    "            mask_sum.append(value)\n",
    "        if max(edge_attribution.values()) == right_edge_value:\n",
    "            accuracy.append(1)\n",
    "        else:\n",
    "            accuracy.append(0)\n",
    "        avg_wrong_masks.append(np.mean(mask_sum))\n",
    "    return avg_wrong_masks, accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-03T16:45:40.890179Z",
     "start_time": "2021-02-03T16:45:38.167167Z"
    },
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "eca6ec8651eb4614885adecfdaa01d79",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=1000.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ba357c3130c34396a7b415ca108db4ed",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=1000.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "avg_wrong_masks_model1, accuracy_model1 = calculate_avg_wrong(model1)\n",
    "avg_wrong_masks_model2, accuracy_model2 = calculate_avg_wrong(model2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-03T16:47:40.847562Z",
     "start_time": "2021-02-03T16:47:40.840571Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Explanation accuracy for model without_preprocess 1.0\n",
      "Explanation accuracy for model with_preprocess 1.0\n"
     ]
    }
   ],
   "source": [
    "print('Explanation accuracy for model ' + model1.desc, np.mean(accuracy_model1))\n",
    "print('Explanation accuracy for model ' + model2.desc, np.mean(accuracy_model2))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Altough the explanation method can detect the most important edge correctly in both models, it generally assigns higher value to unimportant edges when the model does not preprocess the node features."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-03T16:47:12.170891Z",
     "start_time": "2021-02-03T16:47:11.513126Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x13d32b160>"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZAAAAEGCAYAAABLgMOSAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3de3wU9b3/8dcHRCOIKJdDVdRgRTwKIUgMIEYRqgLeFazIEVB7KC2tWls9tNYjtbY/qNb2VK1IK0V7rEVU1OMdRSt4QYMiFy8FNGo4nIog4KWo6Of3x3wThrCbbCbZ7Ca8n4/HPjL7ne/Mfr4zm/3szOx8v+buiIiI1FerXAcgIiLNkxKIiIgkogQiIiKJKIGIiEgiSiAiIpLILrkOoDF17tzZCwsLcx2GiEizsXjx4g/cvUuSZVtUAiksLKS8vDzXYYiINBtm9k7SZXUKS0REElECERGRRJRAREQkkRZ1DUREtvniiy+orKxky5YtuQ5F8kBBQQHdunWjTZs2jbZOJRCRFqqyspL27dtTWFiImeU6HMkhd2f9+vVUVlbSvXv3RluvTmGJtFBbtmyhU6dOSh6CmdGpU6dGPxrNWgIxs/3N7Ckze83MVpjZxaG8o5nNM7OV4e/eaZYfF+qsNLNx2YpTpCVT8pAq2XgvZPMIZCvwQ3c/DBgATDKzw4DJwJPu3gN4Mjzfjpl1BK4C+gOlwFXpEo2IiORG1q6BuPtaYG2Y/sjMXgf2A04DBodqtwFPA/9RY/ETgXnuvgHAzOYBw4A7sxWvSEtXOPmhRl1fxdSTGnV90vw0yTUQMysE+gKLgK4huQD8H9A1xSL7Ae/FnleGslTrnmBm5WZWvm7dukaLuVmY0mHbIxvL1HfdIvU0YsQINm7cyMaNG/n9739fXf70009z8sknN8prPP300zz33HONsi7ZXtYTiJntAdwDXOLum+PzPBoOsUFDIrr7DHcvcfeSLl0SdeciIjny8MMPs9dee+2QQBpTYyWQrVu3Nmh5d+err75qcBz5JKsJxMzaECWPO9z93lD8DzPbJ8zfB3g/xaJrgP1jz7uFMhFpRq699lp+97vfAfCDH/yAIUOGADB//nzGjBlDYWEhH3zwAZMnT2b16tUUFxdz2WWXAfDxxx8zcuRIDj30UMaMGUPV8NtPPvkkffv2pXfv3lxwwQV89tlnANXrAigvL2fw4MFUVFQwffp0fvOb31BcXMyCBQtSxjl+/HgmTpxISUkJhxxyCA8++CAAs2bN4tRTT2XIkCEMHTqUTz75hAsuuIDS0lL69u3L/fffX13vtNNOY/DgwfTo0YOf/exnAFRUVNCzZ0/Gjh1Lr169eO+997jsssvo1asXvXv3Zvbs2dUxTJs2jd69e9OnTx8mT44uDa9evZphw4bRr18/ysrKeOONNwCYM2cOvXr1ok+fPhxzzDEArFixgtLSUoqLiykqKmLlypWNsQtrlbVrIBZd8r8VeN3dr4/NegAYB0wNf+9PsfhjwC9jF85PAH6crVhFJDvKysr49a9/zUUXXUR5eTmfffYZX3zxBQsWLOCYY47h2WefBWDq1KksX76cJUuWANFRwyuvvMKKFSvYd999GTRoEM8++ywlJSWMHz+eJ598kkMOOYSxY8dy8803c8kll6R8/cLCQiZOnMgee+zBj370o1pjraio4MUXX2T16tUcd9xxrFq1CoCXX36ZpUuX0rFjR37yk58wZMgQZs6cycaNGyktLeUb3/gGAC+++CLLly+nbdu2HHnkkZx00kl07tyZlStXcttttzFgwADuuecelixZwquvvsoHH3zAkUceyTHHHMOSJUu4//77WbRoEW3btmXDhg0ATJgwgenTp9OjRw8WLVrEd7/7XebPn8/VV1/NY489xn777cfGjRsBmD59OhdffDFjxozh888/58svv2z4DqxDNo9ABgHnAUPMbEl4jCBKHMeb2UrgG+E5ZlZiZn8ECBfPfw68FB5XV11QF5Hmo1+/fixevJjNmzez2267MXDgQMrLy1mwYAFlZWW1LltaWkq3bt1o1aoVxcXFVFRU8Oabb9K9e3cOOeQQAMaNG8czzzzTKLGeffbZtGrVih49enDQQQdVf9s//vjj6dixIwCPP/44U6dOpbi4mMGDB7Nlyxbefffd6nqdOnVi991358wzz2ThwoUAHHjggQwYMACAhQsXMnr0aFq3bk3Xrl059thjeemll3jiiSc4//zzadu2LQAdO3bk448/5rnnnmPUqFEUFxfz7W9/m7Vro8vHgwYNYvz48fzhD3+oThQDBw7kl7/8JdOmTeOdd95h9913b5TtUpts/gprIZDuh8dDU9QvB74Vez4TmJmd6ESkKbRp04bu3bsza9YsjjrqKIqKinjqqadYtWoV//qv/1rrsrvttlv1dOvWreu8BrHLLrtUX2NIcsNczfskqp63a9euuszdueeee+jZs+d2dRctWpTR8vXx1Vdfsddee1UflcVNnz6dRYsW8dBDD1Un6XPPPZf+/fvz0EMPMWLECG655ZbqU4bZoq5MRHYSufrZbVlZGddddx0zZ86kd+/eXHrppfTr12+7D9z27dvz0Ucf1bmunj17UlFRwapVqzj44IP585//zLHHHgtEp6sWL17M8OHDueeee7Zb9+bNm9OtstqcOXMYN24cb7/9Nm+99RY9e/bklVde2a7OiSeeyA033MANN9yAmfHKK6/Qt29fAObNm8eGDRvYfffdue+++5g5c8fvv2VlZdxyyy2MGzeODRs28Mwzz3Dttdey6667cvXVVzNmzJjqU1gdO3ake/fuzJkzh1GjRuHuLF26lD59+rB69Wr69+9P//79eeSRR3jvvffYtGkTBx10EBdddBHvvvsuS5cuzXoCUVcmIpJVZWVlrF27loEDB9K1a1cKCgp2OH3VqVMnBg0aRK9evaovoqdSUFDAn/70J0aNGkXv3r1p1aoVEydOBOCqq67i4osvpqSkhNatW1cvc8oppzB37txaL6IDHHDAAZSWljJ8+HCmT59OQUHBDnWuvPJKvvjiC4qKijj88MO58sorq+eVlpZy1llnUVRUxFlnnUVJSckOy59xxhkUFRXRp08fhgwZwq9+9Su+9rWvMWzYME499VRKSkooLi7muuuuA+COO+7g1ltvpU+fPhx++OHVF+0vu+wyevfuTa9evTjqqKPo06cPd911F7169aK4uJjly5czduzYtG1tLFb1y4aWoKSkxHeqEQnj92hM2dT4y1TVzXTdkldef/31Ok8TSWT8+PGcfPLJjBw5MtHys2bNory8nBtvvLGRI2tcqd4TZrbY3XfMdhnQEYiIiCSiayDScEmOhERy4Be/+AVz5szZrmzUqFHMmjWrQesdP34848ePb9A6miMlEBHZaVxxxRVcccUVuQ6jxdApLBERSUQJREREElECERGRRHQNRGRn0dhd8+sHEzs9HYGISM40xXggkj1KICKSM00xHkiVho7n0dDlWyIlEBHJmmyMB5JKYWEhl19+Ob1796a0tLS6K/aqcT769+/P5ZdfnnZ8jUzHA9mwYQOnn346RUVFDBgwgKVLl1bHev7559O7d2+Kioqq++J6/PHHGThwIEcccQSjRo3i448/BmDy5MkcdthhFBUVVXczn2qMj3ynayAikjWNPR7I0Ucfnfa1OnTowLJly7j99tu55JJLqpNAZWUlzz33HK1bt2bo0KEpx9eAzMYD+f73v0/fvn257777mD9/PmPHjmXJkiX8/Oc/r359gA8//JAPPviAa665hieeeIJ27doxbdo0rr/+eiZNmsTcuXN54403MLPq8TxSjfGR73QEsrOoa3xzjX8uWdDY44HUZvTo0dV/n3/++eryUaNG0bp161rH14DMxgNZuHAh5513HgBDhgxh/fr1bN68mSeeeIJJkyZVr2vvvffmhRde4LXXXmPQoEEUFxdz22238c4779ChQwcKCgq48MILuffee6vHAEk1xke+0xGIiGRNU44HEu8ePj5dNR5HbeNr1Fwm/jzpeB7uzvHHH8+dd965w7wXX3yRJ598krvvvpsbb7yR+fPnpxzjo1OnToleu6lkc0jbmcDJwPvu3iuUzQaqRmLZC9jo7sUplq0APgK+BLYm7SlSRGJy9LPbxhwPpDazZ89m8uTJzJ49m4EDB+4wf88990w7vgZkNh5IWVkZd9xxB1deeSVPP/00nTt3Zs899+T444/npptu4re//S0QncIaMGAAkyZNqh675JNPPmHNmjXsu+++fPrpp4wYMYJBgwZx0EEHAaQc42OnTSDALOBG4PaqAnf/ZtW0mf0aqO0dfZy7f5C16ESkSZSVlfGLX/yCgQMH0q5duzrHAxk+fDgnnVT/wa8+/PBDioqK2G233VJ+64dofI3vfOc7XHPNNXzxxRecc8451QmkajyQzZs3px0PZMqUKVxwwQUUFRXRtm1bbrvtNgB++tOfMmnSJHr16kXr1q256qqrOPPMM5k1axajR4/ms88+A+Caa66hffv2nHbaaWzZsgV35/rrrweiMT5WrlyJuzN06NDquPJZVscDMbNC4MGqI5BYuQHvAkPcfWWK5SqAkvomEI0HkmHdVMukGvsj0/FA1BtvXtqZxgMpLCykvLyczp07J1q+oeOBNBctZTyQMuAfqZJH4MDjZrbYzCY0YVwiIpKhXF1EHw2kPsaMHO3ua8zsX4B5ZvaGuz+TqmJIMBMgOgQVkZbrjDPO4O23396ubNq0aXX+QqsuDR0PZGfV5AnEzHYBzgT6pavj7mvC3/fNbC5QCqRMIO4+A5gB0SmsRg9YpBlz9x1+XdSczZ07N9chNFvZuFyRi1NY3wDecPfKVDPNrJ2Zta+aBk4AljdhfCItQkFBAevXr8/KB4c0L+7O+vXrU/4woCGy+TPeO4HBQGczqwSucvdbgXOocfrKzPYF/ujuI4CuwNzwrWkX4C/u/mi24hRpqbp160ZlZSXr1q3LdSiSBwoKCujWrVujrjNrCcTdR6cpH5+i7H+BEWH6LSD/f78mkueqbuITyRbdid6SZaNrEv1kV0QC9YUlIiKJKIGIiEgiSiAiIpKIEoiIiCSiBCIiIokogYiISCJKICIikogSiIiIJKIEIiIiiehO9OamMe8uz8ad6iKy09ARiIiIJKIEIiIiiSiBiIhIIkogIiKSiBKIiIgkogQiIiKJZC2BmNlMM3vfzJbHyqaY2RozWxIeI9IsO8zM3jSzVWY2OVsxiohIctk8ApkFDEtR/ht3Lw6Ph2vONLPWwE3AcOAwYLSZHZbFOEVEJIGsJRB3fwbYkGDRUmCVu7/l7p8DfwVOa9TgRESkwXJxJ/r3zGwsUA780N0/rDF/P+C92PNKoH+6lZnZBGACwAEHHNDIoTZD+X53eYIx1QsnP1Q9XTH1pMaOSEQSauqL6DcDXweKgbXArxu6Qnef4e4l7l7SpUuXhq5OREQy1KQJxN3/4e5fuvtXwB+ITlfVtAbYP/a8WygTEZE80qQJxMz2iT09A1ieotpLQA8z625muwLnAA80RXwiIpK5rF0DMbM7gcFAZzOrBK4CBptZMeBABfDtUHdf4I/uPsLdt5rZ94DHgNbATHdfka04RUQkmawlEHcfnaL41jR1/xcYEXv+MLDDT3xFRCR/6E50ERFJRAlEREQSUQIREZFElEBERCQRJRAREUkkF12ZiCSmbk1E8oeOQEREJBElEBERSUQJREREElECERGRRJRAREQkESUQERFJRAlEREQSUQIREZFElEBERCQR3Yku25vSIbOyPFB1V7ruSBfJDR2BiIhIIllLIGY208zeN7PlsbJrzewNM1tqZnPNbK80y1aY2TIzW2Jm5dmKUUREksvmEcgsYFiNsnlAL3cvAv4O/LiW5Y9z92J3L8lSfCIi0gBZSyDu/gywoUbZ4+6+NTx9AeiWrdcXEZHsyuU1kAuAR9LMc+BxM1tsZhNqW4mZTTCzcjMrX7duXaMHKSIiqeUkgZjZFcBW4I40VY529yOA4cAkMzsm3brcfYa7l7h7SZcuXbIQrYiIpNLkCcTMxgMnA2Pc3VPVcfc14e/7wFygtMkCFBGRjDRpAjGzYcDlwKnu/mmaOu3MrH3VNHACsDxVXRERyZ1s/oz3TuB5oKeZVZrZhcCNQHtgXviJ7vRQd18zezgs2hVYaGavAi8CD7n7o9mKU0REksnanejuPjpF8a1p6v4vMCJMvwX0yVZcIiLSONSVST6LdyEyZVPmdXMpZVcodcSeQlU3JQ0RX4e6OxFpfOrKREREElECERGRRDJKIGY2KJMyERHZeWR6BHJDhmUiIrKTqPUiupkNBI4CupjZpbFZewKtsxmYiIjkt7p+hbUrsEeo1z5WvhkYma2gREQk/9WaQNz9b8DfzGyWu7/TRDGJiEgzkOl9ILuZ2QygML6Muw/JRlAiIpL/Mk0gc4DpwB+BL7MXjoiINBeZJpCt7n5zViORSLo7yvPlTnMBdJe7CGT+M97/MbPvmtk+Ztax6pHVyEREJK9legQyLvy9LFbmwEGNG46IiDQXGSUQd++e7UBERKR5ySiBmNnYVOXufnvjhiMiIs1FpqewjoxNFwBDgZcBJRARkZ1Upqewvh9/bmZ7AX/NSkQiItIsJO3O/ROgzusiZjbTzN43s+Wxso5mNs/MVoa/e6dZdlyos9LMxqWqIyIiuZNpd+7/Y2YPhMdDwJvA3AwWnQUMq1E2GXjS3XsAT4bnNV+vI3AV0B8oBa5Kl2hERCQ3Mr0Gcl1seivwjrtX1rWQuz9jZoU1ik8DBofp24Cngf+oUedEYJ67bwAws3lEiejODOMVEZEsy+gIJHSq+AZRj7x7A5834DW7uvvaMP1/QNcUdfYD3os9rwxlOzCzCWZWbmbl69ata0BY0tQaY9zzbCic/FD1Q0TSy/QU1tnAi8Ao4GxgkZk1uDt3d3eiGxIbso4Z7l7i7iVdunRpaEgiIpKhTE9hXQEc6e7vA5hZF+AJ4O4Er/kPM9vH3dea2T7A+ynqrGHbaS6AbkSnukREJE9k+iusVlXJI1hfj2VreoBtXaOMA+5PUecx4AQz2ztcPD8hlImISJ7I9AjkUTN7jG0Xsb8JPFzXQmZ2J9GRRGczqyT6ZdVU4C4zuxB4h+iUGGZWAkx092+5+wYz+znwUljV1VUX1EVEJD/UNSb6wUQXvS8zszOBo8Os54E76lq5u49OM2toirrlwLdiz2cCM+t6DRERyY26jkB+C/wYwN3vBe4FMLPeYd4pWY1ORETyVl3XMbq6+7KahaGsMCsRiYhIs1BXAtmrlnm7N2YgIiLSvNSVQMrN7N9rFprZt4DF2QlJRESag7qugVwCzDWzMWxLGCXArsAZ2QxMRETyW60JxN3/ARxlZscBvULxQ+4+P+uRSf6b0qFR6yXtOiS+XMXUkxKto76vk3S5+sTXVO0SSSrT8UCeAp7KciwiItKMJL2bXEREdnJKICIikogSiIiIJKIEIiIiiSiBiIhIIkogIiKSiBKIiIgkogQiIiKJZDqglGRbpnd1S63S3S2errzqDu/6LpeqTvxu8UzWl/Tuct2hLvlCRyAiIpJIkycQM+tpZktij81mdkmNOoPNbFOszn82dZwiIlK7Jj+F5e5vAsUAZtYaWAPMTVF1gbuf3JSxiYhI5nJ9CmsosNrd38lxHCIiUk+5TiDnAHemmTfQzF41s0fM7PB0KzCzCWZWbmbl69aty06UIiKyg5wlEDPbFTgVmJNi9svAge7eB7gBuC/detx9hruXuHtJly5dshOsiIjsIJdHIMOBl8OgVdtx983u/nGYfhhoY2admzpAERFJL5cJZDRpTl+Z2dfMzMJ0KVGc65swNhERqUNObiQ0s3bA8cC3Y2UTAdx9OjAS+I6ZbQX+CZzj7p6LWEVEJLWcJBB3/wToVKNsemz6RuDGpo5Ldj5JxznPlmzGozvYpbHl+ldYIiLSTCmBiIhIIkogIiKSiBKIiIgkogQiIiKJKIGIiEgiSiAiIpKIEoiIiCSiBCIiIokogYiISCI56cpkpzalQ2x6U+7iyAMVBefCFNJuh4qCcwEo3PKXpguqGavqqqS+3ZSoixNJSkcgIiKSiBKIiIgkogQiIiKJKIGIiEgiSiAiIpKIEoiIiCSSswRiZhVmtszMlphZeYr5Zma/M7NVZrbUzI7IRZwiIpJaru8DOc7dP0gzbzjQIzz6AzeHvyIikgfy+RTWacDtHnkB2MvM9sl1UCIiEsnlEYgDj5uZA7e4+4wa8/cD3os9rwxla+OVzGwCMAHggAMOyF60VepzJ3lddePzd2ZhO1QU1H3XedXd6XHN9U71xrgDPL6O2srqu754PNm8U70+69Yd8/knl0cgR7v7EUSnqiaZ2TFJVuLuM9y9xN1LunTp0rgRiohIWjlLIO6+Jvx9H5gLlNaosgbYP/a8WygTEZE8kJMEYmbtzKx91TRwArC8RrUHgLHh11gDgE3uvhYREckLuboG0hWYa2ZVMfzF3R81s4kA7j4deBgYAawCPgXOz1GsIiKSQk4SiLu/BfRJUT49Nu3ApKaMS0REMpfPP+MVEZE8pgQiIiKJKIGIiEgiSiAiIpJIrvvCEtlOqjvNU5W1NA25e7w5vF6+vHac7mxvOB2BiIhIIkogIiKSiBKIiIgkogQiIiKJKIGIiEgiSiAiIpKIEoiIiCSiBCIiIokogYiISCJKICIikoi6MqlpSofY9Kb6L1efZUQylC/df9RHQ2KuWjZdFyOZdENS1zrquz7ZkY5AREQkkSZPIGa2v5k9ZWavmdkKM7s4RZ3BZrbJzJaEx382dZwiIlK7XJzC2gr80N1fNrP2wGIzm+fur9Wot8DdT85BfCIikoEmPwJx97Xu/nKY/gh4HdivqeMQEZGGyek1EDMrBPoCi1LMHmhmr5rZI2Z2eC3rmGBm5WZWvm7duixFKiIiNeUsgZjZHsA9wCXuvrnG7JeBA929D3ADcF+69bj7DHcvcfeSLl26ZC9gERHZTk4SiJm1IUoed7j7vTXnu/tmd/84TD8MtDGzzk0cpoiI1CIXv8Iy4FbgdXe/Pk2dr4V6mFkpUZzrmy5KERGpSy5+hTUIOA9YZmZLQtlPgAMA3H06MBL4jpltBf4JnOPunoNYRUQkjSZPIO6+ELA66twI3Ng0EWUofod6knqZLi/1VlFwbvV04Za/pCxPNb8+68pomSmpl0myzkzia6x1ppKLO9+Tvma+3EWeLv6kMeVLu2qjO9FFRCQRJRAREUlECURERBJRAhERkUSUQEREJBElEBERSUQJREREElECERGRRJRAREQkEY2JXkV3iu8UUt0Vnuru7nTldd3pXtdd5011J3k6Va+frh2ptkm8bqq7reuznerb1sa+Iz6T9e1wB3j4bEi5P+sx3nom47tnElMqubpTXUcgIiKSiBKIiIgkogQiIiKJKIGIiEgiSiAiIpKIEoiIiCSiBCIiIonkJIGY2TAze9PMVpnZ5BTzdzOz2WH+IjMrbPooRUSkNk2eQMysNXATMBw4DBhtZofVqHYh8KG7Hwz8BpjWtFGKiEhdcnEEUgqscve33P1z4K/AaTXqnAbcFqbvBoaaWa3jqIuISNMyd2/aFzQbCQxz92+F5+cB/d39e7E6y0OdyvB8dajzQYr1TQAmhKc9gTfrEU5nYId1tgAtsV1qU/OgNjUP8TYd6O5dkqyk2feF5e4zgBlJljWzcncvaeSQcq4ltkttah7UpuahsdqUi1NYa4D9Y8+7hbKUdcxsF6ADsL5JohMRkYzkIoG8BPQws+5mtitwDvBAjToPAOPC9Ehgvjf1uTYREalVk5/CcvetZvY94DGgNTDT3VeY2dVAubs/ANwK/NnMVgEbiJJMNiQ69dUMtMR2qU3Ng9rUPDRKm5r8IrqIiLQMuhNdREQSUQIREZFEWmwCSdpdipkVmtk/zWxJeExv6tjTyaBNx5jZy2a2NdxvE583zsxWhse4msvmSgPb9GVsP9X8IUbOZNCmS83sNTNbamZPmtmBsXnNdT/V1qa83E+QUbsmmtmyEPvCeK8ZZvbjsNybZnZi00aeXtI2Jfrsc/cW9yC6OL8aOAjYFXgVOKxGne8C08P0OcDsMF0ILM91GxK2qRAoAm4HRsbKOwJvhb97h+m9m3ObwryPc92GhG06Dmgbpr8Te+815/2Usk35up/q0a49Y9OnAo+G6cNC/d2A7mE9rZt5m+r92ddSj0BaYncpdbbJ3SvcfSnwVY1lTwTmufsGd/8QmAcMa4qg69CQNuWrTNr0lLt/Gp6+QHQvFDTv/ZSuTfksk3Ztjj1tB1T96ug04K/u/pm7vw2sCuvLtYa0qd5aagLZD3gv9rwylKWs4+5bgU1ApzCvu5m9YmZ/M7OybAeboUzalI1ls6mhcRWYWbmZvWBmpzduaInVt00XAo8kXLapNKRNkJ/7CTJsl5lNCt0p/Qq4qD7L5kBD2gT1/Oxr9l2ZZMFa4AB3X29m/YD7zOzwGllb8sOB7r7GzA4C5pvZMndfneugMmVm/waUAMfmOpbGkqZNzXo/uftNwE1mdi7wU7bd5NxspWlTvT/7WuoRSOLuUsIh6XoAd19MdD7xkKxHXLdM2pSNZbOpQXG5+5rw9y3gaaBvYwaXUEZtMrNvAFcAp7r7Z/VZNgca0qZ83U9Q/+39V6DqCKpZ76uY6jYl+uzL9UWfLF1I2oXoAmR3tl1IOrxGnUlsfxH9rjDdhXAxjOhC1BqgY3NoU6zuLHa8iP420YXZvcN0c2/T3sBuYbozsJIaFwvztU1EH6CrgR41ypvtfqqlTXm5n+rRrh6x6VOIessAOJztL6K/RX5cRG9Im+r92ZfznZjFDTkC+Ht4U18Ryq4m+nYEUADMIbr49SJwUCg/C1gBLAFeBk7JdVvq0aYjic55fkLU+eSK2LIXhLauAs7PdVsa2ibgKGBZ+AdZBlyY67bUo01PAP8I77ElwAMtYD+lbFM+76cM2/Vfsc+Dp+IfxkRHW6uJhpAYnuu2NLRNST771JWJiIgk0lKvgYiISJYpgYiISCJKICIikogSiIiIJKIEIiIiiSiB5BEzO93M3GIryMIAAAfeSURBVMwOzXUsSZnZJWbWNvb8J3XUf9jM9go9gS6v52sNNrOjYs8nmtnY+kfdMElir2Vdh4aeUF8xs683xjozfN3nmuq1wusVhrugG7KO7d5rCWNolP22s1ICyS+jgYXhb4OZWevGWE89XQLE/6lTJhCLtHL3Ee6+MeFrDSa6zwAAd5/u7rcnXFe+OB242937egbdfVRtx4a+qLsfVXetxhF6figEGpRA2PG9Jk0t1ze96FF9888eRHd+HgK8GcqGAXNidQYDD4bpE4DniW74mQPsEcorgGmh/Bzg34GXiG7kuodtXW5/najX1GXANcS63AYuC8ssBX6WJt6bgXKiG49+FsouAj4P63wKmAp8SXRj0h1EHxpvEnXNvgI4MMTbOcx7I9R7naiH5LaxNnUO0yVE3WEUAv8XttkSoAyYAvwo1CsO7VsKzCV0ix6WnUZ08+jfgbIUbfsrcFLs+SxgZHjNBWHbvgwcFeYXErrBBsYDN8aWfRAYXNs+i9UdEWvTU6HsUmB5eFwSe73ttmON9eywvcL0FGBm2AZvARfFlvk49h77G3B/qDMVGBO21zLg67FtMp3oPfB34ORQXgD8KdR9BTgutl0eAOaH9b9A1IHpEuAHtWzbwSHeu9n2/jBqvNdS7MN+4XUWA48B+8TKXw2Pa2P7rS1wF/Aa0ftlEVBSx//a1FB/KXBdrj9DcvK5lesA9Ag7IvonvTVMPxfe6LsA7wLtQvnNwL8RfeA+Eyv/D+A/w3QFcHlsvZ1i09cA3w/TDwKjw/TE2AfICcCM8E/aKtQ7JkW8HcPf1uEfvCj2+p1j9eKJqZCoW/YBsbIKtiUQBwaF8plsSwbV62THD8QfxdY1JbbMUuDYMH018Nsw/TTw6zA9AngiRdvOAG4L07sS9W66e/iQKQjlPdjWBUQhdSSQ2vZZjdeOt6Ef0QdkO6IvGCuIugzZYTvWWEdt2+s5ou43OhPd2d8mvp9CrBuBfUK9NWz7gnBxbDvOAh4leo/0IOotoAD4ITAz1DmU6P1bELZLJdveN4MJX4bC83TbdjBRoukWXut54OhU77XYutqEdnYJz78Zi2kp4f3M9gnkR8AtYboXsDVsu5T7jajn7jeh+mbsvXL9GZKLh05h5Y/RRN98CX9He9TN/KPAKeGw/ySib4YDiAa0edbMlhD1pHlgbF2zY9O9zGyBmS0jSlKHh/KBRN+mAP4Sq39CeLxC9I3rUKJ/6JrONrOXQ73DQzyZeMfdX0gz7z13fzZM/zdwdIbr3I6ZdSD6h/5bKLoNOCZW5d7wdzHRh3FNjwDHmdluwHDgGXf/J9EH0x/CtpxD5m2GuvdZKkcDc939E3f/OMRd1cV2bduxNg951GneB8D7QNcUdV5y97UedYi4Gng8lC9j++11l7t/5e4riY5WDg0x/zeAu78BvMO2DvnmufuGNHHVtm1fdPdKd/+K6IilMMXycT2JksC8sK1/CnQzs72I3hfPhHp/ji1zNOH/z92XEyUaSL/fNgFbgFvN7EzgU3ZC6s49D5hZR2AI0NvMnOhbvZvZZURv6u8BG4i+lX0UBr6a5+7prpV8EpueBZzu7q+a2Xiib3S1hgP8P3e/pZZ4uxN9YzvS3T80s1lE3zIz8Ukt82r2q1P1fCvbrtdl+jq1qeop9ktS/A+4+xYze5pogKdvsi2x/4Cov6c+IZ4tKdYdjzUeb137rL5q2461ba/PYtMp21+jzlex51/VqJ9uf6VTW8y1bdtMYo4zoj7TBm5XGCWQ+kq738ysFBhKdHrze0T/wzsVHYHkh5HAn939QHcvdPf9iXpiLSM6j3sE0bWMqg+yF4BBZnYwgJm1M7N03S63B9aaWRuiI5AqLxB1ngbRtZIqjwEXmNkeYd37mdm/1FjnnkQfBpvMrCvRt/QqH4XXrPJFeO1MHGBmVf/05xL9oACiUxX9wvRZsfo1XwsAd98EfBgbEOc8ou1YH7OB84n2waOhrAOwNnwTPo8o0ddUARSbWSsz259to9TVZ59VWQCcbmZtzawd0am1BRnEXkHq7dXYRoV2fp2o99Y3ieIbAxDad0Aor6nmvstk29a1jipvAl2q3ktm1iaMa7ER2GhmVUe28f+HZ4GzQ/3DgN6hPOV+C/8fHdz9YaLk1yeDeFscJZD8MJrowl3cPUSnsb4kOo8+PPzF3dcRnVO+08yWEp0XTvfT3yuJLgg+S3QRssolwKVh+YOJDslx98eJTmk9H04n3E2Nf1J3f5Xo1NUboe6zsdkzgEfN7KnY86VmdkedWyH6x59kZq8TdQN+cyj/GfBfZlZO9A20yv8AZ4SfvdYcPW0ccG1oXzHRdZD6eJxoUKQnPBoaFOD3wDgze5Voe6f6Rv0sUfJ/Dfgd0WnA+u4zwjIvEx1Bvki0D//o7q9kEHu67dXY3g2xPQJMdPctRNuoVXjvzAbGe2xskJilwJdm9qqZ/YDMtm1NNd9rAIT9NRKYFta3hG2/1jufaCClJURHF1V+T5R0XiO6VrgC2FTLfmsPPBjKFhL92GGno954d1Lh9/P/dHc3s3OIklXNceNFUgqnLR9097tzHUtjCD95bxNOX36dqHv6nrEvD5KCroHsvPoBN4brKRuJxqEQ2Vm1BZ4Kp1sN+K6SR910BCIiIonoGoiIiCSiBCIiIokogYiISCJKICIikogSiIiIJPL/AQ3nEr+B5vJSAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(avg_wrong_masks_model1, bins=100, label=model1.desc)\n",
    "plt.hist(avg_wrong_masks_model2, bins=100, label=model2.desc)\n",
    "plt.ylabel('Count')\n",
    "plt.xlabel('Average attribution value for unimportant edges')\n",
    "plt.legend()"
   ]
  }
 ],
 "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": "3.8.3"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}