repoduction_with_boston_housing.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "c30661e9-6256-4345-813c-869e5aeb97ce",
   "metadata": {},
   "source": [
    "# Fooling LIME and SHAP\n",
    "## Reproduction (10)\n",
    "\n",
    "Implement the approach by writting a simple interface/framework and confirm yiur implementation by using any (tabular) raciscm dataset (e.g. Boston Housing)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "50e4479b-2610-4ac4-be4a-b5197590f0b8",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "\n",
    "import pandas as pd\n",
    "from sklearn.ensemble import RandomForestRegressor\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "\n",
    "from fooling_LIME_SHAP_Interface.adversarial_model_toolbox import AdversarialModelToolbox\n",
    "from fooling_LIME_SHAP_Interface.util.explainer_type import ExplainerType\n",
    "from fooling_LIME_SHAP_Interface.util.ml_type import MLType"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d1a77f8f-a355-4d0c-bd3d-85e43b946d1d",
   "metadata": {},
   "source": [
    "First we define all the neccessary variables."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "e68705fa-c888-4d2a-88b9-22d290655ea2",
   "metadata": {},
   "outputs": [],
   "source": [
    "TRAIN_TEST_SPLIT = 0.2\n",
    "SEED = 666\n",
    "\n",
    "dataset_name = 'HousingData'\n",
    "input_ids = [5, 7, 11, 12]\n",
    "categorical_input_ids = []\n",
    "output_id = 13\n",
    "biased_id = 11  # Using LSTAT as biased feature for demonstration purposes\n",
    "\n",
    "categorical_input_ids = [input_ids.index(cat_id) for cat_id in categorical_input_ids]\n",
    "biased_id = input_ids.index(biased_id)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8d5bce39-34e7-45de-82f8-4327d94a0de8",
   "metadata": {},
   "source": [
    "Loading the dataset and doing some preprocessing. There are two datasets. One Dataset is with the biased feature and the other dataset is unbiased because all rows from the biased feature are set to 0."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "e5dadf37-027e-4086-967e-d7bc75e7cafc",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Load dataset\n",
    "house_df = pd.read_csv(os.path.join(\"datasets\", dataset_name + \".csv\"))\n",
    "house_df = house_df.fillna(0)\n",
    "\n",
    "# Split input and output\n",
    "y_df = house_df.iloc[:, output_id]\n",
    "x_df = house_df.iloc[:, input_ids]\n",
    "\n",
    "scaler = MinMaxScaler()\n",
    "scaler.fit(x_df.to_numpy())\n",
    "x = scaler.transform(x_df.to_numpy())\n",
    "\n",
    "# Create train and test data\n",
    "x_train, x_test, y_train, y_test = train_test_split(x, y_df.to_numpy(), test_size=TRAIN_TEST_SPLIT, random_state=SEED)\n",
    "\n",
    "# Create training data without biased column\n",
    "ux_train = x_train.copy()\n",
    "ux_train[:, biased_id] = 0\n",
    "ux_test = x_test.copy()\n",
    "ux_test[:, biased_id] = 0"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1d7dd816-7c40-415a-b9ad-b07ebbafd69e",
   "metadata": {},
   "source": [
    "Train the biased and unbiased ML model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "b41862d9-3705-4712-ad32-993960dd9afa",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy of biased model: 0.77\n"
     ]
    }
   ],
   "source": [
    "biased_ml = RandomForestRegressor(random_state=SEED)\n",
    "biased_ml.fit(x_train, y_train)\n",
    "print(\"Accuracy of biased model: {0:3.2}\".format(biased_ml.score(x_test, y_test)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "23270d54-e32e-45cf-9e68-14c6efbb67a3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy of unbiased model: 0.96\n"
     ]
    }
   ],
   "source": [
    "unbiased_ml = RandomForestRegressor(random_state=SEED)\n",
    "unbiased_ml.fit(ux_test, y_test)\n",
    "print(\"Accuracy of unbiased model: {0:3.2}\".format(unbiased_ml.score(ux_test, y_test)))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d2ce0bcc-2813-40d8-8973-a60157cae461",
   "metadata": {},
   "source": [
    "## Get Lime Explanations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "6c4846aa-5488-4857-87c3-1271a8182bb9",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Calculating Lime explanations\n",
      "Original Lime explanation:\n",
      "{'RM': 4.222782069571837, 'LSTAT': -3.4068982139961905, 'DIS': -1.9202133425684937, 'B': 0.3780400306591363}\n",
      "Adversarial Lime explanation:\n",
      "{'LSTAT': -5.017204349246036, 'RM': 2.2636038288798446, 'DIS': -1.3549235542874367, 'B': 0.010346604645831444}\n",
      "Prediction fidelity between original and adversarial model: 0.75\n"
     ]
    }
   ],
   "source": [
    "adv = AdversarialModelToolbox(biased_model=biased_ml, \n",
    "                              x_train=x_train, \n",
    "                              y_train=y_train, \n",
    "                              x_test=x_test, \n",
    "                              y_test=y_test,\n",
    "                              input_feature_names=house_df.columns[input_ids].tolist(),\n",
    "                              categorical_feature_indices=categorical_input_ids,\n",
    "                              unbiased_model=unbiased_ml,\n",
    "                              biased_id=biased_id, \n",
    "                              fool_explainer_type=ExplainerType.LIME,\n",
    "                              ml_type=MLType.REGRESSION, \n",
    "                              seed=SEED)\n",
    "adv.train()\n",
    "adv.get_explanations(explanation_sample_number=-1)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "209e9d0e-e231-4455-809e-49cc5e08d518",
   "metadata": {},
   "source": [
    "## Get Shap Explanations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "6fb0216d-ed23-4357-8e90-dfa84dba8a91",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Calculating Shap explanations\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a3e1aaf3cfc2482faae7a7cd39a0c79f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/101 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Original Shap explanation:\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfoAAADYCAYAAAADBiHAAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/YYfK9AAAACXBIWXMAAAsTAAALEwEAmpwYAAAbRElEQVR4nO3deZhlRX3w8e8PBlQYAWURFYYlIC6oLOUaNOOGRp2IGl5EERFBIPomr74YEwTljWxRjEvUKCggqBgwEQEFBRUXjJpSXFCRgMyAwIRthrAom+f9o6rhzKVvT09PT9/pmu/nefrpe/aqU3XO75w6dc+NruuQJEltWmvUCZAkSauOgV6SpIYZ6CVJapiBXpKkhhnoJUlq2JxRJ2BVOOecc7oFCxaMOhmSJM2UGDbBO3pJkhpmoJckqWEGekmSGmaglySpYQZ6SZIaZqCXJKlhBnpJkhpmoJckqWEGekmSGmaglySpYQZ6SZIaZqCXJKlhBnpJkhpmoJckqWEGekmSGmaglySpYQZ6SZIaZqCXJKlhBnpJkhpmoJckqWEGekmSGhZd1406DdMujr+3vUxJkprQHTpnVaw2hk3wjl6SpIYZ6CVJapiBXpKkhhnoJUlqmIFekqSGGeglSWqYgV6SpIYZ6CVJatgq+db+MCmli4BnAfcA9wFXAUfnnM/sTf8zYK+c8xm95Z4B/ABYlHPeeibTLEnSbDaKO/r35pznAhsDpwCfTylt15v+a+DAgWUOrOMlSdIKGFnTfc75XuBESqvCTr1J/w7snFLaFiCl9HDg1cDJM51GSZJmu5EF+pTSusAhdfDy3qQ/AJ8D3lSH9wa+DVw/c6mTJKkNowj070opLQV+DxwFHJBz/vnAPCcCb0wpzQHeXIclSdIKGkWgPzrnvBGwCfBV4PmDM+ScLwUWAUcAjwLOn8kESpLUilE+o18CHAC8NKX0inFmOYES6D+dc75vRhMnSVIjRvo9+pzzLcA/AceklAbTcjqwO/DhGU+YJEmNmNHv0Q/xYeBtwL79kTnnPwAXjiRFkiQ1IrquG3Uapl0cf297mZIkNaE7dJXcY8ewCb4CV5KkhhnoJUlqmIFekqSGGeglSWqYgV6SpIatDl+vm3Zn73AeCxYsGHUyJEkaOe/oJUlqmIFekqSGGeglSWqYgV6SpIYZ6CVJapiBXpKkhhnoJUlqmIFekqSGGeglSWqYv0cvSauRVfRb5Wqfv0cvSdKayEAvSVLDDPSSJDXMQC9JUsMM9JIkNcxAL0lSwwz0kiQ1zEAvSVLDlvtmhpTSRcCFOeejBsZvChwHvATYELgN+CmwP/BG4LA6awDrAXcCYy+yOSbnfExKaT3gOuBmYLucc5dS+gSwT51vLeBhwB29TR+Uc/7cimVTkqQ108q8gumzwK3AzjnnG1JKm1GCfpdzPgY4BiCltAVwDfCknPPCgXW8pv7fCnghcEHO+WDg4LrsbsB3c85zVyKdkiStsVYm0D8b2CvnfANA/X/qCq7jIMoFw1b18wUrkR5JkjRgZQL9d4D31zv2/wR+nnO+b7ILp5SeCjwdOIQS6P81pbR5znnxSqRJkiT1rExnvL0od+NvBL4P3JxS+lBK6aGTXP4g4Gc5558A5wJL6rokSdI0mXKgzznfnnM+Nuf8LEpnvH1ZthPeUCml9YHXASfVdd1DafY/MKU09Bd4JEnSipmW30PMOd8NnJ1SuhDYaRKL7A1sALwnpTR2YfAQYCPgRcDXpyNdkiSt6SYb6OeM0yR/DHA68AvgbuC5wPOAYyexvjcDnwMOHRh/GqVJ30AvSdI0mGygf0/96/skcDIwj/L9+GuB44EPTLSilNJOwNOAAwY73qWU3g98JaX06Jzz9ZNMmyRJGiK6rlv+XLNMHH9ve5mStEboDp2WJ6pa8wzt3+YrcCVJapiBXpKkhhnoJUlqmIFekqSGNdnr4+wdzmPBggWjToYkSSPnHb0kSQ0z0EuS1DADvSRJDTPQS5LUMAO9JEkNM9BLktQwA70kSQ0z0EuS1DADvSRJDTPQS5LUMH+PXqslf5NbklaIv0cvSdKayEAvSVLDDPSSJDXMQC9JUsMM9JIkNcxAL0lSwwz0kiQ1bEa+rJxSugh4FnA38EfgZuBi4EM55x/35rkw53xUHX4+cCTwZMoFyWLgiznnd81EmiVJasFM3tG/N+f88JzzhsDzgEXAD1JKrxycMaW0DXAucCKwGbAx8CrgshlMryRJs95IXj+Wc14EHJ5SejTwzymlswZm2QW4Led8Wm/cL+ufJEmapFG/Z/QLwP7ADgPjMzA3pXQa8G/Aj3PO18x04iRJmu1G3Rnvd/X/xv2R9Y7/GcBdwPHAopTSZSmlPWY2eZIkzW6jDvRb1P83D07IOV+acz4g57wdsDlwPnBmSulxM5lASZJms1EH+r2Aa4HfTDRTzvkG4AjKo4YdZyBdkiQ1YSTP6FNKWwIHAPsBe+Wcu5RSf/pzgJ2BsyjN++sD7wR+T3l+L0mSJmEmA/0RKaV3Ah2lqf77wLNzzj8aZ94lwHzgb4GNKAH+Z8BLc85Xz0hqJUlqQHRdN+o0TLs4/t72MrWG6Q4d9RdCJGlWiWETRv2MXpIkrUIGekmSGmaglySpYQZ6SZIaZqCXJKlhTXZtPnuH81iwYMGokyFJ0sh5Ry9JUsMM9JIkNcxAL0lSwwz0kiQ1zEAvSVLDDPSSJDXMQC9JUsMM9JIkNcxAL0lSw/w9+hHwt9YlSdPM36OXJGlNZKCXJKlhBnpJkhpmoJckqWEGekmSGmaglySpYQZ6SZIaZqCXJKlhq+2bW1JKFwHPAu6poxYDH805f2hUaZIkabZZ3e/o35tznptzngvsAxydUtp91ImSJGm2WN0D/f1yzj8AfgXsOOq0SJI0W6y2Tfd9KaUAng08HviPESdHkqRZY3W/o39XSmkpcAfwPeBzwI9GmiJJkmaR1T3QH51z3ijnvB6wJfBE4KQRp0mSpFljdQ/098s5/w44A3jVqNMiSdJsMWsCfUppc2BP4GejToskSbPF6t4Z74iU0t/Vz3cA3wYOHWF6JEmaVVbbQJ9znj/qNEiSNNvNmqZ7SZK04gz0kiQ1zEAvSVLDDPSSJDXMQC9JUsNW2173K+PsHc5jwYIFo06GJEkj5x29JEkNM9BLktQwA70kSQ0z0EuS1DADvSRJDTPQS5LUMAO9JEkNM9BLktQwA70kSQ0z0EuS1DADvSRJDTPQS5LUMAO9JEkNM9BLktQwA70kSQ0z0EuS1DADvSRJDTPQS5LUMAO9JEkNM9BLktQwA70kSQ2LrutGnYZp95CHPOTSu++++w+jTsdMmTNnzib33nvvTaNOx0wyz2uGNS3Pa1p+wTxPo5u6rnvJuFO6rmvub9ddd82jToP5Nc/m2TybX/O8OuTZpntJkhpmoJckqWGtBvoTRp2AGbam5RfM85piTcvzmpZfMM+rXJOd8SRJUtHqHb0kScJAL0lS0+aMOgFTlVJ6HPAZYGPgZmDfnPN/DcyzNvAR4CVABxyXc/7UTKd1Okwyv0cCfwVcV0ddnHN+y0ymczqllI4HXg1sDTw553zpOPO0VMaTye+RtFXGGwOnAX8C3AVcARyUc75xYL4mynkF8nskbZXzWcA2wB+B24H/nXP+6cA8TZTxmEnm+UhmoJxnbaAHPgF8LOf82ZTSPsAngecPzPM6YDtge0qAvCSldGHOeeGMpnR6TCa/AKfmnA+d2aStMmcBHwa+O8E8LZXxWSw/v9BWGXfA+3LOFwGklN4PHAe8aWC+Vsp5svmFtsr5DTnnWwFSSq8ATgJ2GZinlTIeM5k8wwyU86xsuk8pbUbZYafXUacDu6SUNh2YdS/gxJzzH+sV81nAnjOW0GmyAvltSs75eznna5YzWxNlDJPOb1NyzreMBb3qB8BW48zaRDmvQH6bMhbwqg0pd7mDmijjMZPM84yYrXf0WwLX5pzvA8g535dSuq6O7zeBzQMW9YavrvPMNpPNL8BrUkq7A4uB9+Sc/2NmkzrjWinjFdFkGaeU1gIOAc4eZ3Jz5byc/EJj5ZxS+hSwOxCU5vlBLZbx8vIMM1DOs/KOXkN9Atgm5/wU4P3Al+szQbWj5TL+Z8qzzI+OOiEzZKL8NlfOOecDcs7zgMMoeWreJPI8I+U8WwP9NcBja+eNsU4cj6nj+65m2WaxeePMMxtMKr8558U553vq5wvq9B1nOK0zrZUynpRWy7h2RNwe2CvnPF4TZ1PlvLz8tlrOADnn04DnjRPQmirjvmF5nqlynpWBPud8A/BTYO86am/gksGeq8CZwIEppbXq8+w9gH+bqXROl8nmN6X02N7nnSi9t38zI4kcnSbKeLJaLOOU0tHArsAeOee7hszWTDlPJr8tlXNKaW5Kacve8ALglvrX11IZTyrPM1XOs/UZPcDBwGdSSu8GlgD7AqSUvgq8O+ecKV9jeQYw9jW0f8g5/3YUiZ0Gk8nvMSmlXYH7gLuB1+ecF48qwSsrpfQR4FXA5sCFKaWbc85ParWMJ5nf1sr4SZRmzcuB76eUAK7KOb+yxXJegfy2VM7rA2emlNan5OcWYEHOuWuxjKvJ5nlGytlX4EqS1LBZ2XQvSZImx0AvSVLDDPSSJDXMQC9JUsMM9JIkNcxAvxqIiBdHxHd7w/MjYuEIkzRjIuKUiJi2X6iKiK0jousNbxoRiyJik0kse3BEnDZdaZkNIuI5EbF01OlYE0XEPitynE/3saKJrapjYwrl/o8R8d6V2aaBfsQiIoAPAu9ZznyHRMSlEfE/EbEkInJE7NWbvjAi9hlnuQeNj+Lyuq65A9PmR0QXEbfXv+si4uSIeOTK5XQ0uq67Efg8y9+/6wP/ABw5A8labXRd992u6zYadTqGiYgjI+LCUadjTbCq9nVEXBQRh0/3ele1wWNjhHXxOOAtEfHY5c45hIF+9HYH1gW+NWyGiNibEqjeRPkVpMcAb6O8OGcqngdsS/k1pb3HmX5f13Vzu66bC+wGPAv40BS3tTo4CXhjRGwwwTz7AL/ouu7KGUrTMiJi7YjweJS0jK7rlgDnAQdNdR1r1Iml3t0eHhHfqnerv4iIp0TE3hFxRUTcGhGfiog5vWXmRcQXI+L6+ndCRDy8N/2YiPhtXd+VEfF/etO2rnfHr4+IX0XEbRHx9Yh4dC9ZewAXdhO/uejZwHe6rvthV/y+Xm1+fYq74iDgfMqbqCasPF3X/RY4F9h5cFpEzKn75BUD4z8TESfVzy+IiB/WVogbI+ILEbHZsO3V/bVbb3h+RNw7sM3DaovE0oi4OCJ2XU4e/gu4CXjhBLPtAVwwkJa/iYjLarldHRHHRsTaddrxEfGlgfmfV+ddvw7vGBFfi4ibesuvU6eN1Y03RcSvgDuBzSLiNRHxs9racn1EfHJsfXW5zSPinFpXL6/LdxGxdW+eA2vrz60RcUlE7D4s0+Ps31Mi4rSIOKnu32vr8bFTRPxnzd+3IuIxvWUWRsS7I+J79TjIEfG03vQJ60BErFPL9Dd1/VdGxKujtFgdBsyPB1qYth2Sjz+r27i1ltlBvWnzI+LeiNirrvvWiDijfxyPs76pnCueEhHfrPn8bV1+7d70p9d9c3tEfI9ysd3f5nq1Xl0VEbdExPkRsd2wNI6T5o0j4tRabxZHOQ4f2Zu+TOterw5uMWxfR8R+Nb/vrOu9ISI+ME493qK33v0i4or6+aPAc4Aj6jrHfb1rlLvlb0Rppr4xIm6OiLdHxFZ1n94WET+OiCf0llmpYyUeqOsnxgN1/UH1pn6ecP8M5GWZRyzTVO4XUM5RU9N13RrzByykvF7xCcA6wGeBK4ETKK8snAfcALy2zv9Q4ApKk+7DgEcAXwVO6q1zH8oddgDPB34PvLhO2xroKIFyE2AD4GLgxN7yPwT+eiCd84GFveE9gT8ARwEvADYakrd9ljce2BS4i/Kq1Z1q+nYd2Pa9veHtKO9ePmnIPn0fcFZveC7lF7meU4d3A55Ged3y5sB3gNN7858CfKo33AG7TZCeY+o+2xZYm9LKcRPwiP4+Hyed5wBHTVA3/hv4i4Fxrwa2qWW7c53noDrtiZRXVm7am/8zwKfr582AmykXUusCjwUy8O6BuvGNul/Wrfn5c+BJlIvw7YBfAcf2tvENyvu/N6jbuKiuZ+s6/c2UOvvUuo6X1vLYbki+B/fvKZQ6/LK6/MF1+bOBLYD1gG8CJwzUseso729fF/g7ys8nbzDJOvCPNZ9Pqft6C+ApddqRlAvhiY7rbWqa31i38UzKK0f37OWxAz5NqZ+PopwH3jWN54oNa/04AnhIXe63wDt602+u+2bduj8Ws+xx/nnKueJRdZ7/B1wGrDPesTJOms+n1PNH1L+vAF+Z4Fywdd0vWwzb18B+wD3AxyjnwD+hvL7378dbR2+ZK3rDFwGHL6cMj6zbOYAHjoP7gAsHyuDrvWVW9lg5hVJv/qKu41U1DVsNOTaG7Z8rBsbdX07TUe51nl0pLbDrTrQfh+7fqSw0W/9qRX9Hb/ilteD7J+szgA/Wz38JXDmwjl0pgXLtIdv4IvC+gYPgab3pbwEu6Q1fDuw3sI75/YpQx70c+HfKyeQ+SlP/jgN5uwNYOvD3R5Y9uP+WcoIaO3n8BPjkwLa7uuwS4CrKTyluNCS/T6AEvM3q8P7A5ROUwcuBG8Y7KOrw0EBPCQK3Ac8dWOcvxvLI8ED/OeDjE6TrbmD+curP8cAZveEfAm+rnx9OCYh/WocPBb45sPyrqSeFXt147nK2+VbgR/XzFnWZbXvTX8CyJ69LgX0H1nEOQ060jB/o+8Fhvbr+PXvj/opl6/BC4L294aD8Etlrl1cH6ry3Ay8bMu+RLD/QHwZcPDDuWOBrA3W6f5y/H/jSBOtcyIqdK15L+eWx6E0/CPhN/fy6uk/604+mHueUG4EOmNebvhZwK/V4YIJAT7nZ6IDte+N2qOMe3cvTVAL9XcB6vXEHUI/xwXX0lplKoP/lwLgbximDJdN4rJxCr67XcTcCrxhybAzbPxMF+pUu9zpu+zrfZhPtx2F/s/lHbabq+t7nOynPo28cGDfWpLcNMC8e3POyo9yZXBsRfw0cSKlYQbnq/fwE27yjt34owXSiZ8dlg113LuWqj4h4PPBx4NyI2KarNYFyt/nZ/nLR690ZEVHT+tmu6+6poz8NHBcR/7frutvruPu6SXbQ6rru1xHxE0rLxj9R7qpO7m1zV8pd+FMpQSMod1VTsUld9pzo9aynXO1vMf4i99uActEyzIPKIUrfiLdTWg/mUK62f9Cb5WRK0Psg8L+Aa7uuu7hO2wb404G6E5S7lb6FA9t8EfBu4PGUO8O1KSc8KK0CUE4cYxYNrG8b4GMR8ZHeuDnA75i8++tr13V3lmrzoONmsNl7YW+ZLiKuppbJcurAppQ75MtXIH2DtqTcPfddCbyiNzx4nA8eh+NZkXPFlpSTd79eXlnHQ9kXiwam9+vjNvX/z+v+HrNObx0TGZunv84re9OuZ+pu6Lruzt7wQpZ/vE3FYBrvZIJ6Nw3HynjbnEy9WBHTVe4b8MAN2Apbo57RT8EiypXrRgN/D+267tqI+FNKs+NBwCY1OJ5DOZFN1iWUZuBJ67ruMkpw2YrSRDdZL6A0ce1fn+EtpjQTzaXckUzVycB+9bnSM4FTe9O+QGk1eFzXdRswfue/vjsoJ/4xj+l9vqlOf+FAeazfdd1xy1nvjpR9Pcwy5RARW1KaCo+i3BFtSGm+7JftF4DtI2IXypX9yb1piyhX//10btiVDo599/8WeUSsC5xV1zuv7q939rZ5bf0/r7d8//PYdvcf2O7crusOmSDv02HrsQ/1gnIeD1xcTFQHbqSU6fZD1jveb9MPuoYHTphjtmVmf8v8GmCrWPZs3U/DteNM76d5LAhtP1B263Vdd/oktw+9cuCBZ8Fj025n+LEFw/f1ZhGxXm94ax4o27Gbg6msd8qm6VhZUePlY3CfwrL5n65y35HS4nH3VBJuoJ/YucBYR6GHR/HYiHhlnb4BpRn9RqCLiJdRnhutiLMoAXioiNg/IvaM+l3w2vHlYOBXXdcN/qbzRN5MeT76eMrz+Z0oFehkVqJHJ+Vg2w74CHBB13XX9qZtQGmGui0i5lGeVU0kA2+IiHVrp5m3j02oV8UfBo6PiO0BImJulPcQDJ5c7lcvQDalPO8b5iyW7aw3l3J83AjcExHPBF7fX6DruqXAlygXA4MXOKcCqZbdQyNirdp55yUTpGFdSr+QJV3X/T4inkhpjhzb3u8ozaDH1fq4GTD4taUPAkdG6TwXEfGwiNittgKtSvtHxC5ROmm9g3Ln/pU6bWgdqGX6L8D7onReHDvGnlxnWUxpVVt3gm2fDuwaEftG6az5dEp9/vS05nBiX6GU3WG17u5ACTxjaTiXUqfeEaXz4S6Ux1wAdF13A6Ul8ONRv0YVERtFxCtj4Cuw4+m67jrg68AH6nKPAD4AnNd13dhdawb2rsfMppT+BH3D9vValDr3sCidIQ+l9Eeh67qbqBeXUb458mRKq+HgeifdqXCSpuNYWVHj7Z9LKBdCL6/H+CuB5/amT1e5v4hyjpoSA/0EanPVCyh3epdRTlbfoARIgK9Req7/iHK3+ZeUE/+K+Bpwb0TMn2CeJZQm4l9HxB2UZ8NLKc86J6VW9D2A47uuW9z/o7RK7BwRaQXTDkDXdbdS8v3nlK+y9b2Z8kzvNkofgzOXs7q3Uk4Kt1CegZ4yMP09wJeBL0fE/1A6TB3MxHV5f+CUms5hTgOeWk9kdF336962llKC03h3VidT8v21erKlLr+Y8jXGPShNnUso+2jcXuN1mduBQyhB73ZKC8LgY6DXUoLo74Dv8cD+vKuu40RKB8mT6zavppzQ15kg79PhBMqF3hJgL8oz97H9vbw68C5KWZ9V5/k2D9zhn0m5I10cpWf04J07XdddRXl++1ZKx6fTKJ0ez5iuzC1PzevulIvF/6Yc16dSHmeNXRS+jLJvllD21b8MrOZASsfXiyLiNkrfkz0pTbaTsQ9l/11W/5YC+/amH065MbmeEgS/MLD8sH29iHJnehXl3HM+pY6NeQPlXHRrze/gBdYHKRe9SyPil5PMy4Sm41iZggftn658HfdvKPX/FuAllA6AY+lcykqWe0RsRKnfn5hiuv09+tVBvcs7rOu659bh+ZTAtPUIkzUr1VaAq7quizq8CfBjIA08Xx1v2YMpneleP9F8q5OIeDHlYuRh3YgO5ij9QA4f7B+i2S8i9qOU7XTfkc+41eFYmYqIOJbSP2TKLRJrYme81U7XdedTrpI1zWrT4laTnPcTrMRV80yIiKdSrvR/QXnWdxTwr7PpxCXNhFaOla7r/n5l12HT/eppIbP7TXSjtJTSwbBVj6Q0f99OaY78OaXpUNKyPFYqm+4lSWqYd/SSJDXMQC9JUsMM9JIkNcxAL0lSwwz0kiQ17P8DEaHxBun1zx4AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 576x223.2 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0866677389db445296f53aa203de08ab",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/101 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Adversarial Shap explanation:\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfoAAADYCAYAAAADBiHAAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/YYfK9AAAACXBIWXMAAAsTAAALEwEAmpwYAAAbUklEQVR4nO3deZglVX3w8e8PBlQYAWURFYaBgLigshzXoBk3NOpE1PAiiogIAtE3efXFmCDovJEtinGJGgUFBBUDJiKgoKDiglFzFBdUJCAzIDBhmyEsCgzW+8c5zdRc+vbc7unp27fm+3mefvrWfk6dU/WrOnVu3WiaBkmS1E3rDTsBkiRp7THQS5LUYQZ6SZI6zEAvSVKHGeglSeqwOcNOwNpw3nnnNQsXLhx2MiRJminRb4J39JIkdZiBXpKkDjPQS5LUYQZ6SZI6zEAvSVKHGeglSeowA70kSR1moJckqcMM9JIkdZiBXpKkDjPQS5LUYQZ6SZI6zEAvSVKHGeglSeowA70kSR1moJckqcMM9JIkdZiBXpKkDjPQS5LUYQZ6SZI6zEAvSVKHRdM0w07DtIsTV3QvU5KkTmiOmLM2Vhv9JnhHL0lShxnoJUnqMAO9JEkdZqCXJKnDDPSSJHWYgV6SpA4z0EuS1GEGekmSOmy139pPKV0CXJxzPqZn/JbACcBLgE2BO4CfAgcBbwSOrLMGsBFwNzD2Ipvjcs7HpZQ2Am4AbgV2zDk3KaVPAPvX+dYDHgbc1dr0oTnnz00um5IkrZvW5PU8nwVuB3bLOd+UUtqKEvSbnPNxwHEAKaVtgOuAJ+WcF/es4zX1/3bAC4GLcs6HAYfVZfcEvptznrsG6ZQkaZ21JoH+2cC+OeebAOr/0ye5jkMpFwzb1c8XrUF6JElSjzUJ9N8B3l/v2P8T+HnO+f5BF04pPRV4OnA4JdD/a0pp65zz0jVIkyRJalmTznj7Uu7G3wh8H7g1pfShlNJDB1z+UOBnOeefAOcDy+q6JEnSNJlyoM8535lzPj7n/CxKZ7wDWLUTXl8ppY2B1wGn1HXdR2n2PySl1PcXeCRJ0uRMy2/l5ZzvBc5NKV0M7DrAIvsBmwDvSSmNXRg8BNgMeBHw9elIlyRJ67pBA/2ccZrkjwPOBH4B3As8F3gecPwA63sz8DngiJ7xZ1Ca9A30kiRNg0ED/XvqX9sngVOBeZTvx18PnAh8YKIVpZR2BZ4GHNzb8S6l9H7gKymlR+ecbxwwbZIkqY9ommb1c42YOHFF9zIlSeqE5ohpeWreq2//Nl+BK0lShxnoJUnqMAO9JEkdZqCXJKnDDPSSJHXYWun6N2zn7nwBCxcuHHYyJEkaOu/oJUnqMAO9JEkdZqCXJKnDDPSSJHWYgV6SpA4z0EuS1GEGekmSOsxAL0lShxnoJUnqMH+PXpKmaC39rrg0Ff4evSRJ6yIDvSRJHWaglySpwwz0kiR1mIFekqQOM9BLktRhBnpJkjrMQC9JUofN6NseUkqXAM8C7gPuB64Bjs05n92a/mfAvjnns1rLPQP4AbAk5zx/JtMsSdIoG8Yd/XtzznOBzYHTgM+nlHZsTf81cEjPMofU8ZIkaRKG1nSfc14BnExpVdi1Nenfgd1SSjsApJQeDrwaOHWm0yhJ0qgbWqBPKW0IHF4Hr2xN+gPwOeBNdXg/4NvAjTOXOkmSumEYgf5dKaXlwO+BY4CDc84/75nnZOCNKaU5wJvrsCRJmqRhBPpjc86bAVsAXwWe3ztDzvlyYAlwNPAo4MKZTKAkSV0xzGf0y4CDgZemlF4xziwnUQL9p3PO989o4iRJ6oihfo8+53wb8E/AcSml3rScCewFfHjGEyZJUkfM6Pfo+/gw8DbggPbInPMfgIuHkiJJkjoimqYZdhqmXZy4onuZkjTrNEfMhnslCYDoN8FX4EqS1GEGekmSOsxAL0lShxnoJUnqsE72JDl35wtYuHDhsJMhSdLQeUcvSVKHGeglSeowA70kSR1moJckqcMM9JIkdZiBXpKkDjPQS5LUYQZ6SZI6zEAvSVKHGeglSeowf49eD/C3tSVpZPl79JIkrYsM9JIkdZiBXpKkDjPQS5LUYQZ6SZI6zEAvSVKHGeglSeqwGfnidErpEuBZwL3AH4FbgUuBD+Wcf9ya5+Kc8zF1+PnAIuDJlAuSpcAXc87vmok0S5LUBTN5R//enPPDc86bAs8DlgA/SCm9snfGlNL2wPnAycBWwObAq4ArZjC9kiSNvKG8Ci3nvAQ4KqX0aOCfU0rn9MyyO3BHzvmM1rhf1j9JkjSgYb/z9AvAQcDOPeMzMDeldAbwb8CPc87XzXTiJEkadcPujPe7+n/z9sh6x/8M4B7gRGBJSumKlNLeM5s8SZJG27AD/Tb1/629E3LOl+ecD8457whsDVwInJ1SetxMJlCSpFE27EC/L3A98JuJZso53wQcTXnUsMsMpEuSpE4YyjP6lNK2wMHAgcC+OecmpdSe/hxgN+AcSvP+xsA7gd9Tnt9LkqQBzGSgPzql9E6goTTVfx94ds75R+PMuwxYAPwtsBklwP8MeGnO+doZSa0kSR0QTdMMOw3TLk5c0b1MzYDmiGF/CUOSNEXRb8Kwn9FLkqS1yEAvSVKHGeglSeowA70kSR1moJckqcM62c363J0vYOHChcNOhiRJQ+cdvSRJHWaglySpwwz0kiR1mIFekqQOM9BLktRhBnpJkjrMQC9JUocZ6CVJ6jADvSRJHebv0c8i/h68JGmK/D16SZLWRQZ6SZI6zEAvSVKHGeglSeowA70kSR1moJckqcMM9JIkdZiBXpKkDpu1b2hJKV0CPAu4r45aCnw05/yhYaVJkqRRM9vv6N+bc56bc54L7A8cm1Laa9iJkiRpVMz2QP+AnPMPgF8Buww7LZIkjYpZ23TfllIK4NnA44H/GHJyJEkaGbP9jv5dKaXlwF3A94DPAT8aaookSRohsz3QH5tz3iznvBGwLfBE4JQhp0mSpJEx2wP9A3LOvwPOAl417LRIkjQqRibQp5S2BvYBfjbstEiSNCpme2e8o1NKf1c/3wV8GzhiiOmRJGmkzNpAn3NeMOw0SJI06kam6V6SJE2egV6SpA4z0EuS1GEGekmSOsxAL0lSh83aXvdr4tydL2DhwoXDToYkSUPnHb0kSR1moJckqcMM9JIkdZiBXpKkDjPQS5LUYQZ6SZI6zEAvSVKHGeglSeowA70kSR1moJckqcMM9JIkdZiBXpKkDjPQS5LUYQZ6SZI6zEAvSVKHGeglSeowA70kSR1moJckqcMM9JIkdZiBXpKkDjPQS5LUYdE0zbDTMO0e8pCHXH7vvff+YdjpWJvmzJmzxYoVK24ZdjrWpnUhj7Bu5NM8dse6kM8RzeMtTdO8ZNwpTdN07m+PPfbIw06DeTSP5tM8dvFvXchn1/Jo070kSR1moJckqcO6GuhPGnYCZoB57I51IZ/msTvWhXx2Ko+d7IwnSZKKrt7RS5IkDPSSJHXanGEnYKpSSo8DPgNsDtwKHJBz/q+eedYHPgK8BGiAE3LOn5rptE7VgHlcBPwVcEMddWnO+S0zmc41kVI6EXg1MB94cs758nHmGfVyHCSPixjtctwcOAP4E+Ae4Crg0JzzzT3zjWxZTiKPixjhsgRIKZ0DbA/8EbgT+N8555/2zDOyZQkD53ERI16WMMKBHvgE8LGc82dTSvsDnwSe3zPP64AdgZ0owfKylNLFOefFM5rSqRskjwCn55yPmNmkTZtzgA8D351gnlEvx3NYfR5htMuxAd6Xc74EIKX0fuAE4E09841yWQ6aRxjtsgR4Q875doCU0iuAU4Dde+YZ5bKEwfIIo1+Wo9l0n1LailIgZ9ZRZwK7p5S27Jl1X+DknPMf61X3OcA+M5bQNTCJPI60nPP3cs7XrWa2kS1HGDiPIy3nfNtYAKx+AGw3zqwjW5aTyOPIGwuA1aaUu95eI1uWMHAeO2FU7+i3Ba7POd8PkHO+P6V0Qx3fbkabByxpDV9b5xkFg+YR4DUppb2ApcB7cs7/MbNJXetGuRwnoxPlmFJaDzgcOHecyZ0oy9XkETpQlimlTwF7AUFpnu818mU5QB6hA2U5knf0WsUngO1zzk8B3g98uT5L1GjpUjn+M+WZ50eHnZC1aKI8dqIsc84H55znAUdS8tE5A+SxE2U5qoH+OuCxtTPIWKeQx9TxbdeyatPavHHmma0GymPOeWnO+b76+aI6fZcZTuvaNsrlOJCulGPteLgTsG/Oebym0JEvy9XlsStlOSbnfAbwvHEC3MiX5Zh+eexKWY5koM853wT8FNivjtoPuKy39ytwNnBISmm9+mx7b+DfZiqda2LQPKaUHtv6vCulZ/dvZiSRM2dky3FQXSjHlNKxwB7A3jnne/rMNtJlOUgeR70sU0pzU0rbtoYXArfVv7aRLctB8zjqZTlmVJ/RAxwGfCal9G5gGXAAQErpq8C7c86Z8lWYZwBjX0n7h5zzb4eR2CkaJI/HpZT2AO4H7gVen3NeOqwET1ZK6SPAq4CtgYtTSrfmnJ/UpXIcMI+jXo5PojR/Xgl8P6UEcE3O+ZVdKctJ5HGkyxLYGDg7pbQxJQ+3AQtzzk1XypLB8zjqZQn4ClxJkjptJJvuJUnSYAz0kiR1mIFekqQOM9BLktRhBnpJkjrMQD8LRMSLI+K7reEFEbF4iEmaMRFxWkRM2y9eRcT8iGhaw1tGxJKI2GKAZQ+LiDOmKy2jICKeExHLh52OdVFE7D+Z43y6jxVNbG0dG1Mo93+MiPeuyTYN9EMWEQF8EHjPauY7PCIuj4j/iYhlEZEjYt/W9MURsf84yz1ofBRX1nXN7Zm2ICKaiLiz/t0QEadGxCPXLKfD0TTNzcDnWf3+3Rj4B2DRDCRr1mia5rtN02w27HT0ExGLIuLiYadjXbC29nVEXBIRR033ete23mNjiHXxBOAtEfHY1c7Zh4F++PYCNgS+1W+GiNiPEqjeRPmVpccAb6O8RGcqngfsQPm1pv3GmX5/0zRzm6aZC+wJPAv40BS3NRucArwxIjaZYJ79gV80TXP1DKVpFRGxfkR4PEpaRdM0y4ALgEOnuo516sRS726Piohv1bvVX0TEUyJiv4i4KiJuj4hPRcSc1jLzIuKLEXFj/TspIh7emn5cRPy2ru/qiPg/rWnz693x6yPiVxFxR0R8PSIe3UrW3sDFzcRvLno28J2maX7YFL+vV5tfn+KuOBS4kPJmqwkrT9M0vwXOB3brnRYRc+o+eUXP+M9ExCn18wsi4oe1FeLmiPhCRGzVb3t1f+3ZGl4QESt6tnlkbZFYHhGXRsQeq8nDfwG3AC+cYLa9gYt60vI3EXFFLbdrI+L4iFi/TjsxIr7UM//z6rwb1+FdIuJrEXFLa/kN6rSxuvGmiPgVcDewVUS8JiJ+VltbboyIT46try63dUScV+vqlXX5JiLmt+Y5pLb+3B4Rl0XEXv0yPc7+PS0izoiIU+r+vb4eH7tGxH/W/H0rIh7TWmZxRLw7Ir5Xj4McEU9rTZ+wDkTEBrVMf1PXf3VEvDpKi9WRwIJY2cK0Q598/Fndxu21zA5tTVsQESsiYt+67tsj4qz2cTzO+qZyrnhKRHyz5vO3dfn1W9OfXvfNnRHxPcrFdnubG9V6dU1E3BYRF0bEjv3SOE6aN4+I02u9WRrlOHxka/oqrXutOrhNv30dEQfW/L6zrvemiPjAOPV4m9Z6D4yIq+rnjwLPAY6u6xz39bFR7pa/EaWZ+uaIuDUi3h4R29V9ekdE/DgintBaZo2OlVhZ10+OlXX9QfWmfp5w//TkZZVHLNNU7hdRzlFT0zTNOvMHLKa8rvEJwAbAZ4GrgZMor0ScB9wEvLbO/1DgKkqT7sOARwBfBU5prXN/yh12AM8Hfg+8uE6bDzSUQLkFsAlwKXBya/kfAn/dk84FwOLW8D7AH4BjgBcAm/XJ2/6rGw9sCdxDeSXrrjV9e/Rse0VreEfKu51P6bNP3wec0xqeS/lVr+fU4T2Bp1Fet7w18B3gzNb8pwGfag03wJ4TpOe4us92ANantHLcAjyivc/HSed5wDET1I3/Bv6iZ9yrge1r2e5W5zm0Tnsi5ZWYW7bm/wzw6fp5K+BWyoXUhsBjgQy8u6dufKPulw1rfv4ceBLlInxH4FfA8a1tfIPyPvFN6jYuqeuZX6e/mVJnn1rX8dJaHjv2yXfv/j2NUodfVpc/rC5/LrANsBHwTeCknjp2A+Ud8BsCf0f5KeVNBqwD/1jz+ZS6r7cBnlKnLaJcCE90XG9f0/zGuo1nUl5puk8rjw3waUr9fBTlPPCuaTxXbFrrx9HAQ+pyvwXe0Zp+a903G9b9sZRVj/PPU84Vj6rz/D/gCmCD8Y6VcdJ8IaWeP6L+fQX4ygTngvl1v2zTb18DBwL3AR+jnAP/hPIK4L8fbx2tZa5qDV8CHLWaMlxUt3MwK4+D+4GLe8rg661l1vRYOY1Sb/6iruNVNQ3b9Tk2+u2fq3rGPVBO01HudZ49KC2wG060H/vu36ksNKp/taK/ozX80lrw7ZP1WcAH6+e/BK7uWccelEC5fp9tfBF4X89B8LTW9LcAl7WGrwQO7FnHgnZFqONeDvw75WRyP6Wpf5eevN0FLO/5+yOrHtx/SzlBjZ08fgJ8smfbTV12GXAN5acaN+uT3ydQAt5Wdfgg4MoJyuDlwE3jHRR1uG+gpwSBO4Dn9qzzF2N5pH+g/xzw8QnSdS+wYDX150TgrNbwD4G31c8PpwTEP63DRwDf7Fn+1dSTQqtuPHc123wr8KP6eZu6zA6t6S9g1ZPX5cABPes4jz4nWsYP9O3gsFFd/z6tcX/FqnV4MfDe1nBQftnstaurA3XeO4GX9Zl3EasP9EcCl/aMOx74Wk+dbh/n7we+NME6FzO5c8VrKb9sFq3phwK/qZ9fV/dJe/qx1OOcciPQAPNa09cDbqceD0wQ6Ck3Gw2wU2vcznXco1t5mkqgvwfYqDXuYOox3ruO1jJTCfS/7Bl30zhlsGwaj5XTaNX1Ou5m4BV9jo1++2eiQL/G5V7H7VTn22qi/djvb5R/1Gaqbmx9vpvyPPrmnnFjTXrbA/PiwT0vG8qdyfUR8dfAIZSKFZSr3s9PsM27WuuHEkwnenZcNtg051Ou+oiIxwMfB86PiO2bWhMod5ufbS8Xrd6dERE1rZ9tmua+OvrTwAkR8X+bprmzjru/GbCDVtM0v46In1BaNv6Jcld1amube1Duwp9KCRpBuauaii3qsudFq2c95Wp/m/EXecAmlIuWfh5UDlH6Rryd0nowh3K1/YPWLKdSgt4Hgf8FXN80zaV12vbAn/bUnaDcrbQt7tnmi4B3A4+n3BmuTznhQWkVgHLiGLOkZ33bAx+LiI+0xs0BfsfgHqivTdPcXarNg46b3mbvxa1lmoi4llomq6kDW1LukK+cRPp6bUu5e267GnhFa7j3OO89DsczmXPFtpSTd7teXl3HQ9kXS3qmt+vj9vX/z+v+HrNBax0TGZunvc6rW9NuZOpuaprm7tbwYlZ/vE1FbxrvZoJ6Nw3HynjbHKReTMZ0lfsmrLwBm7R16hn9FCyhXLlu1vP30KZpro+IP6U0Ox4KbFGD43mUE9mgLqM0Aw+saZorKMFlO0oT3aBeQGniOqg+w1tKaSaaS7kjmapTgQPrc6VnAqe3pn2B0mrwuKZpNmH8zn9td1FO/GMe0/p8S53+wp7y2LhpmhNWs95dKPu6n1XKISK2pTQVHkO5I9qU0nzZLtsvADtFxO6UK/tTW9OWUK7+2+nctCkdHNse+D3ziNgQOKeud17dX+9sbfP6+n9ea/n257HtHtSz3blN0xw+Qd6nw/yxD/WCch4rLy4mqgM3U8p0pz7rHe837Xtdx8oT5pgdmNnfRr8O2C5WPVu303D9ONPbaR4LQjv1lN1GTdOcOeD2oVUOrHwWPDbtTvofW9B/X28VERu1huezsmzHbg6mst4pm6ZjZbLGy0fvPoVV8z9d5b4LpcXj3qkk3EA/sfOBsY5CD4/isRHxyjp9E0oz+s1AExEvozw3moxzKAG4r4g4KCL2ifpd8Nrx5TDgV03T9P5G9ETeTHk++njK8/ldKRXoVNagRyflYNsR+AhwUdM017embUJphrojIuZRnlVNJANviIgNa6eZt49NqFfFHwZOjIidACJibpT3EPSeXB5QL0C2pDzv6+ccVu2sN5dyfNwM3BcRzwRe316gaZrlwJcoFwO9FzinA6mW3UMjYr3aeeclE6RhQ0q/kGVN0/w+Ip5IaY4c297vKM2gJ9T6uBXQ+7WlDwKLonSei4h4WETsWVuB1qaDImL3KJ203kG5c/9Knda3DtQy/RfgfVE6L44dY0+usyyltKptOMG2zwT2iIgDonTWfDqlPn96WnM4sa9Qyu7IWnd3pgSesTScT6lT74jS+XB3ymMuAJqmuYnSEvjxqF+jiojNIuKV0fMV2PE0TXMD8HXgA3W5RwAfAC5ommbsrjUD+9VjZktKf4K2fvt6PUqde1iUzpBHUPqj0DTNLdSLyyjfHHkypdWwd70Ddyoc0HQcK5M13v65jHIh9PJ6jL8SeG5r+nSV+4so56gpMdBPoDZXvYByp3cF5WT1DUqABPgapef6jyh3m39JOfFPxteAFRGxYIJ5llGaiH8dEXdRng0vpzzrHEit6HsDJzZNs7T9R2mV2C0i0iTTDkDTNLdT8v3nlK+ytb2Z8kzvDkofg7NXs7q3Uk4Kt1GegZ7WM/09wJeBL0fE/1A6TB3GxHX5IOC0ms5+zgCeWk9kNE3z69a2llOC03h3VqdS8v21erKlLr+U8jXGvSlNncso+2jcXuN1mTuBwylB705KC0LvY6DXUoLo74DvsXJ/3lPXcTKlg+SpdZvXUk7oG0yQ9+lwEuVCbxmwL+WZ+9j+Xl0deBelrM+p83yblXf4Z1PuSJdG6Rnde+dO0zTXUJ7fvpXS8ekMSqfHs6Yrc6tT87oX5WLxvynH9emUx1ljF4Uvo+ybZZR99S89qzmE0vH1koi4g9L3ZB9Kk+0g9qfsvyvq33LggNb0oyg3JjdSguAXepbvt6+XUO5Mr6Gcey6k1LExb6Cci26v+e29wPog5aJ3eUT8csC8TGg6jpUpeND+acrXcf+GUv9vA15C6QA4ls7lrGG5R8RmlPr9iSmm29+jnw3qXd6RTdM8tw4voASm+UNM1kiqrQDXNE0TdXgL4MdA6nm+Ot6yh1E6071+ovlmk4h4MeVi5GHNkA7mKP1AjurtH6LRFxEHUsp2uu/IZ9xsOFamIiKOp/QPmXKLxLrYGW/WaZrmQspVsqZZbVrcbsB5P8EaXDXPhIh4KuVK/xeUZ33HAP86SicuaSZ05Vhpmubv13QdNt3PTosZ7TfRDdNySgfDrnokpfn7Tkpz5M8pTYeSVuWxUtl0L0lSh3lHL0lShxnoJUnqMAO9JEkdZqCXJKnDDPSSJHXY/wdRqPEGDvuokAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 576x223.2 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Prediction fidelity between original and adversarial model: 0.68\n"
     ]
    }
   ],
   "source": [
    "adv = AdversarialModelToolbox(biased_model=biased_ml,\n",
    "                              x_train=x_train, \n",
    "                              y_train=y_train,\n",
    "                              x_test=x_test, \n",
    "                              y_test=y_test,\n",
    "                              input_feature_names=house_df.columns[input_ids].tolist(),\n",
    "                              categorical_feature_indices=categorical_input_ids,\n",
    "                              unbiased_model=unbiased_ml,\n",
    "                              biased_id=biased_id, \n",
    "                              fool_explainer_type=ExplainerType.SHAP,\n",
    "                              ml_type=MLType.REGRESSION, \n",
    "                              seed=SEED)\n",
    "adv.train()\n",
    "adv.get_explanations(explanation_sample_number=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "40185c0e-9ee6-46be-b58f-64bf7377ae9f",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}