notebooks/adversarial-ml-examples-exercises.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "416ecc26-322b-4f17-8577-70cc04f0367a",
   "metadata": {},
   "source": [
    "# Examples and exercises for Lecture Adversarial Regularization Regimes for Classification Tasks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "f309f158-f891-423a-9fee-59acf6e607bd",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "from pathlib import Path\n",
    "\n",
    "import pandas as pd\n",
    "from sklearn.linear_model import LogisticRegression, LogisticRegressionCV\n",
    "import numpy as np\n",
    "\n",
    "from risk_learning.arr import (\n",
    "    convert_to_categorical,\n",
    "    make_feature_combination_array,\n",
    "    make_feature_combination_score_array,\n",
    "    make_trend_reports, \n",
    "    make_data_trend_reports\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bdd11940-2d83-4e03-af51-bfd87ef139cf",
   "metadata": {},
   "source": [
    "## Example Simpson's Paradox Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "142da3ce-17d2-4e8a-9bd5-5b25684421db",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>default</th>\n",
       "      <th>gender</th>\n",
       "      <th>occupation</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>595</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>596</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>597</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>598</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>599</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>600 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     default  gender  occupation\n",
       "0          0       0           1\n",
       "1          1       0           0\n",
       "2          1       1           1\n",
       "3          0       0           0\n",
       "4          0       1           1\n",
       "..       ...     ...         ...\n",
       "595        0       0           0\n",
       "596        0       0           1\n",
       "597        1       0           0\n",
       "598        1       0           0\n",
       "599        0       0           1\n",
       "\n",
       "[600 rows x 3 columns]"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "datadir = Path(os.getcwd()) / 'data'\n",
    "data_path = datadir / 'adversarial-default-for-x-validation.csv'\n",
    "\n",
    "df = pd.read_csv(data_path)\n",
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "1a34a6fd-c73b-4e9e-b1cc-f6cdefd7e935",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>default</th>\n",
       "      <th>gender</th>\n",
       "      <th>occupation</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  default gender occupation\n",
       "0       0      0          1\n",
       "1       1      0          0\n",
       "2       1      1          1\n",
       "3       0      0          0\n",
       "4       0      1          1\n",
       "5       1      0          0\n",
       "6       0      0          0\n",
       "7       1      0          0\n",
       "8       0      0          0\n",
       "9       1      0          0"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "label_mapping_values = dict(gender=[0, 1], occupation=[0, 1])\n",
    "data_categories = label_mapping_values.copy()\n",
    "data_categories['default'] = [0, 1]\n",
    "df = convert_to_categorical(df, data_categories)\n",
    "df.head(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "527e5b7b-66e6-4218-a0c1-df4e709df14b",
   "metadata": {},
   "source": [
    "## Exercise: Simpson or not?\n",
    "\n",
    "Difficulty: (*)\n",
    "\n",
    "Prove that this dataset exhibites Simpson's paradox."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fde5b290-34d0-431c-86d4-fc1259c11c7b",
   "metadata": {},
   "source": [
    "## Exercises: non-trivial regularization regime\n",
    "\n",
    "* Which optimizer (\"solver\") for logistic regression seems best suited for the above dataset? https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html. Difficulty (*)\n",
    "* Calculate the \"true\" trends for female default for each occupation subgroup. Note that in sklearn, the inverse regularization parameter is  used, $C$, so to approximate the usual $c=0$, set $C$ to a large value. Difficulty (**)\n",
    "* Show that this dataset is adversarial for logistic regression for inverse regularization parameter $C=0.05$. Difficulty: (**)\n",
    "* Show that this dataset is still adversarial for k-fold cross-validated logistic regression if $k=5$, the default setting."
   ]
  }
 ],
 "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": "3.10.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}