notebooks/debiased_representation.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from sklearn.svm import LinearSVC\n",
    "from sklearn.linear_model import SGDClassifier, LogisticRegression\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "from sklearn.utils import shuffle\n",
    "\n",
    "from collections import Counter, defaultdict\n",
    "\n",
    "import json\n",
    "from collections import defaultdict\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import seaborn as sns\n",
    "# sns.set()\n",
    "sns.set_style(\"whitegrid\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "sys.path.append('..')\n",
    "from src import debias"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_data(path, size, ratio=0.5):\n",
    "    fnames = [\"neg_neg.npy\", \"neg_pos.npy\", \"pos_neg.npy\", \"pos_pos.npy\"]\n",
    "    protected_labels = [0, 1, 0, 1]\n",
    "    main_labels = [0, 0, 1, 1]\n",
    "    X, Y_p, Y_m = [], [], []\n",
    "    n1 = int(size * ratio / 2)\n",
    "    n2 = int(size * (1 - ratio) / 2)\n",
    "#     print(n1, n2)\n",
    "\n",
    "    for fname, p_label, m_label, n in zip(fnames, protected_labels, main_labels, [n1, n2, n2, n1]):\n",
    "#         print(path + '/' + fname)\n",
    "#         print(np.load(path + '/' + fname).shape)\n",
    "        data = np.load(path + '/' + fname)[:n]\n",
    "        for x in data:\n",
    "            X.append(x)\n",
    "        for _ in data:\n",
    "            Y_p.append(p_label)\n",
    "        for _ in data:\n",
    "            Y_m.append(m_label)\n",
    "\n",
    "    Y_p = np.array(Y_p)\n",
    "    Y_m = np.array(Y_m)\n",
    "    X = np.array(X)\n",
    "    X, Y_p, Y_m = shuffle(X, Y_p, Y_m, random_state=0)\n",
    "    return X, Y_p, Y_m"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_TPR(y_main, y_hat_main, y_protected):\n",
    "    \n",
    "    all_y = list(Counter(y_main).keys())\n",
    "    \n",
    "    protected_vals = defaultdict(dict)\n",
    "    for label in all_y:\n",
    "        for i in range(2):\n",
    "            used_vals = (y_main == label) & (y_protected == i)\n",
    "            y_label = y_main[used_vals]\n",
    "            y_hat_label = y_hat_main[used_vals]\n",
    "            protected_vals['y:{}'.format(label)]['p:{}'.format(i)] = (y_label == y_hat_label).mean()\n",
    "            \n",
    "    diffs = {}\n",
    "    for k, v in protected_vals.items():\n",
    "        vals = list(v.values())\n",
    "        diffs[k] = vals[0] - vals[1]\n",
    "    return protected_vals, diffs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dict_values([-0.30746933667083864, 0.49425093867334163])"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "get_TPR(y_m_dev, biased_classifier.predict(x_dev), y_p_dev)[1].values()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def rms(arr):\n",
    "    return np.sqrt(np.mean(np.square(arr)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "results = defaultdict(dict)\n",
    "\n",
    "for ratio in [0.5, 0.6, 0.7, 0.8]:\n",
    "\n",
    "    x_train, y_p_train, y_m_train = load_data(\n",
    "        '../data/emoji_sent_race_{}/train/'.format(ratio),\n",
    "        size=100000, ratio=ratio)\n",
    "    x_dev, y_p_dev, y_m_dev = load_data(\n",
    "        '../data/emoji_sent_race_{}/test/'.format(ratio),\n",
    "        size=100000, ratio=0.5)\n",
    "    \n",
    "    biased_classifier = LinearSVC(fit_intercept=True, class_weight='balanced', dual=False, C=0.1, max_iter=10000)\n",
    "\n",
    "    biased_classifier.fit(x_train, y_m_train)\n",
    "    biased_score = biased_classifier.score(x_dev, y_m_dev)\n",
    "    \n",
    "    P = np.load('../data/emoji_sent_race_{}/P_svm.num-clfs=300.npy'.format(ratio), allow_pickle=True)\n",
    "    P = P[1]\n",
    "    n_dims = 120\n",
    "#     n_dims = 70\n",
    "    if ratio == 0.5:\n",
    "        n_dims = 200\n",
    "    elif ratio == 0.6:\n",
    "        n_dims = 100\n",
    "    elif ratio == 0.7:\n",
    "        n_dims = 115\n",
    "    elif ratio == 0.8:\n",
    "        n_dims = 200\n",
    "    P = debias.get_projection_to_intersection_of_nullspaces(P[:n_dims], input_dim=300)\n",
    "    \n",
    "    debiased_x_train = P.dot(x_train.T).T\n",
    "    debiased_x_dev = P.dot(x_dev.T).T\n",
    "\n",
    "    classifier = LinearSVC(fit_intercept=True, class_weight='balanced', dual=False, C=0.1, max_iter=10000)\n",
    "\n",
    "    classifier.fit(debiased_x_train, y_m_train)\n",
    "    debiased_score = classifier.score(debiased_x_dev, y_m_dev)\n",
    "    \n",
    "    p_classifier = SGDClassifier(warm_start=True, loss='log', n_jobs=64, max_iter=10000, random_state=0, tol=1e-3)\n",
    "    p_classifier.fit(debiased_x_train, y_p_train)\n",
    "    p_score = p_classifier.score(debiased_x_dev, y_p_dev)\n",
    "    results[ratio]['p_acc'] = p_score\n",
    "    \n",
    "    _, biased_diffs = get_TPR(y_m_dev, biased_classifier.predict(x_dev), y_p_dev)\n",
    "    \n",
    "    _, debiased_diffs = get_TPR(y_m_dev, classifier.predict(debiased_x_dev), y_p_dev)\n",
    "    \n",
    "#     results[ratio]['biased_diff_tpr'] = biased_diffs['y:0']\n",
    "    results[ratio]['biased_diff_tpr'] = rms(list(biased_diffs.values()))\n",
    "#     results[ratio]['debiased_diff_tpr'] = debiased_diffs['y:0']\n",
    "    results[ratio]['debiased_diff_tpr'] = rms(list(debiased_diffs.values()))\n",
    "    \n",
    "    results[ratio]['biased_acc'] = biased_score\n",
    "    results[ratio]['debiased_acc'] = debiased_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "defaultdict(dict,\n",
       "            {0.5: {'p_acc': 0.517,\n",
       "              'biased_diff_tpr': 0.19000000000000006,\n",
       "              'debiased_diff_tpr': 0.16200000000000003,\n",
       "              'biased_acc': 0.7595,\n",
       "              'debiased_acc': 0.7525},\n",
       "             0.6: {'p_acc': 0.6165,\n",
       "              'biased_diff_tpr': 0.2891666666666667,\n",
       "              'debiased_diff_tpr': 0.21500000000000008,\n",
       "              'biased_acc': 0.783,\n",
       "              'debiased_acc': 0.7375},\n",
       "             0.7: {'p_acc': 0.623,\n",
       "              'biased_diff_tpr': 0.3771428571428571,\n",
       "              'debiased_diff_tpr': 0.24095238095238097,\n",
       "              'biased_acc': 0.8085,\n",
       "              'debiased_acc': 0.6645},\n",
       "             0.8: {'p_acc': 0.6331331331331331,\n",
       "              'biased_diff_tpr': 0.44906407035175877,\n",
       "              'debiased_diff_tpr': 0.15108668341708542,\n",
       "              'biased_acc': 0.8423423423423423,\n",
       "              'debiased_acc': 0.6721721721721722}})"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "defaultdict(dict,\n",
       "            {0.5: {'p_acc': 0.5157539384846211,\n",
       "              'biased_diff_tpr': 0.152341531246204,\n",
       "              'debiased_diff_tpr': 0.11717221236744547,\n",
       "              'biased_acc': 0.7580645161290323,\n",
       "              'debiased_acc': 0.7534383595898975},\n",
       "             0.6: {'p_acc': 0.5878969742435609,\n",
       "              'biased_diff_tpr': 0.23854565356548424,\n",
       "              'debiased_diff_tpr': 0.17784194035108863,\n",
       "              'biased_acc': 0.7576894223555889,\n",
       "              'debiased_acc': 0.7149287321830458},\n",
       "             0.7: {'p_acc': 0.5740185046261566,\n",
       "              'biased_diff_tpr': 0.3248818642517531,\n",
       "              'debiased_diff_tpr': 0.15601789337104066,\n",
       "              'biased_acc': 0.7408102025506377,\n",
       "              'debiased_acc': 0.6475368842210553},\n",
       "             0.8: {'p_acc': 0.5703925981495374,\n",
       "              'biased_diff_tpr': 0.40913329572191975,\n",
       "              'debiased_diff_tpr': 0.15847001555011755,\n",
       "              'biased_acc': 0.7161790447611903,\n",
       "              'debiased_acc': 0.6224056014003501}})"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "plot_results = defaultdict(list)\n",
    "for r in [0.5, 0.6, 0.7, 0.8]:\n",
    "    plot_results['biased_diff_tpr'].append(results[r]['biased_diff_tpr'])\n",
    "    plot_results['debiased_diff_tpr'].append(results[r]['debiased_diff_tpr'])\n",
    "    plot_results['biased_acc'].append(results[r]['biased_acc'])\n",
    "    plot_results['debiased_acc'].append(results[r]['debiased_acc'])\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "defaultdict(list,\n",
       "            {'biased_diff_tpr': [0.19000000000000006,\n",
       "              0.2891666666666667,\n",
       "              0.3771428571428571,\n",
       "              0.44906407035175877],\n",
       "             'debiased_diff_tpr': [0.16200000000000003,\n",
       "              0.21500000000000008,\n",
       "              0.24095238095238097,\n",
       "              0.15108668341708542],\n",
       "             'biased_acc': [0.7595, 0.783, 0.8085, 0.8423423423423423],\n",
       "             'debiased_acc': [0.7525, 0.7375, 0.6645, 0.6721721721721722]})"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "plot_results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "def data2plot(results):\n",
    "    x = [0.5, 0.6, 0.7, 0.8]\n",
    "\n",
    "    fig, ax1 = plt.subplots()\n",
    "\n",
    "    color = 'tab:red'\n",
    "    ax1.set_xlabel('ratio')\n",
    "    ax1.set_ylabel('diff tpr', color=color)\n",
    "    ax1.plot(x, results['biased_diff_tpr'], '.--', label='biased tpr diff', color=color)\n",
    "    ax1.plot(x, results['debiased_diff_tpr'], '*:', label='debiased tpr diff', color=color)\n",
    "    ax1.tick_params(axis='y', labelcolor=color)\n",
    "    plt.ylim(0.,0.6)\n",
    "\n",
    "    ax2 = ax1.twinx()  # instantiate a second axes that shares the same x-axis\n",
    "\n",
    "    color = 'tab:blue'\n",
    "    ax2.set_ylabel('acc', color=color)  # we already handled the x-label with ax1\n",
    "    ax2.plot(x, results['biased_acc'], '^-', label='biased acc', color=color)\n",
    "    ax2.plot(x, results['debiased_acc'], 'o--', label='debiased acc', color=color)\n",
    "    ax2.tick_params(axis='y', labelcolor=color)\n",
    "    \n",
    "    # ask matplotlib for the plotted objects and their labels\n",
    "    lines, labels = ax1.get_legend_handles_labels()\n",
    "    lines2, labels2 = ax2.get_legend_handles_labels()\n",
    "    ax2.legend(lines, labels, loc='lower right')\n",
    "    \n",
    "    from matplotlib.legend import Legend\n",
    "    leg = Legend(ax2, lines2, labels2,\n",
    "             loc='upper left', frameon=False)\n",
    "    ax2.add_artist(leg);\n",
    "\n",
    "    plt.title('TPR rates as a function of the ratio')\n",
    "    plt.ylim(0.5,.85)\n",
    "#     plt.legend()\n",
    "#     plt.savefig('tpr_rates_ratio.png', dpi=1000)\n",
    "    plt.show()\n",
    "    \n",
    "# data2plot(plot_results)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "data2plot(plot_results)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "defaultdict(dict,\n",
       "            {0.5: {'p_acc': 0.517,\n",
       "              'biased_diff_tpr': 0.19000000000000006,\n",
       "              'debiased_diff_tpr': 0.16200000000000003,\n",
       "              'biased_acc': 0.7595,\n",
       "              'debiased_acc': 0.7525},\n",
       "             0.6: {'p_acc': 0.6165,\n",
       "              'biased_diff_tpr': 0.2891666666666667,\n",
       "              'debiased_diff_tpr': 0.21500000000000008,\n",
       "              'biased_acc': 0.783,\n",
       "              'debiased_acc': 0.7375},\n",
       "             0.7: {'p_acc': 0.623,\n",
       "              'biased_diff_tpr': 0.3771428571428571,\n",
       "              'debiased_diff_tpr': 0.24095238095238097,\n",
       "              'biased_acc': 0.8085,\n",
       "              'debiased_acc': 0.6645},\n",
       "             0.8: {'p_acc': 0.6331331331331331,\n",
       "              'biased_diff_tpr': 0.44906407035175877,\n",
       "              'debiased_diff_tpr': 0.15108668341708542,\n",
       "              'biased_acc': 0.8423423423423423,\n",
       "              'debiased_acc': 0.6721721721721722}})"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Ratio & Sentiment (Acc) & Sentiment + INLP (Acc) & Diff TPR & Diff TPR + INLP\n",
      "0.5 \t 0.76 & \t 0.75 & \t\t 0.15 & \t\t 0.12\n",
      "0.6 \t 0.76 & \t 0.71 & \t\t 0.24 & \t\t 0.18\n",
      "0.7 \t 0.74 & \t 0.65 & \t\t 0.32 & \t\t 0.16\n",
      "0.8 \t 0.72 & \t 0.62 & \t\t 0.41 & \t\t 0.16\n"
     ]
    }
   ],
   "source": [
    "print('Ratio & Sentiment (Acc) & Sentiment + INLP (Acc) & Diff TPR & Diff TPR + INLP')\n",
    "for ratio, vals in results.items():\n",
    "    print('{} \\t {:.2f} & \\t {:.2f} & \\t\\t {:.2f} & \\t\\t {:.2f}'\n",
    "          .format(ratio, vals['biased_acc'], vals['debiased_acc'], vals['biased_diff_tpr'], vals['debiased_diff_tpr'], '.2f'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Accuracy - TPR rate tradeoff"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "ratio = 0.8\n",
    "x_train, y_p_train, y_m_train = load_data(\n",
    "        '../data/emoji_sent_race_{}/train/'.format(ratio),\n",
    "        size=40000, ratio=ratio)\n",
    "x_dev, y_p_dev, y_m_dev = load_data(\n",
    "    '../data/emoji_sent_race_{}/dev/'.format(ratio),\n",
    "    size=8000, ratio=ratio)\n",
    "\n",
    "all_Ps = np.load('../data/emoji_sent_race_{}/P_svm.num-clfs=300.npy'.format(ratio), allow_pickle=True)\n",
    "rowspaces = all_Ps[1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tqdm import tqdm_notebook as tqdm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "231"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(rowspaces)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(300, 300)"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rowspaces[0].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bcf624468a514891b24043c91bdcdf7c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=200), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "ratio_tradeoff = defaultdict(dict)\n",
    "\n",
    "for p_iteration in tqdm(range(1, 201)):\n",
    "    \n",
    "    P = debias.get_projection_to_intersection_of_nullspaces(rowspaces[:p_iteration], input_dim=300)\n",
    "    \n",
    "    debiased_x_train = P.dot(x_train.T).T\n",
    "    debiased_x_dev = P.dot(x_dev.T).T\n",
    "    \n",
    "    classifier = LinearSVC(fit_intercept=True, class_weight='balanced', dual=False, C=0.1, max_iter=10000)\n",
    "\n",
    "    classifier.fit(debiased_x_train, y_m_train)\n",
    "    debiased_score = classifier.score(debiased_x_dev, y_m_dev)\n",
    "    \n",
    "    _, debiased_diffs = get_TPR(y_m_dev, classifier.predict(debiased_x_dev), y_p_dev)\n",
    "    \n",
    "    ratio_tradeoff[p_iteration]['debiased_diff_tpr'] = debiased_diffs['y:0']\n",
    "    ratio_tradeoff[p_iteration]['debiased_acc'] = debiased_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# ratio_tradeoff"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def data2plot(x, y1, y2, x_label=None, y_label=None, title=None):\n",
    "\n",
    "    # label is optional, and is used for the legend\n",
    "    # here, the third arguemtns are a mix of the `marker' and the `linestyle'\n",
    "    ax = plt.plot(x, y1, '.--', label='TPR-Gap')\n",
    "    ax = plt.plot(x, y2, '*:', label='accuracy')\n",
    "\n",
    "    if y_label:\n",
    "        plt.ylabel(y_label)\n",
    "    if x_label:\n",
    "        plt.xlabel(x_label)\n",
    "    if title:\n",
    "        plt.title(title)\n",
    "\n",
    "    # Optional (and many times useful)\n",
    "    # plt.ylim(0.5,0.90)\n",
    "    plt.legend(loc='lower left', frameon=True)\n",
    "    \n",
    "    plt.tight_layout()\n",
    "    # \n",
    "    plt.savefig('acc_tpr.pdf', dpi=1000)\n",
    "    plt.show()\n",
    "    \n",
    "data2plot(list(ratio_tradeoff.keys()),\n",
    "          [x['debiased_diff_tpr'] for x in ratio_tradeoff.values()],\n",
    "          [x['debiased_acc'] for x in ratio_tradeoff.values()], 'P iter', title='')"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "nullspace",
   "language": "python",
   "name": "nullspace"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}