ROC_analysis_HoeffdingTest.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from __future__ import division"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "%run util.py"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "data_dir = \"/home/jzh/Dropbox/tsp_rev_refs/anomaly-weakconv-waze/\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def wcKLs_testSet(N, n):\n",
    "    # mu_0: the initial distribution; 1 x (N**2) \n",
    "    # mu: the actual stationary distribution; 1 x (N**2)\n",
    "    # mu_1: the estimated stationary distribution \n",
    "    # P: the new transition matrix (the old transition matrix is Q)\n",
    "    # G_1: the estimate of the gradient\n",
    "    # H_1: the estimate of the Hessian\n",
    "\n",
    "    # P is the ground truth transition matrix\n",
    "    mu_0, mu, mu_1, P, G_1, H_1, Sigma_1, W_1, rho_1, Chi_1 = ChainGen(N)  \n",
    "\n",
    "    # PP is a different transition matrix for testing purposes\n",
    "    mu_0P, muP, mu_1P, PP, G_1P, H_1P, Sigma_1P, W_1P, rho_1P, Chi_1P = ChainGen(N) \n",
    "\n",
    "    KL_wc_1, KL_wc_2, _, _ = ThresWeakConv(N, 0.1, rho_1, n, mu_0, mu, mu_1, P, \\\n",
    "                                                           G_1, H_1, Sigma_1, W_1, Chi_1).ThresCal()\n",
    "    \n",
    "    negative_sample = {}\n",
    "    positive_sample = {}\n",
    "\n",
    "    # Get sample paths of the Markov chain with length n; \n",
    "    # these paths will be the test set\n",
    "    for idx in range(int(num_test_sample/2)):\n",
    "        negative_sample[idx] = chain(mu, P, n)\n",
    "    for idx in range(int(num_test_sample/2)):\n",
    "        positive_sample[idx] = chain(muP, PP, n)\n",
    "    \n",
    "    mu = np.reshape(mu, (N, N))\n",
    "    negative_sample_KL = {}\n",
    "    positive_sample_KL = {}\n",
    "    \n",
    "    # Compute KL divergence values w.r.t. mu for each sample\n",
    "    for idx in range(int(num_test_sample/2)):\n",
    "        negative_sample_KL[idx] = KL_est(negative_sample[idx], mu)\n",
    "        \n",
    "    for idx in range(int(num_test_sample/2)):\n",
    "        positive_sample_KL[idx] = KL_est(positive_sample[idx], mu)\n",
    "        \n",
    "    return KL_wc_1, KL_wc_2, negative_sample_KL, positive_sample_KL"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def roc_data(n, beta, KL_wc_1, KL_wc_2, negative_sample_KL, positive_sample_KL):\n",
    "    \n",
    "    # Get thresholds for Hoeffding's test corresponding to sample length n    \n",
    "    key = str(n) + \"_\" + str(beta)\n",
    "    \n",
    "    eta_wc_1 = {}\n",
    "    eta_wc_2 = {}\n",
    "    eta_sv = {}\n",
    "\n",
    "    eta_wc_v1 = prctile(KL_wc_1, 100 * (1 - beta))\n",
    "    eta_wc_v2 = prctile(KL_wc_2, 100 * (1 - beta))\n",
    "    eta_sav = - log(beta) / n\n",
    "    \n",
    "    eta_wc_1[key] = eta_wc_v1\n",
    "    eta_wc_2[key] = eta_wc_v2\n",
    "    eta_sv[key] = eta_sav\n",
    "\n",
    "    negative_sample_wc_1 = {}\n",
    "    negative_sample_wc_2 = {}\n",
    "    negative_sample_sv = {}\n",
    "\n",
    "    positive_sample_wc_1 = {}\n",
    "    positive_sample_wc_2 = {}\n",
    "    positive_sample_sv = {}\n",
    "\n",
    "    for idx in range(int(num_test_sample/2)):\n",
    "        if negative_sample_KL[idx] > eta_wc_1[key]:\n",
    "            negative_sample_wc_1[idx] = 1\n",
    "        else:\n",
    "            negative_sample_wc_1[idx] = -1\n",
    "        if negative_sample_KL[idx] > eta_wc_2[key]:\n",
    "            negative_sample_wc_2[idx] = 1\n",
    "        else:\n",
    "            negative_sample_wc_2[idx] = -1\n",
    "        if negative_sample_KL[idx] > eta_sv[key]:\n",
    "            negative_sample_sv[idx] = 1\n",
    "        else:\n",
    "            negative_sample_sv[idx] = -1    \n",
    "\n",
    "\n",
    "    for idx in range(int(num_test_sample/2)):\n",
    "        if positive_sample_KL[idx] > eta_wc_1[key]:\n",
    "            positive_sample_wc_1[idx] = 1\n",
    "        else:\n",
    "            positive_sample_wc_1[idx] = -1\n",
    "        if positive_sample_KL[idx] > eta_wc_2[key]:\n",
    "            positive_sample_wc_2[idx] = 1\n",
    "        else:\n",
    "            positive_sample_wc_2[idx] = -1\n",
    "        if positive_sample_KL[idx] > eta_sv[key]:\n",
    "            positive_sample_sv[idx] = 1\n",
    "        else:\n",
    "            positive_sample_sv[idx] = -1 \n",
    "\n",
    "    TP_wc_1 = count_one(positive_sample_wc_1.values())[0]\n",
    "    FP_wc_1 = count_one(negative_sample_wc_1.values())[0]\n",
    "    TN_wc_1 = count_one(negative_sample_wc_1.values())[1]\n",
    "    FN_wc_1 = count_one(positive_sample_wc_1.values())[1]\n",
    "    TPR_wc_1 = TP_wc_1 / (TP_wc_1 + FN_wc_1)\n",
    "    FPR_wc_1 = FP_wc_1 / (FP_wc_1 + TN_wc_1)\n",
    "    Precision_wc_1 = TP_wc_1 / (TP_wc_1 + FP_wc_1)\n",
    "    Recall_wc_1 = TP_wc_1 / (TP_wc_1 + FN_wc_1)\n",
    "    assert(TP_wc_1 + FN_wc_1 == int(num_test_sample/2))\n",
    "    assert(FP_wc_1 + TN_wc_1 == int(num_test_sample/2))\n",
    "    assert(Recall_wc_1 == TPR_wc_1)\n",
    "\n",
    "    TP_wc_2 = count_one(positive_sample_wc_2.values())[0]\n",
    "    FP_wc_2 = count_one(negative_sample_wc_2.values())[0]\n",
    "    TN_wc_2 = count_one(negative_sample_wc_2.values())[1]\n",
    "    FN_wc_2 = count_one(positive_sample_wc_2.values())[1]\n",
    "    TPR_wc_2 = TP_wc_2 / (TP_wc_2 + FN_wc_2)\n",
    "    FPR_wc_2 = FP_wc_2 / (FP_wc_2 + TN_wc_2)\n",
    "    Precision_wc_2 = TP_wc_2 / (TP_wc_2 + FP_wc_2)\n",
    "    Recall_wc_2 = TP_wc_2 / (TP_wc_2 + FN_wc_2)\n",
    "    assert(TP_wc_2 + FN_wc_2 == int(num_test_sample/2))\n",
    "    assert(FP_wc_2 + TN_wc_2 == int(num_test_sample/2))\n",
    "    assert(Recall_wc_2 == TPR_wc_2)\n",
    "\n",
    "    TP_sv = count_one(positive_sample_sv.values())[0]\n",
    "    FP_sv = count_one(negative_sample_sv.values())[0]\n",
    "    TN_sv = count_one(negative_sample_sv.values())[1]\n",
    "    FN_sv = count_one(positive_sample_sv.values())[1]\n",
    "    TPR_sv = TP_sv / (TP_sv + FN_sv)\n",
    "    FPR_sv = FP_sv / (FP_sv + TN_sv)\n",
    "    Precision_sv = TP_sv / (TP_sv + FP_sv)\n",
    "    Recall_sv = TP_sv / (TP_sv + FN_sv)\n",
    "    assert(TP_sv + FN_sv == int(num_test_sample/2))\n",
    "    assert(FP_sv + TN_sv == int(num_test_sample/2))\n",
    "    assert(Recall_sv == TPR_sv)\n",
    "\n",
    "    data_wc_1 = {\"TPR\": TPR_wc_1, \"FPR\":FPR_wc_1, \"Precision\":Precision_wc_1, \"Recall\":Recall_wc_1}\n",
    "    data_wc_2 = {\"TPR\": TPR_wc_2, \"FPR\":FPR_wc_2, \"Precision\":Precision_wc_2, \"Recall\":Recall_wc_2}\n",
    "    data_sv = {\"TPR\": TPR_sv, \"FPR\":FPR_sv, \"Precision\":Precision_sv, \"Recall\":Recall_sv}\n",
    "    \n",
    "    return data_wc_1, data_wc_2, data_sv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "N = 2\n",
    "n = 20\n",
    "\n",
    "# N = 4\n",
    "# n = 50\n",
    "\n",
    "# N = 6\n",
    "# n = 100\n",
    "\n",
    "# N = 8\n",
    "# n = 200\n",
    "\n",
    "num_test_sample = 2000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "beta_list = [1e-3] + list(np.arange(0, 0.21, 0.01)[1:-1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0.001,\n",
       " 0.01,\n",
       " 0.02,\n",
       " 0.029999999999999999,\n",
       " 0.040000000000000001,\n",
       " 0.050000000000000003,\n",
       " 0.059999999999999998,\n",
       " 0.070000000000000007,\n",
       " 0.080000000000000002,\n",
       " 0.089999999999999997,\n",
       " 0.10000000000000001,\n",
       " 0.11,\n",
       " 0.12,\n",
       " 0.13,\n",
       " 0.14000000000000001,\n",
       " 0.14999999999999999,\n",
       " 0.16,\n",
       " 0.17000000000000001,\n",
       " 0.17999999999999999,\n",
       " 0.19]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "beta_list"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<string>:330: RuntimeWarning: covariance is not positive-semidefinite.\n"
     ]
    }
   ],
   "source": [
    "KL_wc_1, KL_wc_2, negative_sample_KL, positive_sample_KL = wcKLs_testSet(N, n)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "roc_data_wc_1 = {}\n",
    "roc_data_wc_2 = {}\n",
    "roc_data_sv = {}\n",
    "\n",
    "for beta in beta_list:\n",
    "    roc_data_wc_1[beta], roc_data_wc_2[beta], roc_data_sv[beta] = \\\n",
    "    roc_data(n, beta, KL_wc_1, KL_wc_2, negative_sample_KL, positive_sample_KL)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAm8AAACyCAYAAAAUCShQAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nU6JJKA8AAAANSURBVBRFURRFySL+f6rRQLBhrQZZAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1de7ba9b50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import pylab\n",
    "from pylab import *\n",
    "%matplotlib inline  \n",
    "import itertools \n",
    "\n",
    "\n",
    "plt.figure(figsize=(10, 2))\n",
    "\n",
    "font = {\"family\" : \"normal\",\n",
    "        \"weight\" : \"normal\",\n",
    "        \"size\"   : 14}\n",
    "\n",
    "matplotlib.rc('font', **font)\n",
    "\n",
    "# from matplotlib import rc\n",
    "# rc('font',**{'family':'sans-serif','sans-serif':['Helvetica']})\n",
    "# ## for Palatino and other serif fonts use:\n",
    "# #rc('font',**{'family':'serif','serif':['Palatino']})\n",
    "rc('text', usetex=True)\n",
    "\n",
    "roc_sv_pr, = plt.plot([beta for beta in beta_list], \\\n",
    "                   [roc_data_sv[beta][\"Precision\"] for beta in beta_list], \"g-^\", linewidth=1,\n",
    "                      markerfacecolor=\"None\", markeredgecolor=\"g\", markeredgewidth=1)\n",
    "\n",
    "roc_wc_1_pr, = plt.plot([beta for beta in beta_list], \\\n",
    "                     [roc_data_wc_1[beta][\"Precision\"] for beta in beta_list], \"b-s\", linewidth=1, \\\n",
    "                        markersize=5, markerfacecolor=\"None\", markeredgecolor=\"b\", markeredgewidth=1)\n",
    "\n",
    "roc_wc_2_pr, = plt.plot([beta for beta in beta_list], \\\n",
    "                     [roc_data_wc_2[beta][\"Precision\"] for beta in beta_list], \"m-o\", linewidth=1, \\\n",
    "                        markerfacecolor=\"None\", markeredgecolor=\"m\", markeredgewidth=1)\n",
    "\n",
    "legend_list = [r\"HTSV\", r\"HTWC-1\", r\"HTWC-2\"]\n",
    "\n",
    "# legend(loc=0, fancybox=\"true\", ncol=2, mode=\"expand\", borderaxespad=0.2) \n",
    "\n",
    "plt.legend([roc_sv_pr, roc_wc_1_pr, roc_wc_2_pr], \\\n",
    "           legend_list, loc=4, ncol=3, mode=\"expand\", borderaxespad=0.2, numpoints=1, fancybox=\"true\", \\\n",
    "           frameon=False) \n",
    "plt.title(r\"$N = %s, n = %s$\"%(N, n))\n",
    "plt.xticks(np.arange(-0.02, 0.24, 0.02)[1:-1])\n",
    "\n",
    "pylab.xlim(-0.01, 0.2)\n",
    "pylab.ylim(0.25, 1.05)\n",
    "\n",
    "plt.xlabel(r\"target \\textit{FPR} ($\\beta$)\")\n",
    "plt.ylabel(r\"\\textit{precision}\")\n",
    "# plt.grid()\n",
    "\n",
    "\n",
    "savefig(data_dir + \"ROC_data_precision_N_%s_n_%s.eps\"%(N, n), dpi=300, bbox_inches=\"tight\")\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAm8AAACyCAYAAAAUCShQAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzt3Xt0VNed4PvvrwTmYaD0sGMTx0YSEMePxHoUTneSbssg\nhJOZdGKDRJLb47TbCKH07e6VG7AgXnNj31l+gJysns69AUk4Gc9NX1sSOJ309LSBElGn7STNQ4I4\nHscBJHASY2xTeoDNU/rdP86pUlWpSlKhUpVK/D5r1ao6++xz9t6qourH3vvsI6qKMcYYY4zJDJ50\nV8AYY4wxxoydBW/GGGOMMRnEgjdjjDHGmAxiwZsxxhhjTAax4M0YY4wxJoNY8GaMMcYYk0EseDPG\nmAwgIt5018EYMzmIrfNmjLmaiMhuVa1Idz3GQkSWAXvCko4By1X1eFieakABAbyq+kxKK2mMSTnr\neTPGJJ2IFIvINhEZFJGtUfu8ItIiIgER2Soi81JYr1XAslSVlwTZQIn7KFTVxTECN6+qblfVJqBb\nRJ5OT1WNMaliPW/GmAnh9hqVA4+oalbUPi9QqarbU1gfL7AWeDq6PpOViKwEDoYHbFH7jwLlUQFd\nQFVzU1NDY0w6WM+bMWaiFKjqJqBPRNZE7SsHWlJcn0qgMcVlThg3GC2IEdhli0hRGqpkjEkRC96M\nMROtEaiLSstV1f5UVUBEioEDqSovyZaLyAMiUh01JFoYJ3/vCPuMMVPAtHRXwBgzZQXc5wZgg4gU\nqeohNy3V8zVKVXV7oldsunPKFgLNDAVEhQCqWp/cKsZ0DDgW/Lu5Adw2VV0HxBsaDYywzxgzBdic\nN2NM0rnz3fYHe9dEZDfQo6qr3QCqVFX3jnKOFoauoowWTFdgz0hz50RkparudF97gcBY5ry5vXXH\ngBqgClgW1p5BVR1x5CJZ9Y86Z4Fbp2xgCbA7xnzCozjz+lI2n9AYk1rW82aMmQiFqtoWtr0Z2O1e\nWVoeDKZGoqpV462EG+z0hiclcHiPqvaLyBJgW1jgFgzqRpSM+sc4Z7eIgNP7F4iTLXeEfcaYKcCC\nN2PMRIjo0lfVNhHpwunF6klhPcqBQhEpd7dzAETkKZyewRfjHRh2IUA5EH7BRTngT35VI7m9hN1A\nSbAuYcO+6u5DROZFzR/MBromun7GmPSx4M0Yk1TBockYuxqBTcAjYzxPcNgxbhZGGXZ01z4LP2cx\nUO1eBTuWOhQDp6OCoxpgpdtOjXfhRTLqjxNgHg/bXuiWedgtowunpy28Dho2t9AYMwVZ8GaMSbYq\nYi8D0gg8zRh7hSZi2JEYw6bu0Gp5dKDniuhlc/PmqOphEamOcwww/vqrap+I7IlK3khk8LsZJ5jc\n5NavmuFX9hpjphi7YMEYkxRuT9RmnPXU/Dg9XP1Rebaqam2a6lft1m0ZsANoUNW9bvrTOGumRde3\nOZgvLG0rzi2rulLRwyUiG9yXC4ED0b10IrIeZ15fDs4SLGPqVTTGZC4L3owxVz0RyYeIeW7GGDNp\n2SK9xhgTdbN3Y4yZzCx4M8Zc1dzh3lGX/jDGmMnChk2NMcYYYzJIRl9tKiIWeRpjjDEmY6hqIouF\nx5Txw6aqOuGPb33rWykpJ5WPqdimqdoua1NmPKxNmfGwNmXOYyq2K1kyPngzxhhjjLmaWPBmjDHG\nGJNBUhq8icjuMeQpEJH1IrLMffaOdsxEKysrS3cVkm4qtgmmZrusTZnB2pQZrE2ZY6q2KxlScrWp\niCwDCoFtqpo1St4DqupzX3uBJo1zmxkR0VTU3xhjjDFmvEQEzZQLFlS1TUe4B2BQ8CbQYcf14dxb\n8Kpw8sxJ7vvhfbx99u0pUc5ULcvalBllTcU2pbIsa1NmlDUV25TKslLZpmSabHPeCnHu0RcuICJF\n6ahMtIl+k7e8soV9f9jHlle2TMj5U13OVC3L2jT5yzr1/Cn2f3w/Gx7cwL4793Hq+VNW1iQsJ5Vl\nWZusrHSVMxFSukiviAyMNGzq3iC6XFVXh6UdBVZpjBtAp3rY9OsvfZ3nDj/HXxT9Bd9Z8Z2YeVQV\nRRN+PnnmJH/07B/x4y/9mC++8EV++fAvuWHODUlvw9tn305JOQC3f3Q2b/1uekTazbcM8uobZzK2\nrIdvep//EPg9t/ABbzKbH7KAI7dcPyFtSlVZU61NgeYA+x9+m6cGPsarePk4fWzgDf4xL5+m389K\nWjlTtSxrU2aUNRXblMqy4pXz4+sKaH53Yn4TIXnDppMteFsJrFXVFWFpAWBpvODtW9/6Vmi7rKxs\nwiY4njxzkoL/WsCADnB58DJZ4jQjPACLqBuCiIz5+cLlC4gIM6fN5Pzl8wDMnDYz6e3o33wYevMj\n65p9grl1nxj3uVWBgekwcA0MXMOZ/3Kc6d9YzAzPLC4MXADg0nfeYM6G8ZcV7e76PfynPD83B+bz\nu9y3+f+WvMSel/471/4fJYCAysjPEHuf+29MVbjnyJ38uf8rfOc/bOW3Hz7JR9+azzd+9gW2n/k0\nv1z3vyW1PfccuYM/b/uyU9b8U3z0Dx/mG//2ebaf/RS/XLMG1AODHlBB1ePW1xP1kIjXOizNwz1v\n3sKDB/6Ub//xLt7IO8Ot73lZf7iE7ed8vPIfvz22v9GwZ9zXkfnvefs6vvraJ/j2Xft5PecDbuuZ\nzTfeWMyz54t5eekPhhqvTmmi4EFCrwXB4z47+0DU3R/aB5s7bucH50r4zSf/O9M80xgYvMztfXN5\n8DfFPPrHP0/q+/TkoY/xg3PFvPHJH5LlmcbAwAC398/jP/3mE3zr7v1u/Zy/xdBrt8ZRaeF5nf3B\nljn5Nh69kZYLt3HsY+1kkYXqIIvfn8HK33+EZ2590/27DZ03eI6h8wdr7aa771UwT/CYvzo5g3++\nXMjxm/4XWSIMDioFFzx87rSXhpv6CH4awn99ht6jsO1gnqj08P1//u41/PTyzfzhhqNkiQfVQT5y\nwcM9vXN54YYzI54v5jlj1AHgzwLT+feB+ZzKexOPCIOq3HhJuLv/Wv7Hde8n/L6P5PM90/j3gfm8\nPcFlhZeT5QQE3HhJWNJ/Lf+SN1RO9N8q+DmI+bcNpkX8jYWyvmm8OphHIPtt9/OkXHdZuOP9Wbyc\nfS7m3z7iOcZ7FvxcR9ez+P0suga99M05HQx08F4WCs/P4NDcC+P4i0UqCiun/+7zXNqwl+t+dR0f\nrlvOQ3p30sppb2+nvb09tP34449PyeCtGGhU1SVhaadVNS9O/pT1vH39pa8zoANsWb6FR/Y8gojw\nzPJnYgZiiTp55iR3fO8OXvvaa8yfO5+TZ05y59Y7ee1rr3HjnBuvqL6qcPYs9PY6j54e6DoZ4KEv\n5fJ/be5DLno53X+Gbft+wPl//Rtqa+HixaHHhQsjb8dLmz4drrkGpk8fpKT3Xf5i2gluuvwBb02f\nzX+75lra3r+D/PwralJcS86e5IvvneC5GxbRNSuXhecCPPju6zw7+FF+t9j5H5TI6I9QPpTpDHKN\nDj2m6wBfPfIqz1/8CDM+eg0zJIvBgUvIO118rt9D26JCBPCoIqpDr1E3qFB3G0SHXsfKLwq+d05y\nYCAbz4emMd3jQXUA+t/hjnNZHJ1/PR4UEZxn4j87D3UCnVCau63Kde/2c/zyLObc6GEaHgYHBzjf\nH+CG8zP54EPXhuoj7r8zUefDFWyH80MQnUfdYCVyX9b5y1xQD9OuAY97noGBAaapBzwydByh+AV1\n3pDQMwxPi36+5sxF3uEa5npB3ICg/3w/N1yYyfns5P6HaGbved5mBl6vIh6nrL7z/dx4fibn89xe\ngtCvloRi2tDXRES7Irdxt8XdN+vEGV5nDjcvGGRa1jQu6wBv9r3J7YE8zi/2RpYl4pQRXU7YuSU8\nLWx7+oHT/JxcinyXmZ51DZcGL/Krtw/zR78r4NIfXx/1ax9V36g2hd4XiMrnvld7TvI/uJF7ll1k\nxvSZnB+4wMsn/o3P/fZ2Ln7upvjnGWPZwT/0jNYT/L/cwucfOM/Ma2Zz7tI5XjryL3zpVyVc+FJ+\nnHf3ysx84TjPsYA/W3meWdNnTVhZw8q5fI5/OfI/+fLhUi58pWD0v00wLbgtON8Qw6JfYdb2I/y9\nLuIrf3mOa6+5lvcvvU/rqy2s2/dpzn/to6OfJ7z8Ueozu/41HtPbqf36+8ydMYczF87yg4Pfp+5n\nyzi38Y7x/+Fcsze/xrf0dr729fdZ/EfX8+HPnuGu//sunv9mC8u0LGnlREtWz1uqb481rMIiUgAE\nVLVPVTvDgx93nz+F9Yvp5JmTPHf4OV772mvMnDaTTZ/ZxJ1b72TTZzZdcXAVbssrW7j8naN8eGOu\nmzIfOM1Hn+nh9c6hACwYhIVvx9vX1wczZ0J2NuTkOM9/uHgSyOWd33mZNw9y5syltLCAV/4V7rjD\nCbpmzHCeg49EtqdPB487i/KZjd/hw5s/w2d334r3M176Xu4j5ys/R9+/jrbuxLukBy8NMvjBIAMf\nDDjP5wZC279Y3Ukzt/HclssMnnubgXMDtL10jP99l4dPfOGsk/fcYOgRvT0s7fwgMl3wzPKQNSsL\nzywPnlke3r94nnt4l3sWZCHXCJIlvHrdCa7/5Yf5m5L3kCwBD0iWXOFrCb0+tvES/4t5fPubHsgC\n8Qg7fr2bOVvv4kuPz0U87rEe9wsw+DosLWLb7ZyLTjtccZhtLKTtBY+zP0v4+1/8PTdv+BM+9U+L\nI88rErOcsZb9yode4Ut8ktPvZDltEqHupTo+u+o/cu/5Pxk6LhhIXKGf5P+Ep0/8CZ29OaG0JzY/\nwbmNy3moJ3n/ow6WtSVGWec3Lueh95JfVtOJu+g8Hl7Wbo5vXM5Dvy1Oajk7TtzFo/uHyjm0+V84\nvvEOHvp58n48nbLeYO+Ju/i2f6isX2/+Ccc3zuahf16cxHJepfNEDt/fOVTO65tf5MSvZvPQ8/lJ\nKwfgJ7/4FYdPZPPfdkxsWXHLOTybh/5hQdLKAfjJ7oMcP3Etlc9+JJT2281wYt9sHvp/Ppzcspr7\n6T8xnS99p8BNuZ5jm2dy4mezeeip65NXzvP9nIkoZw5/6/lb3mR20sqYSCkJ3twetXJAReQpYI+q\n7nV3bwZ2A9vd7WoRWQ90AsVAdSrqOJItr2zhq3d9lflz5wMwf+58HvzEg2x5ZUvcuW+J+LdfvcmZ\nd3KhcLczpHk+B85nc+bdHJYscQKv4CMYiGVnww03wK23xt7n9TrBVLjPfL+Ggn07KWn+n9zy3i28\ned2bzPjTHwKf56/+atzNYPDiIBf7LnO57zIF3y+gmZvYtu0LzPzuTGZemknhRwr5q7ez6Xr0/aFA\n7FxYQBb1HL5PVcmanYVntmfY86x3ZrGIs3z3me1cmH6BC9MvcNFzkdXkM/366cyYNcMJwGZGBmPR\nwVloe6bHCaaiNN/UzA/eKuJvP50bSivqLuJvftnAvc3J/VH7xX/9Bb99q4ibe8PK+qCIv6GBpdXJ\n+7J8Z/47DL4l5LQP/QAUdRfxEVaw9O55SSsnWFbhWx8w/e8i/353UIVnevKunXppxUtsaPwQxX9R\ny6u3vMrH3/w4G368gWdZwENJK2XqlmVtyoyypmKbUllWKts0EVI6bJpsqRo2/cz3P8Mrv3tlWPqn\nb/40L//lywmfTxV+9Sv4x390Hr//Pbz3HvzoR3DbbU4Q5vU6PWfJbN6p50/R9pVuPrt3qDfsjYff\n4NHuAvyXP8RlN/C63HeZgb4BLvdejki73Oumx0kbvDjINO80pnmncb77PK9f4+X0xWl8wDQu4OEi\nwgO8RcF/yY8diM0aHpgFn0f6cd935z7+z/cWs+tUWABCD9+YfoQ/v5i83o9Tz5/iXx/s5qnLt074\nBNdUlTUV2wSw+vpT3PfeiYgLI44tuIHjx5NazJQty9qUGWVNxTalsqxUtikoIy9YSLZMWqT38mV4\n+eWhgC0rC77wBfjiF+FTn3J6yaKbIpK84G3ggwH2f3w/P343j0tnBlnA+8zlMtlcZB6X8Xgga24W\n07KnhQKwad5pTMueRpY3a+Q09xjPbE9ouGvfnftY/N3F5Nw7FFD1/LSHI399hLt/ndzhpFPPn6L7\n0W5ufTYyKC14ooAbvpz8AOTEEyf44PUPmH3bbBY8uiDpZaS6rKnYJmOMmYwseGPyB2/vvw+7dsGP\nfwz//M+Qnz8UsN15Z9ikZWIHalcavF3qucTZzrOc6TjD2c6znO08y/nj5xk8N8gND93APN88rr3j\nWqblTsNzrYd9i/dxz6V7nPlGSZLKgCpYngUFxhhjJjML3picwds778A//ZMTsLW3wyc/6QRsf/Zn\ncMst8Y/Lz4cTJyLTFixgxO5bVeXiWxc50zkUpJ3pOMPl05eZUzSHOcVDj2tvv5YDJQdS1hsGFlAZ\nY4wx4Sx4Y/IEb0ePOkOhP/4xvPoqVFQ4AdvnPufMXxuL0QIdHVTOdZ3jbIcbpLkBG4OEArS5JXOZ\nUzyHWYtmxexFS3VvmDHGGGOGWPBGaoO3WD1jXi/cdBOcPj00HHrvvc6FBomIDqp623v5zUO/4brP\nX4dcI06v2uGzTMue5gRpxU6QNqdkDjNumpHQsgrWG2aMMcakhwVvpDZ4E3EuOPiHf3B62ObOhTfe\ngF/8Au6+e2h9syux7/Z9fOgrH+LiyYuc2XeG9197n2nXTWOgf4AFjy4IBWzT86aPfjJjjDHGTEoW\nvJH64G3bNmcB3C98AT72sSu/oEBVOXv4LD27egi8FKC3vRfvvV7yPpuH91Nerr3rWjwzPPxs5s8o\nGyhLeluMMcYYk3qZeoeFjFZTc+XHXjp9icCeAIGXAvTs6iFrTha59+XykW98hIunLpL/n/OHXUgw\n+7bMWOnZGGOMMaljwdsEGbw8yJl9ZwjscgK2D17/gOyybHLvyyX/P+cza+GsUN6BMwO88fAbMS8k\nMMYYY4wJl7Lgzb1P6UqGbnvVpKp9cfIWAz530wvsVNXulFQ0jgULItdlC6aFu/CHC6Fgrcffw4yb\nZ5B7Xy6FTxfi/ZQXz4zYE+OCFwwc+esjoQsJ7ApQY4wxxsSSsjlvInJAVX3uay9O8FYVJ+8GVa0P\n296mquti5EvZnLdYV2le/8D19P5bb2ju2oW3LpCzPIfc+3LJrchlxodnpKRuxhhjjJn8MuqCBbcn\n7WlVXRGWFlDV3Dj5jwKlwZ45EdmqqrUx8qUkeAtfymP6jdN5a+tbvL39bVSVucVzyV2RS+59ucz1\nzY15Q3NjjDHGmEy7YKEQ6I1KC4hIkaoeipF/M9AtInXudl2MPClz4okT3Prsrby7813e+9F75N6X\ny83rb+bd1ncp+XlJOqtmjDHGmKtMqoK3mD1s8ahqkzu0utZN8gP9Sa/VGH3w+gd4P+NlbulcFn93\nMSLC4KVBTjxxYvSDjTHGGGOSKFXBWwDIjkqLG9CFzXl7RkSqgd3A4lh5H3vssdDrsrIyysrKxlvX\nYWbfNpu+l/silvLoe7nPlvIwxhhjTFzt7e20t7cn/bypnPPWqKpLwtJOq2pejLzLAFXVvWFpTwHN\n0UOs6ZjzZvcENcYYY8yVyKg5b6raGX7/TXfZEH/UdsC9QCEALAP2Rp0j1ty4lLClPIwxxhgzWaRy\nqZAioJyhdd4aVbXf3dcC7FbV7e72AzjDqoKzzps/VvCWyqVCjDHGGGPGI6OWCpkoFrwZY4wxJlMk\nK3iLveS/McYYY4yZlCx4M8YYY0xMTU1NLFq0CI/Hg8fjITc3lyVLQtce4vP5Qvs8Hg+LFy/mxRdf\npL6+PpSWl5c3LE/wdW3t0Pr7fr8fn8/HkiVLWLx4MatXr6a7uxufz8fx48fp6+uLqIvH42HJkiX0\n9zsriXV2dpKbmxtR7qFDaZsuP7FUNWMfTvWNMcYYM5FKS0vV4/HooUOHhu3z+/0qIrpixYpQWl1d\nnXo8Hn3xxRdVVbWxsVFFRGtra1VVtaOjQz0ej1ZVVamq6p49e3TRokV6/PjxiPPm5OSox+PRzs7O\nUHplZaV6PB6tr6+PW9clS5aMv9ETwI1bxh3/WM+bMcYYY0aUm+sszer1ese0r7e3l7Vr13L//ffH\nzFNcXMy2bdvo7XVuvrRlyxZWrVrFggULQudYtmwZra2tAAQCgVD6pk2bUFUaGhpi1rW3t5empqYr\na2iGsODNGGOMmcROnjnJfT+8j7fPvp3uqoxZIBCgsrJyxDzl5eWhoKyrq4uOjo5heZYtW0ZJSeRt\nKIuLiykpKaGrq2vYsGh3dzc5OTncdddd42zB5GbBmzHGGDOJbXllC/v+sI8tr2xJd1WCU5ZGTauo\nqMDn8414roKCAlavXg1ASUkJe/bsYdGiRdTX19Pd3R3K19raOuxc8XrfGhoaqKmpGXN7MlXc4E1E\nikXkiIhsFZF74+RZJiL5E1U5Y4wx5mp28sxJnjv8HG0PtvHc4efS2vumqvh8PvLy8iIe5eXlhC/E\nD7BmzRrmzZs36jnXr18POBdGlJaW0t3dTV1dHQsXLiQ3N5d169ahqsPOtXLlSrKzs2lsbIxI37Fj\nB2vWrBlnSye/ke6w0IuzSG4esNO9UXwH0Iy7aK6qtonIBhE5qGG3szLGGGOMQx4f97JelDQ6Q4fz\nvz1/XOfRb1352qgiQkdHR8S8NHCu8iwtLR1XvbxeL/v372fv3r20trbi9/vp6uqisbGRxsZGOjo6\nKCoqijhm7dq11NfXs337dtasWcPOnTtZvnz5uOqRKeIGb6raLSJ1qroTQEQKgZVABbBFRHpwbnHl\nBxqIc+N4Y4wx5mp2pQHTyTMnueN7d/Da115j/tz5nDxzkju33slrX3uNG+fcmORajk2sIdLx6uvr\n46mnnuLpp59m6dKlLF26FIDjx4+zY8cO6urqeOqpp2hubo44rqamhi1bttDQ0MCaNWtoaGhgy5b0\nDy2nwohz3oKBm/u6S1XrVbVCVT1AFXAAqAR2xjtHkIgUiMh6d6h1vduTN1L+lSLygPu8cmzNMcYY\nY6aGLa9s4at3fZX5c53etvlz5/PgJx6cFHPfkikQCAwb/gTIz89n/fr1VFdXx7yYoaCggPLycjo6\nOmhra6Onp2dY79xUdcU3plfVNqANqB/jIa2q6gMQkQNAE04AOIyIbACOqeqLbpDnZwwBojHGGDNV\n7H9rP6/87hX+7t//LiL90zd/Ok01mhi5ubn09vZSX1/Phg0bhu3Pzs4edsVpUE1NDX6/n8rKSurr\nxxqOZL64wZuIFAMtOIFTi6r+NEaeZThB1vGRCnHPdTq4rap9IlIeJ68X2KSqucG8wJJYeY0xxpip\n6uW/fDndVQg5fdr5Ce/p6SE/Pz/mvuCabSMdP1KejRs3IiKhixjAWUKkqamJvXtjT6sPXrjQ19fH\nww8/PKa2TAUjDZtGX7AwICL73SHPIgj1vlWKyNJRyil0zxcuEDxPFB/Q5Q6ZBodYC8bWHGOMMcYk\nS/D2WMH11MrLy4fdHmvFihWICH6/P3R7rKDOzk4WLVpEbW0tIkJjY+OwPAALFy6kp6eH9957jxUr\nVrB69Wqqqqqora1l7969I67b9s1vfpO1a9cmueWTm4w0+VBEVsa4YGE5UA6EX7DwiKrGvWBBRKqB\nclVdHZZ2FFilqodi5N0G5Khqv9sTd1BVF8U4r07E5EljjDHGmGQTEVR13JcfjzjnLfqCBZz5bfVu\nBZYBJYztgoUAkB2VlhsnbxfQpar9brl9IlIoIvmxhmcfe+yx0OuysjLKyspGqYoxxhhjzMRrb2+n\nvb096ecdsectaYU4c94aVXVJWNppVc2LkbcA2B3ekyciAaAkOniznjdjjDHGZIpk9bwldHssEXkq\n/G4L7vIfW0VkwUjHqWpn1HkKcIZbw8/jdfN2A70iMs/dl80YLoowxhhjjLkaJNTzJiItOPPdgst3\n7AE6gafDe9XiHFvkHtsJBHvi+sPOu1tVt7vb+cA64BjOxQ4NsYI363kzxhhjTKZIVs9bosFbcAGW\nLpzlO1YBBTi3y1ox3sokyoI3Y4wxxmSKdAVv+6N72ERkLfCeqr4Y57AJY8GbMcYYYzJFWua8AT0i\n8o3wBFVtBDaNtyLGGGOMMWZ0ifa8ZeMMmZ4GdgD7cYZPa4J3REgl63kzxhhjTKZIS8+bqva6QVoT\nzmK9O3DWeascb0WMMcYYY8zoEu15K8C5y8KOybB0h/W8GWOMMSZTpGvOWw3wJfc5WJGtwTXZjDHG\nGGPMxEo0ePOqqk9Vwy9Q2IQzjGqMMcaYKSR4Y3qPx4PH4yE3N3fYjemD+zweT+im8/X19aG0vLy8\nYXmCr9etW0dFRUXE+VevXk1fX9+wcmtrawGoqqoKpWdlZbF3795Qffr6+li3bh0+nw+fz8fixYtp\namrC7/dTUVEx5nb39fVRV1dHbm7Kp/OPjaqO+YFzw/j7Y6TvH8OxBcB6YJn77B1rmSPsU2OMMcZM\nrNLSUvV4PHro0KFh+/x+v4qIrlixIpRWV1enHo9HX3zxRVVVbWxsVBHR2tpaVVXt6OhQj8ejVVVV\nqqpaU1OjHo9H6+vrI84dTN+5c2dE+o4dO9Tj8Wh/f38o7eDBg5qTk6ObNm0KpfX19WllZaWKiPp8\nvlHb2dHRoQsXLtTc3FwVEc3NzR31mES4cUtCsVesR6I9b3VAvYjsd2+VtUZEto7x2FZVfUZV23B6\n6kbtrRORZUB1gnU0xhhjTBIFe6C8Xu+Y9vX29rJ27Vruv//+mHmKi4vZtm0bvb29ANTU1KCqNDc3\nR5y7srIyZnogEGDt2rXMnTs3VJ7P5+Puu+/mySefDOWbN28eLS0tlJSUhMoaSXFxMUePHuX06dMU\nFhaOmj9dEr3atE9VFwGtOFebbgEWAlUjHefemP50+HlwbpU10jFeIACM/tc2xhhjzKQRCASorBx5\nIYry8nICgQDgBE2FhYV0dHTQ398fyrNnzx4AduzYEXFsa2trxPmrq6sREerq6mKWtWlT4svRZmdn\nJ3xMqiTa84aILMUZyvSpaq6qVqhzM/mRFDI8CAu49zuNZ5k6N7Qf91UZxhhjTCbKzweRyEd+fvrq\nozFWeIhnT69JAAAXhElEQVSVVlFRgc/nG/FcBQUFrF69OrS9atUqAFpaWkJpfr8/FKSFz207ePAg\nS5cuDW3v3LkTgNLS0phlrVy5clL3pCUqoeBNRHYDjUC3iMx1054aw9WmCc34c4dL/e6mrQVijDHm\nqnTiBKhGPk6cSE9dVBWfz0deXl7Eo7y8HJHIfpY1a9Ywb97oC1GsX78+9Hr16tWoKq2trYBz0UBe\nXh6bNm1CVWloaACgra2N8vKhwbvu7qH+o5HK3LVr19gamgES7XnrcodNg0OmAJvdx0gCQHT/Y8yA\nzl1LLqCqwX5T63kzxhiTsaJ7zhJ5xDp+POccXzuEjo4OTp8+HfFoa2uL2fuWqODQqd/vp7+/n5aW\nFpYvX05RURHZ2dmhodPW1taIHrsr0dfXF/HINNMSPUBE7lLVDhHpAOeuCyIyWl9kF8ODNVXVQzHy\nlgA5IlKKE7hli8gawK8xFgZ+7LHHQq/LysooKysba1OMMcaYCTeeuEZk+PGx0lIlGUHaSFatWkV9\nfT3Nzc20trbS2NgIOMuDNDU10dbWRltbG9u2bQsdU1BQEHrd398ft/etu7ubgoIC2traWL58eSg9\nJyeH06dPxzxmvNrb22lvb0/6eRMN3jYCB0TkINDlBlXZOHPa4lLVzvAuVbd3zR+1HXAviNgZfqyI\nNKjq9njnDg/ejDHGGJO5ampq2LJlCw0NDYgI+e4Ev5qaGhobG6mrq4s5d62kpITOzk4OHDgQMRcu\n3PLlyzl69Cg+nw+/PxSCTOhabtGdSo8//nhSzptQ8Ob2spXiDJnW4ARuPYzt3qbVIrIe6ASKiVwC\nZDOwGwgFae7VpmsBdY+bFLfkMsYYY1JlwYLhw50LFqSnLqlQUFBAYWEhnZ2dPPLII6H04uJisrOz\n6ezsDPXGhWtqaqK0tJSGhoaYwZvf7w9dzOD1euMGeJki0QsWWnCGL2vcK009qpqnqntHO1ZVDwXX\neXOf+8P2VUX3rrm9cPWqmuXmP55IXY0xxphMd/z48AsWjh9PfT2Cw4o9PT1x9420jtpY8gTV1Dh3\n4Iye11ZV5axKFmsJkuLiYhoaGtixY8ewZUF27NhBbW0tTU2J3Qyqp6dnTPVNh0QvWCgBGiaiIsYY\nY4yZXIK3xzp0yJmiXl5ePuz2WCtWrEBE8Pv9odtjBXV2drJo0SJqa2sRERobG4flibZq1Sqys7Mp\nKopcTaympoaSkpK4c9qqq6s5duwYXV1dLFmyJPRoa2vj4MGDY7r6ta2tDZ/Px6JFizjuRsl5eXks\nWbIk9DeYDCSRyYfuEh7FqvrMxFVp7EREJ3rypDHGGGNMMogIqjruVTQSDd4CDK279iTQFD78mWoW\nvBljjDEmUyQreEt4nTecK073Ao8CPSKyy72gwBhjjDHGTLBEe96K3VtWBbdLcO5RWq6qFRNQv9Hq\nYz1vxhhjjMkIaRk2nWwseDPGGGNMpkjXsKkxxhhjjEkjC96MMcYYYzKIBW/GGGOMMRnEgjdjjDHG\nmAyS6I3pr5h78/mVDN3btElV++LkLQZ8OPdO9QEbVbU7VXU1xhhjjJmsUha8Aa2q6gMQkQNAE1AV\nncm9Ib1PVZvc7WXAHmBRCutqjDHGGDMppWTY1O1JOx3cdnvcyuNkLwQeCds+ABSKyOg3JTPGGGOM\nmeJSNeetEOiNSguISFF0RncR4OVhSUuAnnTehssYY4y5GgVvTO/xePB4POTm5g67MX1wn8fjCd10\nvr6+PpSWl5c3LE/w9bp166ioqIg4/+rVq+nr6xtWbm1tLQBVVVWh9KysLPbu3RuqT19fH+vWrcPn\n8+Hz+Vi8eDFNTU34/X4qKsZ2L4GmpiYqKiqoqKhg0aJFbNy4Mbl/1GRQ1Ql/ANVAc1TaUaBoDMe2\nAPfH2afGGGOMmVilpaXq8Xj00KFDw/b5/X4VEV2xYkUora6uTj0ej7744ouqqtrY2KgiorW1taqq\n2tHRoR6PR6uqqlRVtaamRj0ej9bX10ecO5i+c+fOiPQdO3aox+PR/v7+UNrBgwc1JydHN23aFErr\n6+vTyspKFRH1+XyjtvORRx7RjRs3hra7u7s1JydHKyoqRj12LNy4ZdxxVarmvAVwLj4IlzvaQSJS\nDbygqj+Kl+exxx4LvS4rK6OsrOzKamiMMcaYmHJznZ9sr9c7pn29vb2sXbuW+++/P2ae4uJitm3b\nxo4dOwCoqamhsbGR5uZm1q8ful16ZWVlKP2BBx4IpQcCAdauXcvcuXND5fl8PioqKnjyySdD+ebN\nm0dLSws+n4/e3ugBwOHq6+vp6uoKbefn51NVVUVTUxOHDh2iqGjYgOGI2tvbaW9vT+iYsUhV8NbF\n8GBNVfVQvAPcCxWOqereeHkgMngzxhhjppJTz5/ixBMn+OD1D5h922wWPLqAG758Q7qrNapAIMC6\ndetGzFNeXk5jYyPgBHOFhYV0dHTQ39/PvHnONPc9e/YAhIK8oNbW1ojhzOrqakSEurq6mGVt2rRp\n1OHPvj5nAYyFCxfS09MTqkNQV1dXwsFbdKfS448/ntDx8aRkzpuG3cweQsuG+MO33atMg9vFQCAY\nuInIylTU0xhjjJksTj1/iu5Hu1n83cX86fk/ZfF3F9P9aDennj+VtjppjPuJx0qrqKjA5/ONeK6C\nggJWr14d2l61ahUALS0toTS/309lZSVAxNy2gwcPsnTp0tD2zp07ASgtLY1Z1sqVKyksLByxPl6v\nl1WrVpGTkxOR3tzcDEBJScmIx6dSKhfprRaR9W6P2kqceXBBm4FKCAV2B4EDIjIgIoPA0ymspzHG\nGJN2J544wa3P3krOvTl4pnvIuTeHW5+9lRNPnEhLfVQVn89HXl5exKO8vByRyHutr1mzZljPVSzh\nQ6SrV69GVWltbQWcnrC8vDw2bdqEqtLQ0ABAW1sb5eVDC1Z0dw8tAztSmbt27Rq1Pi0tLbz33nuh\n8zQ2NtLX10dlZSX5+fmjHp8qKVvnzR0iDQ6TtkXtqwp73Y3d+cEYY8wU0S7tV3zs4aWHk3rOMi27\n4rqICB0dHSxYsCAivbOzM26PVyKCQ6d+v5/+/n5aWlpYvnw5RUVFZGdnh4ZOW1tbI3rsrkRwiDQo\n1ly+jo4O1q1bR1VVFS+88MK4yku2VC7Sa4wxxlx1rjRg2nfnPhZ/dzE59w4N4/X8tIcjf32Eu399\nd5Jql5hYQ6TJtGrVKurr62lubqa1tTU0Jy540UBbWxttbW1s27YtdExBQUHodfh8uWjd3d0UFBTQ\n1tbG8uVDK5Ll5ORw+vTpiLy9vb2Ul5ezbt06vve97yWziUlhPVzGGGPMJLTg0QW88fAb9Py0h8FL\ng/T8tIc3Hn6DBY8uGP3gDFVTUxMaIu3p6QkNVQbT6+rqYs5dC85HO3DgQNxzBwM2n8+H3+8PPdra\nIgYDQ1euPvroo6HArbOzc1i+dLKeN2OMMWYSCl5VeuSvj4SuNi14oiAjrja9UgUFBRQWFtLZ2ckj\njwzdbKm4uJjs7Gw6OztDvXHhmpqaKC0tpaGhIeJChiC/3x8a2vV6vTHzBJWXl7Nx40bWrFkTSmto\naBjzIr+pYD1vxhhjzCR1w5dv4O5f303ZQBl3//rutAVuwWHFnp6euPtGWkdtLHmCampqAIbNa6uq\ncqbHB68+DVdcXExDQwM7duxg06ZNEft27NhBbW0tTU1NYyq7t7eXAwcOsG7dOtatWxdaay47O3q5\n2vSRiR6/nkgioplcf2OMMWYya2pqYvPmzaErOrOzsyksLGT//v2AMwTZ2Tm0GlhhYSGbN28OLajb\n2dlJZWVlxBWh0XmidXd34/P5hs1D6+zsZO3ataGyYzl+/Dh1dXURC+36fD42b9486tWvO3fuDAWI\n0USEo0ePjvuKUxFBVWX0nKOcJ5ODHwvejDHGGJMpkhW82bCpMcYYY0wGseDNGGOMMSaDWPBmjDHG\nGJNBUrZUiHvbq5VAJ1AMNKlq33jzGmOMMcZcTVJ2wYKIHFBVn/vaixOQxbysY6x57YIFY4wxxmSK\njLpgQUSKgdA1v24vWvl48xpjjDHGXG1SNeetEIhemS8gIkXjzGuMMcYYc1VJ1Zy33AnKy2OPPRZ6\nXVZWRllZWSKHG2OMMcZMiPb2dtrb25N+3pTMeRORlcBaVV0RlhYAlqrqoXHktTlvxhhjjMkIGTXn\nDehieI+aRgdjV5DXGGOMMeaqkpLgTVU7w7fdpUD84dvuVaWj5jXGGGOMuZqlcqmQIpyrRoNrtzWq\nar+7rwXYrarbR8sbdU4bNjXGGGNMRrAb02PBmzHGGGMyR6bNectoE3GlSLpNxTbB1GyXtSkzWJsy\ng7Upc0zVdiWDBW9jMBU/QFOxTTA122VtygzWpsxgbcocU7VdyWDBmzHGGGNMBrHgzRhjjDEmg2T8\nBQvproMxxhhjzFhd9VebGmOMMcZcbWzY1BhjjDEmg1jwZowxxhiTQaaluwLGjMS9PdpKhu620aSq\nfePJKyLbVHXdxNXaTBWJfP7CjtmtqhXjPc9ESWKbigEfkO0+b1TV7omp9ciS1aao/Wn9nkhmm0Rk\nJaCAAKjqzuTXeHQT8NkD8AI70/XZSxtVvSofQAGwHljmPnuvJG8i58mgNhUD1cAGoBkoSGObDoS9\n9gIt48nrtndgqnz+3P0rgQfc55WZ3qawz1+1uy9TPn/L3DoP+3wlcp5MaJN7XHVUvqOZ3KYYedL6\nPZHEz94G4IGw8+yfCm2K2t6W5vcqke++uL+vCZ0nnQ1O8x87KUFBBn8px8w7mb6U3Q/5rqi0wJXm\nddtWDJxO13uUzPfK3c7UL+YR2xSVNy1fzIl8/qLyRAc6V3SeDGjTkaj3cBCYl6ltimpLWr8nkvg+\nedP1WZvI9wk4SuR/+LamuW1j+u4b7fc1ke/Qq3LOm9vlejq4rU63bXkCeZclep6JloQ2BfMWAo+E\nZT8AFIrIvGTXeQwKgd6otICIFF1h3mWq2ok7dJAuyXqvRCQb2KSqLwb3qeqSCax6XEn8/AHUiIg3\nbDtdl8Qn8vlLxXmSISl1cf8dLQ9LWgL0qGr/OOt3JZL9950M3xPJapMP6BKRB0RkmYisd4cu0yGZ\n79NmoFtEqkWkGqgbd+2uUIJxQNzf10TjiasyeCN5QUGmfinHzTvJvpRzk5VXRJYBfncz3evjJOvz\nV0pmfjGPlneyfDEn8vlLxXmSIWl1UdXjYZtrgTXJOneCktamSfQ9kaw2FeL0ePlVtQ1oAvYk6dyJ\nSuZnrwl4EudztxbIS9a5r8CYv/tG+X1NKJ64Wi9YGG9QICPsS5ekBTqT6Es5gDMZOly8usfN6wY0\ngbAANK09byTvvQr/Yu4XkQPAQWDReCp3hZL5+Wtye97Wukl+IB3/eUjk85eK8yRD0uviBtgvqOqP\nxnOecUhKmybZ90Sy3qcuoCvYJlXtE5FCEcmP+p5PhaR99kRkg6rWA8+4n7/dwOJx1u9KJdSGEX5f\nEzrP1Rq8JSUoSPA8Ey1ZbQqZBF/KXQyvl6rqoUTyulda5YhIKc4XcraIrMEJeo4nu9JjkKz3KlO/\nmEfMO4m+mBP5/KXiPMmQ1Lq4PVXHVHXvuGt25ZLVphImz/dEMj970aJ7d1IlKW1yP3MHw07Q5H7v\nFaXp39QVxQExfl8TOs/VGrwlKyiQBM4z0ZLSpuDGZPhSVtVO508cqlMBQ0Ma4f9T7hspr0ZdFi8i\nDaq6fYKrP5Jkff5iDZFmwhfzSG2aNF/MiXz+og6N6LEZ7TyplKw2uXmL3byd7vbK6H9rqZDE92nS\nfE8ksU3dItIrIvPc3vlsnO/14xNX+9iS+NkL4Mw7j/htStPvLlxBUBrn9zWh81yVc96CXzZBsT5E\nwQnTI+Ud7TyplKw2udvBL+W97vbKCaz6aKrduVzLcJbCqA7btxmoHGNeRMQrIhsAdfPlT2zVY0vi\n568b6BX3YpJ0fzHHq2dwe4yfvwBOD0j0+dP1xTzmz5+IFId9vp4SkaVjPE+qjbtN7nt2EDggIgMi\nMgg8nbomDJOs92nSfE+QvDZVAt+Uofmj4d+ZqTbuNrnfH10iskacebHrcZbcSItEvvvc7Zi/r4nG\nE1ftvU3FmQRYztBigY3B4ScRaQF2B//XNUreuPtSLRltcj8wxxiarCs4AUG65hNMSUn8/OUD63De\ns0KgIU1Dwcls0wM4/wMVnMvl/WkM3owxZkRj/e4b7fc1kXjiqg3ejDHGGGMy0VU5bGqMMcYYk6ks\neDPGGGOMySAWvBljjDHGZBAL3owxxhhjMogFb8YYY4wxGcSCN2OMMcaYDGLBmzHG4Kx6LiIHRGTQ\nfewXkd0iEnC3t8bIs8vd3u2u4YSIrIyRZ7eIHB3rgtfuIqZjyVcQLNcYc/Ww4M0YM6HcRSonfR1U\ntU1Vfe7mNlVdoqoVqpqLs7Dmsag8T6vqCne7F+cerMHbLAVXit/g5qnAWUG+NXpF/xh1fZqhRTxH\nq3M3sNldtd0Yc5Ww4M0YM2FEpBDnLgkZUQcRCd6aK/q2NG1Axwh59gCFwVuVAaU4AVhbWJ5g/mG3\n/worvxwoiHdPYbfnb5l7W6BgL1410Bq/VcaYqcaCN2PMRGpIdwVIrA7lOEHXHgAR2eamH8W5cXQo\nT1SA5cNJDN7KZrm7HX5br2BAd2yE8re5j2HEufF7m/toAmrcMvqAgyKyZkwtNMZkPAvejDETwr2p\ndAHgE5Fm9wbSiEi5iLS4j6NuvuAxT7tzzHa5PUy7g0OeIlLibm91n592j1/qzv3a7Z5zV/Bm4vHq\nMIJyoEtVz4jIKpyAC1V9JuyeseW4vXDB9gBrgLVh51lG5M2pi4FGYI+q/ijO36vArevBGPu8hAV9\n7k3Gw4eCD+AGc8aYqc/ubWqMmTBu4OVV1RVhaQGcuWDPht2ouTAYHLm9XaU4PWY5QJ2qXuce97Cq\n/sgdMmwCVgLHcXrKnnLPuRLYrKqL4tVhhPoOAj04N4zOds+zKUaeYzgBXI6bf5uq/tTd73XTjgHd\nOL1yLUBLvOFQ97hVQLOqZsXYt1JVd7rz4cDp+dsUtr/arcOwY40xU8+0dFfAGHPVeZKh3iXBGUrM\nDtvfizMvrMUdhqx3gzwv0Onm6XK3DwIVOD1WwXlfbTjzz4qihi1HFDbp/xE3CAzgDp+G5QnOV1sb\nDNZiWO62aZWqHnavHG0FHhmlCiNdNaoAqrrRrcfWqP0BN31e2NCtMWaKsuDNGJNqTThXSCrQ56bl\nRuXpCg9CVLVbRLpwhiy34wRIx1S13w3sBGgSEdzXx4BCYMzBG0NBV/Aigz04w5HhgvPd4gVuwfOg\nqofd5zYRyXaPfXGE47pG2NcbtS2xglML3Iy5OljwZoyZSAHcKz3d+We7cXrPlqnqT93AK1aPVHSw\nAs6csToRWehul4blVaA6TvASXYfWsPlr4YLz3YL76mKcr5yRgyxw5ruFeuzcYVRwegO97gUGscQ8\nr9tztz8q2ReVv5CweXjGmKnNLlgwxkykXpzAApweqUU4gVa3m1bD8GFTcIcBg9whzdVufj/OHLLg\n91eLW05NWP5tIlIUow7lcQI3iLoQYYQAL26Q5PawFTI0vBvuOmBT2HIiEVS1E+iKsQ5cxNIibjD3\nQlRguQRojlcvY8zUYsGbMWYiNQCIyGlgq7uAbRvO0ha7gH04PWqbRaTInXi/FiiPmtfVhRPg7cHp\nvTsIBETkAbcnaxmwWkSOiMh+YHfYkGJ4HYYtw+HeEeEoThBZ7tYrOs+ysDyr4uSpduupwFoRaYbQ\nUh6NwAbgyChDmxuBdVFpHThXyy5zy/Cq6jNh5WYDxeFpxpipza42NcZMeu5yHBtwLgI446Y9jbOg\n7eq0Vi7J3KDvKVU95A65Fox04YV7dW7zKPPwjDFTiPW8GWMywSrc9dfC0k4zfC5YxnOD0eAwr2+U\nwK2AsGVKjDFXB+t5M8ZMesEgxd3sxRlC7Yheg22qEZH8EeboGWOuUha8GWOMMcZkEBs2NcYYY4zJ\nIBa8GWOMMcZkEAvejDHGGGMyiAVvxhhjjDEZxII3Y4wxxpgMYsGbMcYYY0wG+f8B6kXfnHlzp3oA\nAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1dddf80d10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import pylab\n",
    "from pylab import *\n",
    "%matplotlib inline  \n",
    "import itertools \n",
    "\n",
    "\n",
    "plt.figure(figsize=(10, 2))\n",
    "\n",
    "font = {\"family\" : \"normal\",\n",
    "        \"weight\" : \"normal\",\n",
    "        \"size\"   : 14}\n",
    "\n",
    "matplotlib.rc('font', **font)\n",
    "\n",
    "from matplotlib import rc\n",
    "# rc('font',**{'family':'sans-serif','sans-serif':['Helvetica']})\n",
    "# ## for Palatino and other serif fonts use:\n",
    "# #rc('font',**{'family':'serif','serif':['Palatino']})\n",
    "rc('text', usetex=True)\n",
    "\n",
    "roc_sv_pr, = plt.plot([beta for beta in beta_list], \\\n",
    "                   [roc_data_sv[beta][\"Recall\"] for beta in beta_list], \"g-^\", linewidth=1, \\\n",
    "                      markerfacecolor=\"None\", markeredgecolor=\"g\", markeredgewidth=1)\n",
    "\n",
    "roc_wc_1_pr, = plt.plot([beta for beta in beta_list], \\\n",
    "                     [roc_data_wc_1[beta][\"Recall\"] for beta in beta_list], \"b-s\", linewidth=1, markersize=5, \\\n",
    "                        markerfacecolor=\"None\", markeredgecolor=\"b\", markeredgewidth=1)\n",
    "\n",
    "roc_wc_2_pr, = plt.plot([beta for beta in beta_list], \\\n",
    "                     [roc_data_wc_2[beta][\"Recall\"] for beta in beta_list], \"m-o\", linewidth=1, \\\n",
    "                        markerfacecolor=\"None\", markeredgecolor=\"m\", markeredgewidth=1)\n",
    "\n",
    "legend_list = [r\"HTSV\", r\"HTWC-1\", r\"HTWC-2\"]\n",
    "\n",
    "# legend(loc=0, fancybox=\"true\", ncol=2, mode=\"expand\", borderaxespad=0.2) \n",
    "\n",
    "plt.legend([roc_sv_pr, roc_wc_1_pr, roc_wc_2_pr], \\\n",
    "           legend_list, loc=4, numpoints=1, fancybox=\"false\", frameon=False) \n",
    "plt.title(r\"$N = %s, n = %s$\"%(N, n))\n",
    "plt.xticks(np.arange(-0.02, 0.24, 0.02)[1:-1])\n",
    "\n",
    "pylab.xlim(-0.01, 0.2)\n",
    "pylab.ylim(-0.1, 1.1)\n",
    "\n",
    "plt.xlabel(r\"target \\textit{FPR} ($\\beta$)\")\n",
    "plt.ylabel(r\"\\textit{recall}\")\n",
    "# plt.grid()\n",
    "\n",
    "\n",
    "savefig(data_dir + \"ROC_data_recall_N_%s_n_%s.eps\"%(N, n), dpi=300, bbox_inches=\"tight\")\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAmMAAACyCAYAAAAONcjeAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1de40fcc90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import pylab\n",
    "from pylab import *\n",
    "%matplotlib inline  \n",
    "import itertools \n",
    "\n",
    "\n",
    "plt.figure(figsize=(10, 2))\n",
    "\n",
    "font = {\"family\" : \"normal\",\n",
    "        \"weight\" : \"normal\",\n",
    "        \"size\"   : 14}\n",
    "\n",
    "matplotlib.rc('font', **font)\n",
    "\n",
    "from matplotlib import rc\n",
    "# rc('font',**{'family':'sans-serif','sans-serif':['Helvetica']})\n",
    "# ## for Palatino and other serif fonts use:\n",
    "# #rc('font',**{'family':'serif','serif':['Palatino']})\n",
    "rc('text', usetex=True)\n",
    "\n",
    "roc_sv_re, = plt.plot([roc_data_sv[beta][\"FPR\"] for beta in beta_list], \\\n",
    "                   [roc_data_sv[beta][\"Recall\"] for beta in beta_list], \"g-^\", linewidth=1, \\\n",
    "                      markerfacecolor=\"None\", markeredgecolor=\"g\", markeredgewidth=1)\n",
    "\n",
    "roc_wc_1_re, = plt.plot([roc_data_wc_1[beta][\"FPR\"] for beta in beta_list], \\\n",
    "                     [roc_data_wc_1[beta][\"Recall\"] for beta in beta_list], \"b-s\", linewidth=1, markersize=5, \\\n",
    "                        markerfacecolor=\"None\", markeredgecolor=\"b\", markeredgewidth=1)\n",
    "\n",
    "roc_wc_2_re, = plt.plot([roc_data_wc_2[beta][\"FPR\"] for beta in beta_list], \\\n",
    "                     [roc_data_wc_2[beta][\"Recall\"] for beta in beta_list], \"m-o\", linewidth=1, \\\n",
    "                        markerfacecolor=\"None\", markeredgecolor=\"m\", markeredgewidth=1)\n",
    "\n",
    "legend_list = [\"HTSV\", \"HTWC-1\", \"HTWC-2\"]\n",
    "\n",
    "# legend(loc=0, fancybox=\"true\", ncol=2, mode=\"expand\", borderaxespad=0.2) \n",
    "\n",
    "plt.legend([roc_sv_re, roc_wc_1_re, roc_wc_2_re], \\\n",
    "           legend_list, loc=4, numpoints=1, fancybox=\"false\", frameon=False) \n",
    "plt.title(r\"$N = %s, n = %s$\"%(N, n))\n",
    "# plt.xticks(N_list)\n",
    "\n",
    "pylab.xlim(-0.02, 1.02)\n",
    "pylab.ylim(-0.1, 1.1)\n",
    "\n",
    "plt.xlabel(r\"\\textit{FPR}\")\n",
    "plt.ylabel(r\"\\textit{TPR}\")\n",
    "# plt.grid()\n",
    "\n",
    "savefig(data_dir + \"ROC_graph_N_%s_n_%s.eps\"%(N, n), dpi=300, bbox_inches=\"tight\")\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "roc_data_all = {\"wc_1\": roc_data_wc_1, \"wc_2\": roc_data_wc_2, \"sv\": roc_data_sv}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'sv': {0.001: {'FPR': 0.402,\n",
       "   'Precision': 0.7130620985010707,\n",
       "   'Recall': 0.999,\n",
       "   'TPR': 0.999},\n",
       "  0.01: {'FPR': 0.752,\n",
       "   'Precision': 0.5707762557077626,\n",
       "   'Recall': 1.0,\n",
       "   'TPR': 1.0},\n",
       "  0.02: {'FPR': 0.844,\n",
       "   'Precision': 0.5422993492407809,\n",
       "   'Recall': 1.0,\n",
       "   'TPR': 1.0},\n",
       "  0.029999999999999999: {'FPR': 0.898,\n",
       "   'Precision': 0.5268703898840885,\n",
       "   'Recall': 1.0,\n",
       "   'TPR': 1.0},\n",
       "  0.040000000000000001: {'FPR': 0.927,\n",
       "   'Precision': 0.5189413596263622,\n",
       "   'Recall': 1.0,\n",
       "   'TPR': 1.0},\n",
       "  0.050000000000000003: {'FPR': 0.945,\n",
       "   'Precision': 0.5141388174807198,\n",
       "   'Recall': 1.0,\n",
       "   'TPR': 1.0},\n",
       "  0.059999999999999998: {'FPR': 0.953,\n",
       "   'Precision': 0.5120327700972862,\n",
       "   'Recall': 1.0,\n",
       "   'TPR': 1.0},\n",
       "  0.070000000000000007: {'FPR': 0.965,\n",
       "   'Precision': 0.5089058524173028,\n",
       "   'Recall': 1.0,\n",
       "   'TPR': 1.0},\n",
       "  0.080000000000000002: {'FPR': 0.971,\n",
       "   'Precision': 0.5073566717402334,\n",
       "   'Recall': 1.0,\n",
       "   'TPR': 1.0},\n",
       "  0.089999999999999997: {'FPR': 0.974,\n",
       "   'Precision': 0.5065856129685917,\n",
       "   'Recall': 1.0,\n",
       "   'TPR': 1.0},\n",
       "  0.10000000000000001: {'FPR': 0.979,\n",
       "   'Precision': 0.5053057099545225,\n",
       "   'Recall': 1.0,\n",
       "   'TPR': 1.0},\n",
       "  0.11: {'FPR': 0.985,\n",
       "   'Precision': 0.5037783375314862,\n",
       "   'Recall': 1.0,\n",
       "   'TPR': 1.0},\n",
       "  0.12: {'FPR': 0.985,\n",
       "   'Precision': 0.5037783375314862,\n",
       "   'Recall': 1.0,\n",
       "   'TPR': 1.0},\n",
       "  0.13: {'FPR': 0.985,\n",
       "   'Precision': 0.5037783375314862,\n",
       "   'Recall': 1.0,\n",
       "   'TPR': 1.0},\n",
       "  0.14000000000000001: {'FPR': 0.988,\n",
       "   'Precision': 0.5030181086519114,\n",
       "   'Recall': 1.0,\n",
       "   'TPR': 1.0},\n",
       "  0.14999999999999999: {'FPR': 0.99,\n",
       "   'Precision': 0.5025125628140703,\n",
       "   'Recall': 1.0,\n",
       "   'TPR': 1.0},\n",
       "  0.16: {'FPR': 0.993,\n",
       "   'Precision': 0.5017561465127948,\n",
       "   'Recall': 1.0,\n",
       "   'TPR': 1.0},\n",
       "  0.17000000000000001: {'FPR': 0.994,\n",
       "   'Precision': 0.5015045135406219,\n",
       "   'Recall': 1.0,\n",
       "   'TPR': 1.0},\n",
       "  0.17999999999999999: {'FPR': 0.995,\n",
       "   'Precision': 0.5012531328320802,\n",
       "   'Recall': 1.0,\n",
       "   'TPR': 1.0},\n",
       "  0.19: {'FPR': 0.997,\n",
       "   'Precision': 0.5007511266900351,\n",
       "   'Recall': 1.0,\n",
       "   'TPR': 1.0}},\n",
       " 'wc_1': {0.001: {'FPR': 0.002,\n",
       "   'Precision': 0.9977452085682075,\n",
       "   'Recall': 0.885,\n",
       "   'TPR': 0.885},\n",
       "  0.01: {'FPR': 0.011,\n",
       "   'Precision': 0.9887295081967213,\n",
       "   'Recall': 0.965,\n",
       "   'TPR': 0.965},\n",
       "  0.02: {'FPR': 0.033,\n",
       "   'Precision': 0.96751968503937,\n",
       "   'Recall': 0.983,\n",
       "   'TPR': 0.983},\n",
       "  0.029999999999999999: {'FPR': 0.062,\n",
       "   'Precision': 0.94106463878327,\n",
       "   'Recall': 0.99,\n",
       "   'TPR': 0.99},\n",
       "  0.040000000000000001: {'FPR': 0.066,\n",
       "   'Precision': 0.9375,\n",
       "   'Recall': 0.99,\n",
       "   'TPR': 0.99},\n",
       "  0.050000000000000003: {'FPR': 0.076,\n",
       "   'Precision': 0.9287722586691659,\n",
       "   'Recall': 0.991,\n",
       "   'TPR': 0.991},\n",
       "  0.059999999999999998: {'FPR': 0.088,\n",
       "   'Precision': 0.9184430027803522,\n",
       "   'Recall': 0.991,\n",
       "   'TPR': 0.991},\n",
       "  0.070000000000000007: {'FPR': 0.098,\n",
       "   'Precision': 0.9101741521539871,\n",
       "   'Recall': 0.993,\n",
       "   'TPR': 0.993},\n",
       "  0.080000000000000002: {'FPR': 0.114,\n",
       "   'Precision': 0.8971119133574007,\n",
       "   'Recall': 0.994,\n",
       "   'TPR': 0.994},\n",
       "  0.089999999999999997: {'FPR': 0.134,\n",
       "   'Precision': 0.8812056737588653,\n",
       "   'Recall': 0.994,\n",
       "   'TPR': 0.994},\n",
       "  0.10000000000000001: {'FPR': 0.139,\n",
       "   'Precision': 0.8773168578993822,\n",
       "   'Recall': 0.994,\n",
       "   'TPR': 0.994},\n",
       "  0.11: {'FPR': 0.149,\n",
       "   'Precision': 0.8697552447552448,\n",
       "   'Recall': 0.995,\n",
       "   'TPR': 0.995},\n",
       "  0.12: {'FPR': 0.164,\n",
       "   'Precision': 0.8586206896551725,\n",
       "   'Recall': 0.996,\n",
       "   'TPR': 0.996},\n",
       "  0.13: {'FPR': 0.179,\n",
       "   'Precision': 0.8476595744680852,\n",
       "   'Recall': 0.996,\n",
       "   'TPR': 0.996},\n",
       "  0.14000000000000001: {'FPR': 0.193,\n",
       "   'Precision': 0.8376787216148024,\n",
       "   'Recall': 0.996,\n",
       "   'TPR': 0.996},\n",
       "  0.14999999999999999: {'FPR': 0.207,\n",
       "   'Precision': 0.827930174563591,\n",
       "   'Recall': 0.996,\n",
       "   'TPR': 0.996},\n",
       "  0.16: {'FPR': 0.228,\n",
       "   'Precision': 0.8137254901960784,\n",
       "   'Recall': 0.996,\n",
       "   'TPR': 0.996},\n",
       "  0.17000000000000001: {'FPR': 0.24,\n",
       "   'Precision': 0.8058252427184466,\n",
       "   'Recall': 0.996,\n",
       "   'TPR': 0.996},\n",
       "  0.17999999999999999: {'FPR': 0.254,\n",
       "   'Precision': 0.7968,\n",
       "   'Recall': 0.996,\n",
       "   'TPR': 0.996},\n",
       "  0.19: {'FPR': 0.263,\n",
       "   'Precision': 0.7911040508339953,\n",
       "   'Recall': 0.996,\n",
       "   'TPR': 0.996}},\n",
       " 'wc_2': {0.001: {'FPR': 0.0, 'Precision': 1.0, 'Recall': 0.816, 'TPR': 0.816},\n",
       "  0.01: {'FPR': 0.002,\n",
       "   'Precision': 0.9977528089887641,\n",
       "   'Recall': 0.888,\n",
       "   'TPR': 0.888},\n",
       "  0.02: {'FPR': 0.003,\n",
       "   'Precision': 0.9968287526427061,\n",
       "   'Recall': 0.943,\n",
       "   'TPR': 0.943},\n",
       "  0.029999999999999999: {'FPR': 0.01,\n",
       "   'Precision': 0.9896907216494846,\n",
       "   'Recall': 0.96,\n",
       "   'TPR': 0.96},\n",
       "  0.040000000000000001: {'FPR': 0.018,\n",
       "   'Precision': 0.9817997977755308,\n",
       "   'Recall': 0.971,\n",
       "   'TPR': 0.971},\n",
       "  0.050000000000000003: {'FPR': 0.029,\n",
       "   'Precision': 0.9712871287128713,\n",
       "   'Recall': 0.981,\n",
       "   'TPR': 0.981},\n",
       "  0.059999999999999998: {'FPR': 0.037,\n",
       "   'Precision': 0.9637610186092067,\n",
       "   'Recall': 0.984,\n",
       "   'TPR': 0.984},\n",
       "  0.070000000000000007: {'FPR': 0.045,\n",
       "   'Precision': 0.95635305528613,\n",
       "   'Recall': 0.986,\n",
       "   'TPR': 0.986},\n",
       "  0.080000000000000002: {'FPR': 0.055,\n",
       "   'Precision': 0.947267497603068,\n",
       "   'Recall': 0.988,\n",
       "   'TPR': 0.988},\n",
       "  0.089999999999999997: {'FPR': 0.064,\n",
       "   'Precision': 0.9392789373814042,\n",
       "   'Recall': 0.99,\n",
       "   'TPR': 0.99},\n",
       "  0.10000000000000001: {'FPR': 0.066,\n",
       "   'Precision': 0.9375,\n",
       "   'Recall': 0.99,\n",
       "   'TPR': 0.99},\n",
       "  0.11: {'FPR': 0.079,\n",
       "   'Precision': 0.9261682242990654,\n",
       "   'Recall': 0.991,\n",
       "   'TPR': 0.991},\n",
       "  0.12: {'FPR': 0.095,\n",
       "   'Precision': 0.9126034958601656,\n",
       "   'Recall': 0.992,\n",
       "   'TPR': 0.992},\n",
       "  0.13: {'FPR': 0.107,\n",
       "   'Precision': 0.9027272727272727,\n",
       "   'Recall': 0.993,\n",
       "   'TPR': 0.993},\n",
       "  0.14000000000000001: {'FPR': 0.117,\n",
       "   'Precision': 0.8946894689468947,\n",
       "   'Recall': 0.994,\n",
       "   'TPR': 0.994},\n",
       "  0.14999999999999999: {'FPR': 0.134,\n",
       "   'Precision': 0.8812056737588653,\n",
       "   'Recall': 0.994,\n",
       "   'TPR': 0.994},\n",
       "  0.16: {'FPR': 0.139,\n",
       "   'Precision': 0.8773168578993822,\n",
       "   'Recall': 0.994,\n",
       "   'TPR': 0.994},\n",
       "  0.17000000000000001: {'FPR': 0.148,\n",
       "   'Precision': 0.8705161854768154,\n",
       "   'Recall': 0.995,\n",
       "   'TPR': 0.995},\n",
       "  0.17999999999999999: {'FPR': 0.159,\n",
       "   'Precision': 0.8623376623376623,\n",
       "   'Recall': 0.996,\n",
       "   'TPR': 0.996},\n",
       "  0.19: {'FPR': 0.177,\n",
       "   'Precision': 0.8491048593350383,\n",
       "   'Recall': 0.996,\n",
       "   'TPR': 0.996}}}"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "roc_data_all"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [default]",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}