3-8-13-Autoencoders.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xfNT-mlFwxVM"
      },
      "source": [
        "# Autoencoders for Anomaly Detection"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ITZuApL56Mny"
      },
      "source": [
        "In this example, you will train an autoencoder to detect anomalies on the [ECG5000 dataset](http://www.timeseriesclassification.com/description.php?Dataset=ECG5000). This dataset contains 5,000 [Electrocardiograms](https://en.wikipedia.org/wiki/Electrocardiography), each with 140 data points. You will use a simplified version of the dataset, where each example has been labeled either `0` (corresponding to an abnormal rhythm), or `1` (corresponding to a normal rhythm). You are interested in identifying the abnormal rhythms.\n",
        "\n",
        "Note: This is a labeled dataset, so you could phrase this as a supervised learning problem. The goal of this example is to illustrate anomaly detection concepts you can apply to larger datasets, where you do not have labels available (for example, if you had many thousands of normal rhythms, and only a small number of abnormal rhythms).\n",
        "\n",
        "How will you detect anomalies using an autoencoder? Recall that an autoencoder is trained to minimize reconstruction error. You will train an autoencoder on the normal rhythms only, then use it to reconstruct all the data. Our hypothesis is that the abnormal rhythms will have higher reconstruction error. You will then classify a rhythm as an anomaly if the reconstruction error surpasses a fixed threshold.\n",
        "\n",
        "Note: This Colab is adapted from the tensorflow [Intro to Autoencoders](https://www.tensorflow.org/tutorials/generative/autoencoder) example."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "e1_Y75QXJS6h"
      },
      "source": [
        "### Import TensorFlow and other libraries"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "import tensorflow as tf\n",
        "if tf.__version__ != \"2.10.0\":\n",
        "    print(f\"Current TensorFlow version: {tf.__version__}, switching to 2.10.0\")\n",
        "    \n",
        "    # Uninstall current TensorFlow version\n",
        "    !pip uninstall -y tensorflow\n",
        "    \n",
        "    # Install TensorFlow 2.10\n",
        "    !pip install tensorflow==2.10\n",
        "    \n",
        "    # After installation, restart runtime\n",
        "    print(\"TensorFlow 2.10 installed.\")\n",
        "    print(\"Please click on the Runtime > Restart session and run all.\")\n",
        "else:\n",
        "    print(\"TensorFlow 2.10 is already installed.\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "YfIk2es3hJEd"
      },
      "outputs": [],
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import tensorflow as tf\n",
        "\n",
        "from sklearn.metrics import accuracy_score, precision_score, recall_score, roc_curve, auc\n",
        "from sklearn.model_selection import train_test_split\n",
        "from tensorflow.keras import layers, losses\n",
        "from tensorflow.keras.models import Model"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "i5estNaur_Mh"
      },
      "source": [
        "### Load ECG data"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "y35nsXLPsDNX"
      },
      "source": [
        "The dataset you will use is based on one from [timeseriesclassification.com](http://www.timeseriesclassification.com/description.php?Dataset=ECG5000).\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 253
        },
        "id": "KmKRDJWgsFYa",
        "outputId": "24e5998d-35b6-422c-e5db-0b4a79d36a87"
      },
      "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>0</th>\n",
              "      <th>1</th>\n",
              "      <th>2</th>\n",
              "      <th>3</th>\n",
              "      <th>4</th>\n",
              "      <th>5</th>\n",
              "      <th>6</th>\n",
              "      <th>7</th>\n",
              "      <th>8</th>\n",
              "      <th>9</th>\n",
              "      <th>10</th>\n",
              "      <th>11</th>\n",
              "      <th>12</th>\n",
              "      <th>13</th>\n",
              "      <th>14</th>\n",
              "      <th>15</th>\n",
              "      <th>16</th>\n",
              "      <th>17</th>\n",
              "      <th>18</th>\n",
              "      <th>19</th>\n",
              "      <th>20</th>\n",
              "      <th>21</th>\n",
              "      <th>22</th>\n",
              "      <th>23</th>\n",
              "      <th>24</th>\n",
              "      <th>25</th>\n",
              "      <th>26</th>\n",
              "      <th>27</th>\n",
              "      <th>28</th>\n",
              "      <th>29</th>\n",
              "      <th>30</th>\n",
              "      <th>31</th>\n",
              "      <th>32</th>\n",
              "      <th>33</th>\n",
              "      <th>34</th>\n",
              "      <th>35</th>\n",
              "      <th>36</th>\n",
              "      <th>37</th>\n",
              "      <th>38</th>\n",
              "      <th>39</th>\n",
              "      <th>...</th>\n",
              "      <th>101</th>\n",
              "      <th>102</th>\n",
              "      <th>103</th>\n",
              "      <th>104</th>\n",
              "      <th>105</th>\n",
              "      <th>106</th>\n",
              "      <th>107</th>\n",
              "      <th>108</th>\n",
              "      <th>109</th>\n",
              "      <th>110</th>\n",
              "      <th>111</th>\n",
              "      <th>112</th>\n",
              "      <th>113</th>\n",
              "      <th>114</th>\n",
              "      <th>115</th>\n",
              "      <th>116</th>\n",
              "      <th>117</th>\n",
              "      <th>118</th>\n",
              "      <th>119</th>\n",
              "      <th>120</th>\n",
              "      <th>121</th>\n",
              "      <th>122</th>\n",
              "      <th>123</th>\n",
              "      <th>124</th>\n",
              "      <th>125</th>\n",
              "      <th>126</th>\n",
              "      <th>127</th>\n",
              "      <th>128</th>\n",
              "      <th>129</th>\n",
              "      <th>130</th>\n",
              "      <th>131</th>\n",
              "      <th>132</th>\n",
              "      <th>133</th>\n",
              "      <th>134</th>\n",
              "      <th>135</th>\n",
              "      <th>136</th>\n",
              "      <th>137</th>\n",
              "      <th>138</th>\n",
              "      <th>139</th>\n",
              "      <th>140</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>-0.112522</td>\n",
              "      <td>-2.827204</td>\n",
              "      <td>-3.773897</td>\n",
              "      <td>-4.349751</td>\n",
              "      <td>-4.376041</td>\n",
              "      <td>-3.474986</td>\n",
              "      <td>-2.181408</td>\n",
              "      <td>-1.818287</td>\n",
              "      <td>-1.250522</td>\n",
              "      <td>-0.477492</td>\n",
              "      <td>-0.363808</td>\n",
              "      <td>-0.491957</td>\n",
              "      <td>-0.421855</td>\n",
              "      <td>-0.309201</td>\n",
              "      <td>-0.495939</td>\n",
              "      <td>-0.342119</td>\n",
              "      <td>-0.355336</td>\n",
              "      <td>-0.367913</td>\n",
              "      <td>-0.316503</td>\n",
              "      <td>-0.412374</td>\n",
              "      <td>-0.471672</td>\n",
              "      <td>-0.413458</td>\n",
              "      <td>-0.364617</td>\n",
              "      <td>-0.449298</td>\n",
              "      <td>-0.471419</td>\n",
              "      <td>-0.424777</td>\n",
              "      <td>-0.462517</td>\n",
              "      <td>-0.552472</td>\n",
              "      <td>-0.475375</td>\n",
              "      <td>-0.694200</td>\n",
              "      <td>-0.701868</td>\n",
              "      <td>-0.593812</td>\n",
              "      <td>-0.660684</td>\n",
              "      <td>-0.713831</td>\n",
              "      <td>-0.769807</td>\n",
              "      <td>-0.672282</td>\n",
              "      <td>-0.653676</td>\n",
              "      <td>-0.639406</td>\n",
              "      <td>-0.559302</td>\n",
              "      <td>-0.591670</td>\n",
              "      <td>...</td>\n",
              "      <td>1.258179</td>\n",
              "      <td>1.433789</td>\n",
              "      <td>1.700533</td>\n",
              "      <td>1.999043</td>\n",
              "      <td>2.125341</td>\n",
              "      <td>1.993291</td>\n",
              "      <td>1.932246</td>\n",
              "      <td>1.797437</td>\n",
              "      <td>1.522284</td>\n",
              "      <td>1.251168</td>\n",
              "      <td>0.998730</td>\n",
              "      <td>0.483722</td>\n",
              "      <td>0.023132</td>\n",
              "      <td>-0.194914</td>\n",
              "      <td>-0.220917</td>\n",
              "      <td>-0.243737</td>\n",
              "      <td>-0.254695</td>\n",
              "      <td>-0.291136</td>\n",
              "      <td>-0.256490</td>\n",
              "      <td>-0.227874</td>\n",
              "      <td>-0.322423</td>\n",
              "      <td>-0.289286</td>\n",
              "      <td>-0.318170</td>\n",
              "      <td>-0.363654</td>\n",
              "      <td>-0.393456</td>\n",
              "      <td>-0.266419</td>\n",
              "      <td>-0.256823</td>\n",
              "      <td>-0.288694</td>\n",
              "      <td>-0.162338</td>\n",
              "      <td>0.160348</td>\n",
              "      <td>0.792168</td>\n",
              "      <td>0.933541</td>\n",
              "      <td>0.796958</td>\n",
              "      <td>0.578621</td>\n",
              "      <td>0.257740</td>\n",
              "      <td>0.228077</td>\n",
              "      <td>0.123431</td>\n",
              "      <td>0.925286</td>\n",
              "      <td>0.193137</td>\n",
              "      <td>1.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>-1.100878</td>\n",
              "      <td>-3.996840</td>\n",
              "      <td>-4.285843</td>\n",
              "      <td>-4.506579</td>\n",
              "      <td>-4.022377</td>\n",
              "      <td>-3.234368</td>\n",
              "      <td>-1.566126</td>\n",
              "      <td>-0.992258</td>\n",
              "      <td>-0.754680</td>\n",
              "      <td>0.042321</td>\n",
              "      <td>0.148951</td>\n",
              "      <td>0.183527</td>\n",
              "      <td>0.294876</td>\n",
              "      <td>0.190233</td>\n",
              "      <td>0.235575</td>\n",
              "      <td>0.253487</td>\n",
              "      <td>0.221742</td>\n",
              "      <td>0.050233</td>\n",
              "      <td>0.178042</td>\n",
              "      <td>0.139563</td>\n",
              "      <td>0.046794</td>\n",
              "      <td>0.043007</td>\n",
              "      <td>0.106544</td>\n",
              "      <td>0.012654</td>\n",
              "      <td>0.003995</td>\n",
              "      <td>0.045724</td>\n",
              "      <td>-0.045999</td>\n",
              "      <td>-0.072667</td>\n",
              "      <td>-0.071078</td>\n",
              "      <td>-0.153866</td>\n",
              "      <td>-0.227254</td>\n",
              "      <td>-0.249270</td>\n",
              "      <td>-0.253489</td>\n",
              "      <td>-0.332835</td>\n",
              "      <td>-0.264330</td>\n",
              "      <td>-0.345825</td>\n",
              "      <td>-0.310781</td>\n",
              "      <td>-0.334160</td>\n",
              "      <td>-0.306178</td>\n",
              "      <td>-0.174563</td>\n",
              "      <td>...</td>\n",
              "      <td>1.808428</td>\n",
              "      <td>2.164346</td>\n",
              "      <td>2.070747</td>\n",
              "      <td>1.903614</td>\n",
              "      <td>1.764455</td>\n",
              "      <td>1.507769</td>\n",
              "      <td>1.293428</td>\n",
              "      <td>0.894562</td>\n",
              "      <td>0.578016</td>\n",
              "      <td>0.244343</td>\n",
              "      <td>-0.286443</td>\n",
              "      <td>-0.515881</td>\n",
              "      <td>-0.732707</td>\n",
              "      <td>-0.832465</td>\n",
              "      <td>-0.803318</td>\n",
              "      <td>-0.836252</td>\n",
              "      <td>-0.777865</td>\n",
              "      <td>-0.774753</td>\n",
              "      <td>-0.733404</td>\n",
              "      <td>-0.721386</td>\n",
              "      <td>-0.832095</td>\n",
              "      <td>-0.711982</td>\n",
              "      <td>-0.751867</td>\n",
              "      <td>-0.757720</td>\n",
              "      <td>-0.853120</td>\n",
              "      <td>-0.766988</td>\n",
              "      <td>-0.688161</td>\n",
              "      <td>-0.519923</td>\n",
              "      <td>0.039406</td>\n",
              "      <td>0.560327</td>\n",
              "      <td>0.538356</td>\n",
              "      <td>0.656881</td>\n",
              "      <td>0.787490</td>\n",
              "      <td>0.724046</td>\n",
              "      <td>0.555784</td>\n",
              "      <td>0.476333</td>\n",
              "      <td>0.773820</td>\n",
              "      <td>1.119621</td>\n",
              "      <td>-1.436250</td>\n",
              "      <td>1.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>-0.567088</td>\n",
              "      <td>-2.593450</td>\n",
              "      <td>-3.874230</td>\n",
              "      <td>-4.584095</td>\n",
              "      <td>-4.187449</td>\n",
              "      <td>-3.151462</td>\n",
              "      <td>-1.742940</td>\n",
              "      <td>-1.490658</td>\n",
              "      <td>-1.183580</td>\n",
              "      <td>-0.394229</td>\n",
              "      <td>-0.282897</td>\n",
              "      <td>-0.356926</td>\n",
              "      <td>-0.287297</td>\n",
              "      <td>-0.399489</td>\n",
              "      <td>-0.473244</td>\n",
              "      <td>-0.379048</td>\n",
              "      <td>-0.399039</td>\n",
              "      <td>-0.178594</td>\n",
              "      <td>-0.339522</td>\n",
              "      <td>-0.498447</td>\n",
              "      <td>-0.337251</td>\n",
              "      <td>-0.425480</td>\n",
              "      <td>-0.423952</td>\n",
              "      <td>-0.463170</td>\n",
              "      <td>-0.493253</td>\n",
              "      <td>-0.549749</td>\n",
              "      <td>-0.529831</td>\n",
              "      <td>-0.530935</td>\n",
              "      <td>-0.502365</td>\n",
              "      <td>-0.417368</td>\n",
              "      <td>-0.526346</td>\n",
              "      <td>-0.471005</td>\n",
              "      <td>-0.676784</td>\n",
              "      <td>-0.898612</td>\n",
              "      <td>-0.610571</td>\n",
              "      <td>-0.530164</td>\n",
              "      <td>-0.765674</td>\n",
              "      <td>-0.581937</td>\n",
              "      <td>-0.537848</td>\n",
              "      <td>-0.556386</td>\n",
              "      <td>...</td>\n",
              "      <td>1.810988</td>\n",
              "      <td>2.185398</td>\n",
              "      <td>2.262985</td>\n",
              "      <td>2.052920</td>\n",
              "      <td>1.890488</td>\n",
              "      <td>1.793033</td>\n",
              "      <td>1.564784</td>\n",
              "      <td>1.234619</td>\n",
              "      <td>0.900302</td>\n",
              "      <td>0.551957</td>\n",
              "      <td>0.258222</td>\n",
              "      <td>-0.128587</td>\n",
              "      <td>-0.092585</td>\n",
              "      <td>-0.168606</td>\n",
              "      <td>-0.495989</td>\n",
              "      <td>-0.395034</td>\n",
              "      <td>-0.328238</td>\n",
              "      <td>-0.448138</td>\n",
              "      <td>-0.268230</td>\n",
              "      <td>-0.456415</td>\n",
              "      <td>-0.357867</td>\n",
              "      <td>-0.317508</td>\n",
              "      <td>-0.434112</td>\n",
              "      <td>-0.549203</td>\n",
              "      <td>-0.324615</td>\n",
              "      <td>-0.268082</td>\n",
              "      <td>-0.220384</td>\n",
              "      <td>-0.117429</td>\n",
              "      <td>0.614059</td>\n",
              "      <td>1.284825</td>\n",
              "      <td>0.886073</td>\n",
              "      <td>0.531452</td>\n",
              "      <td>0.311377</td>\n",
              "      <td>-0.021919</td>\n",
              "      <td>-0.713683</td>\n",
              "      <td>-0.532197</td>\n",
              "      <td>0.321097</td>\n",
              "      <td>0.904227</td>\n",
              "      <td>-0.421797</td>\n",
              "      <td>1.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>0.490473</td>\n",
              "      <td>-1.914407</td>\n",
              "      <td>-3.616364</td>\n",
              "      <td>-4.318823</td>\n",
              "      <td>-4.268016</td>\n",
              "      <td>-3.881110</td>\n",
              "      <td>-2.993280</td>\n",
              "      <td>-1.671131</td>\n",
              "      <td>-1.333884</td>\n",
              "      <td>-0.965629</td>\n",
              "      <td>-0.183319</td>\n",
              "      <td>-0.101657</td>\n",
              "      <td>-0.273874</td>\n",
              "      <td>-0.127818</td>\n",
              "      <td>-0.195983</td>\n",
              "      <td>-0.213523</td>\n",
              "      <td>-0.176473</td>\n",
              "      <td>-0.156932</td>\n",
              "      <td>-0.149172</td>\n",
              "      <td>-0.181510</td>\n",
              "      <td>-0.180074</td>\n",
              "      <td>-0.246151</td>\n",
              "      <td>-0.274260</td>\n",
              "      <td>-0.140960</td>\n",
              "      <td>-0.277449</td>\n",
              "      <td>-0.382549</td>\n",
              "      <td>-0.311937</td>\n",
              "      <td>-0.360093</td>\n",
              "      <td>-0.405968</td>\n",
              "      <td>-0.571433</td>\n",
              "      <td>-0.524106</td>\n",
              "      <td>-0.537886</td>\n",
              "      <td>-0.606778</td>\n",
              "      <td>-0.661446</td>\n",
              "      <td>-0.683375</td>\n",
              "      <td>-0.746683</td>\n",
              "      <td>-0.635662</td>\n",
              "      <td>-0.625231</td>\n",
              "      <td>-0.540094</td>\n",
              "      <td>-0.674995</td>\n",
              "      <td>...</td>\n",
              "      <td>1.772155</td>\n",
              "      <td>2.000769</td>\n",
              "      <td>1.925003</td>\n",
              "      <td>1.898426</td>\n",
              "      <td>1.720953</td>\n",
              "      <td>1.501711</td>\n",
              "      <td>1.422492</td>\n",
              "      <td>1.023225</td>\n",
              "      <td>0.776341</td>\n",
              "      <td>0.504426</td>\n",
              "      <td>0.056382</td>\n",
              "      <td>-0.233161</td>\n",
              "      <td>-0.406388</td>\n",
              "      <td>-0.327528</td>\n",
              "      <td>-0.460868</td>\n",
              "      <td>-0.402536</td>\n",
              "      <td>-0.345752</td>\n",
              "      <td>-0.354206</td>\n",
              "      <td>-0.439959</td>\n",
              "      <td>-0.425326</td>\n",
              "      <td>-0.439789</td>\n",
              "      <td>-0.451835</td>\n",
              "      <td>-0.395926</td>\n",
              "      <td>-0.448762</td>\n",
              "      <td>-0.391789</td>\n",
              "      <td>-0.376307</td>\n",
              "      <td>-0.461069</td>\n",
              "      <td>-0.253524</td>\n",
              "      <td>0.213006</td>\n",
              "      <td>0.491173</td>\n",
              "      <td>0.350816</td>\n",
              "      <td>0.499111</td>\n",
              "      <td>0.600345</td>\n",
              "      <td>0.842069</td>\n",
              "      <td>0.952074</td>\n",
              "      <td>0.990133</td>\n",
              "      <td>1.086798</td>\n",
              "      <td>1.403011</td>\n",
              "      <td>-0.383564</td>\n",
              "      <td>1.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>0.800232</td>\n",
              "      <td>-0.874252</td>\n",
              "      <td>-2.384761</td>\n",
              "      <td>-3.973292</td>\n",
              "      <td>-4.338224</td>\n",
              "      <td>-3.802422</td>\n",
              "      <td>-2.534510</td>\n",
              "      <td>-1.783423</td>\n",
              "      <td>-1.594450</td>\n",
              "      <td>-0.753199</td>\n",
              "      <td>-0.298107</td>\n",
              "      <td>-0.428928</td>\n",
              "      <td>-0.491351</td>\n",
              "      <td>-0.361304</td>\n",
              "      <td>-0.339296</td>\n",
              "      <td>-0.324952</td>\n",
              "      <td>-0.290113</td>\n",
              "      <td>-0.363051</td>\n",
              "      <td>-0.525684</td>\n",
              "      <td>-0.597423</td>\n",
              "      <td>-0.575523</td>\n",
              "      <td>-0.567503</td>\n",
              "      <td>-0.504555</td>\n",
              "      <td>-0.618406</td>\n",
              "      <td>-0.682814</td>\n",
              "      <td>-0.743849</td>\n",
              "      <td>-0.815588</td>\n",
              "      <td>-0.826902</td>\n",
              "      <td>-0.782374</td>\n",
              "      <td>-0.929462</td>\n",
              "      <td>-0.999672</td>\n",
              "      <td>-1.060969</td>\n",
              "      <td>-1.007877</td>\n",
              "      <td>-1.028735</td>\n",
              "      <td>-1.122629</td>\n",
              "      <td>-1.028650</td>\n",
              "      <td>-1.046515</td>\n",
              "      <td>-1.063372</td>\n",
              "      <td>-1.122423</td>\n",
              "      <td>-0.983242</td>\n",
              "      <td>...</td>\n",
              "      <td>1.155363</td>\n",
              "      <td>1.336254</td>\n",
              "      <td>1.627534</td>\n",
              "      <td>1.717594</td>\n",
              "      <td>1.696487</td>\n",
              "      <td>1.741686</td>\n",
              "      <td>1.674078</td>\n",
              "      <td>1.546928</td>\n",
              "      <td>1.331738</td>\n",
              "      <td>1.110168</td>\n",
              "      <td>0.922210</td>\n",
              "      <td>0.521777</td>\n",
              "      <td>0.154852</td>\n",
              "      <td>-0.123861</td>\n",
              "      <td>-0.202998</td>\n",
              "      <td>-0.247956</td>\n",
              "      <td>-0.219122</td>\n",
              "      <td>-0.214695</td>\n",
              "      <td>-0.319215</td>\n",
              "      <td>-0.198597</td>\n",
              "      <td>-0.151618</td>\n",
              "      <td>-0.129593</td>\n",
              "      <td>-0.074939</td>\n",
              "      <td>-0.196807</td>\n",
              "      <td>-0.174795</td>\n",
              "      <td>-0.208833</td>\n",
              "      <td>-0.210754</td>\n",
              "      <td>-0.100485</td>\n",
              "      <td>0.197446</td>\n",
              "      <td>0.966606</td>\n",
              "      <td>1.148884</td>\n",
              "      <td>0.958434</td>\n",
              "      <td>1.059025</td>\n",
              "      <td>1.371682</td>\n",
              "      <td>1.277392</td>\n",
              "      <td>0.960304</td>\n",
              "      <td>0.971020</td>\n",
              "      <td>1.614392</td>\n",
              "      <td>1.421456</td>\n",
              "      <td>1.0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>5 rows × 141 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "        0         1         2         3    ...       137       138       139  140\n",
              "0 -0.112522 -2.827204 -3.773897 -4.349751  ...  0.123431  0.925286  0.193137  1.0\n",
              "1 -1.100878 -3.996840 -4.285843 -4.506579  ...  0.773820  1.119621 -1.436250  1.0\n",
              "2 -0.567088 -2.593450 -3.874230 -4.584095  ...  0.321097  0.904227 -0.421797  1.0\n",
              "3  0.490473 -1.914407 -3.616364 -4.318823  ...  1.086798  1.403011 -0.383564  1.0\n",
              "4  0.800232 -0.874252 -2.384761 -3.973292  ...  0.971020  1.614392  1.421456  1.0\n",
              "\n",
              "[5 rows x 141 columns]"
            ]
          },
          "execution_count": 2,
          "metadata": {
            "tags": []
          },
          "output_type": "execute_result"
        }
      ],
      "source": [
        "# Download the dataset\n",
        "dataframe = pd.read_csv('http://storage.googleapis.com/download.tensorflow.org/data/ecg.csv', header=None)\n",
        "raw_data = dataframe.values\n",
        "dataframe.head()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "id": "UmuCPVYKsKKx"
      },
      "outputs": [],
      "source": [
        "# The last element contains the labels\n",
        "labels = raw_data[:, -1]\n",
        "\n",
        "# The other data points are the electrocadriogram data\n",
        "data = raw_data[:, 0:-1]\n",
        "\n",
        "train_data, test_data, train_labels, test_labels = train_test_split(\n",
        "    data, labels, test_size=0.2, random_state=21\n",
        ")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "byK2vP7hsMbz"
      },
      "source": [
        "Normalize the data to `[0,1]` to improve training accuracy.\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "id": "tgMZVWRKsPx6"
      },
      "outputs": [],
      "source": [
        "min_val = tf.reduce_min(train_data)\n",
        "max_val = tf.reduce_max(train_data)\n",
        "\n",
        "train_data = (train_data - min_val) / (max_val - min_val)\n",
        "test_data = (test_data - min_val) / (max_val - min_val)\n",
        "\n",
        "train_data = tf.cast(train_data, tf.float32)\n",
        "test_data = tf.cast(test_data, tf.float32)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BdSYr2IPsTiz"
      },
      "source": [
        "You will train the autoencoder using only the normal rhythms, which are labeled in this dataset as `1`. Separate the normal rhythms from the abnormal rhythms."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "id": "VvK4NRe8sVhE"
      },
      "outputs": [],
      "source": [
        "train_labels = train_labels.astype(bool)\n",
        "test_labels = test_labels.astype(bool)\n",
        "\n",
        "normal_train_data = train_data[train_labels]\n",
        "normal_test_data = test_data[test_labels]\n",
        "\n",
        "anomalous_train_data = train_data[~train_labels]\n",
        "anomalous_test_data = test_data[~test_labels]"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wVcTBDo-CqFS"
      },
      "source": [
        "Plot a normal ECG. "
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 281
        },
        "id": "ZTlMIrpmseYe",
        "outputId": "33ca8596-c640-4ecd-a42f-dcb55e83d6c6"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light",
            "tags": []
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "plt.grid()\n",
        "plt.plot(np.arange(140), normal_train_data[0])\n",
        "plt.title(\"A Normal ECG\")\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "QpI9by2ZA0NN"
      },
      "source": [
        "Plot an anomalous ECG."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 281
        },
        "id": "zrpXREF2siBr",
        "outputId": "73114b9c-132e-4f23-a845-e582abc5a7d4"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light",
            "tags": []
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "plt.grid()\n",
        "plt.plot(np.arange(140), anomalous_train_data[0])\n",
        "plt.title(\"An Anomalous ECG\")\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0DS6QKZJslZz"
      },
      "source": [
        "### Build the model\n",
        "\n",
        "After training and evaluating the example model, try modifying the size and number of layers to build an understanding for autoencoder architectures.\n",
        "\n",
        "Note: Changing the size of the embedding (the smallest layer) can produce interesting results. Feel free to play around with that layer size."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {
        "id": "bf6owZQDsp9y"
      },
      "outputs": [],
      "source": [
        "class AnomalyDetector(Model):\n",
        "  def __init__(self):\n",
        "    super(AnomalyDetector, self).__init__()\n",
        "    self.encoder = tf.keras.Sequential([\n",
        "      layers.Dense(32, activation=\"relu\"),\n",
        "      layers.Dense(16, activation=\"relu\"),\n",
        "      layers.Dense(8, activation=\"relu\")]) # Smallest Layer Defined Here\n",
        "    \n",
        "    self.decoder = tf.keras.Sequential([\n",
        "      layers.Dense(16, activation=\"relu\"),\n",
        "      layers.Dense(32, activation=\"relu\"),\n",
        "      layers.Dense(140, activation=\"sigmoid\")])\n",
        "    \n",
        "  def call(self, x):\n",
        "    encoded = self.encoder(x)\n",
        "    decoded = self.decoder(encoded)\n",
        "    return decoded\n",
        "\n",
        "autoencoder = AnomalyDetector()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {
        "id": "gwRpBBbg463S"
      },
      "outputs": [],
      "source": [
        "autoencoder.compile(optimizer='adam', loss='mae')"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zuTy60STBEy4"
      },
      "source": [
        "### Train the model\n",
        "\n",
        "Notice that the autoencoder is trained using only the normal ECGs, but is evaluated using the full test set."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "V6NFSs-jsty2",
        "outputId": "8d98755d-9a1a-4b92-ad80-d91e7c5cb74a"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "5/5 [==============================] - 0s 27ms/step - loss: 0.0583 - val_loss: 0.0530\n",
            "Epoch 2/20\n",
            "5/5 [==============================] - 0s 5ms/step - loss: 0.0554 - val_loss: 0.0511\n",
            "Epoch 3/20\n",
            "5/5 [==============================] - 0s 5ms/step - loss: 0.0524 - val_loss: 0.0490\n",
            "Epoch 4/20\n",
            "5/5 [==============================] - 0s 5ms/step - loss: 0.0481 - val_loss: 0.0477\n",
            "Epoch 5/20\n",
            "5/5 [==============================] - 0s 5ms/step - loss: 0.0439 - val_loss: 0.0461\n",
            "Epoch 6/20\n",
            "5/5 [==============================] - 0s 5ms/step - loss: 0.0398 - val_loss: 0.0433\n",
            "Epoch 7/20\n",
            "5/5 [==============================] - 0s 5ms/step - loss: 0.0361 - val_loss: 0.0418\n",
            "Epoch 8/20\n",
            "5/5 [==============================] - 0s 5ms/step - loss: 0.0332 - val_loss: 0.0411\n",
            "Epoch 9/20\n",
            "5/5 [==============================] - 0s 5ms/step - loss: 0.0308 - val_loss: 0.0396\n",
            "Epoch 10/20\n",
            "5/5 [==============================] - 0s 5ms/step - loss: 0.0289 - val_loss: 0.0388\n",
            "Epoch 11/20\n",
            "5/5 [==============================] - 0s 5ms/step - loss: 0.0275 - val_loss: 0.0379\n",
            "Epoch 12/20\n",
            "5/5 [==============================] - 0s 5ms/step - loss: 0.0266 - val_loss: 0.0372\n",
            "Epoch 13/20\n",
            "5/5 [==============================] - 0s 5ms/step - loss: 0.0259 - val_loss: 0.0367\n",
            "Epoch 14/20\n",
            "5/5 [==============================] - 0s 5ms/step - loss: 0.0253 - val_loss: 0.0362\n",
            "Epoch 15/20\n",
            "5/5 [==============================] - 0s 6ms/step - loss: 0.0247 - val_loss: 0.0357\n",
            "Epoch 16/20\n",
            "5/5 [==============================] - 0s 5ms/step - loss: 0.0242 - val_loss: 0.0353\n",
            "Epoch 17/20\n",
            "5/5 [==============================] - 0s 5ms/step - loss: 0.0236 - val_loss: 0.0347\n",
            "Epoch 18/20\n",
            "5/5 [==============================] - 0s 5ms/step - loss: 0.0230 - val_loss: 0.0345\n",
            "Epoch 19/20\n",
            "5/5 [==============================] - 0s 5ms/step - loss: 0.0225 - val_loss: 0.0340\n",
            "Epoch 20/20\n",
            "5/5 [==============================] - 0s 5ms/step - loss: 0.0220 - val_loss: 0.0340\n"
          ]
        }
      ],
      "source": [
        "history = autoencoder.fit(normal_train_data, normal_train_data, \n",
        "          epochs=20, \n",
        "          batch_size=512,\n",
        "          validation_data=(test_data, test_data),\n",
        "          shuffle=True)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 287
        },
        "id": "OEexphFwwTQS",
        "outputId": "baaf7936-f929-4b7d-8dbe-9c8c477c1453"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7eff7e729be0>"
            ]
          },
          "execution_count": 11,
          "metadata": {
            "tags": []
          },
          "output_type": "execute_result"
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light",
            "tags": []
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "plt.plot(history.history[\"loss\"], label=\"Training Loss\")\n",
        "plt.plot(history.history[\"val_loss\"], label=\"Validation Loss\")\n",
        "plt.legend()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ceI5lKv1BT-A"
      },
      "source": [
        "### Evaluate Training\n",
        "\n",
        "You will soon classify an ECG as anomalous if the reconstruction error is greater than one standard deviation from the normal training examples. First, let's plot a normal ECG from the training set, the reconstruction after it's encoded and decoded by the autoencoder, and the reconstruction error."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 12,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 265
        },
        "id": "hmsk4DuktxJ2",
        "outputId": "280a04fe-1f7e-4139-c164-ad55c133691c"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light",
            "tags": []
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "encoded_imgs = autoencoder.encoder(normal_test_data).numpy()\n",
        "decoded_imgs = autoencoder.decoder(encoded_imgs).numpy()\n",
        "\n",
        "plt.plot(normal_test_data[0],'b')\n",
        "plt.plot(decoded_imgs[0],'r')\n",
        "plt.fill_between(np.arange(140), decoded_imgs[0], normal_test_data[0], color='lightcoral' )\n",
        "plt.legend(labels=[\"Input\", \"Reconstruction\", \"Error\"])\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ocA_q9ufB_aF"
      },
      "source": [
        "Create a similar plot, this time for an anomalous test example."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 13,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 266
        },
        "id": "vNFTuPhLwTBn",
        "outputId": "627769b9-9494-4eaa-96ef-a15bea8a21cb"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light",
            "tags": []
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "encoded_imgs = autoencoder.encoder(anomalous_test_data).numpy()\n",
        "decoded_imgs = autoencoder.decoder(encoded_imgs).numpy()\n",
        "\n",
        "plt.plot(anomalous_test_data[0],'b')\n",
        "plt.plot(decoded_imgs[0],'r')\n",
        "plt.fill_between(np.arange(140), decoded_imgs[0], anomalous_test_data[0], color='lightcoral' )\n",
        "plt.legend(labels=[\"Input\", \"Reconstruction\", \"Error\"])\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ocimg3MBswdS"
      },
      "source": [
        "### Detect anomalies"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Xnh8wmkDsypN"
      },
      "source": [
        "Detect anomalies by calculating whether the reconstruction loss is greater than a fixed threshold. In this tutorial, you will calculate the mean average error for normal examples from the training set, then classify future examples as anomalous if the reconstruction error is higher than one standard deviation from the training set.\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TeuT8uTA5Y_w"
      },
      "source": [
        "Plot the reconstruction error on normal ECGs from the training set"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 14,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "id": "gwLuxrb-s0ss",
        "outputId": "3f81e86b-90c0-4875-8282-141c7af39a2e"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light",
            "tags": []
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "reconstructions = autoencoder.predict(normal_train_data)\n",
        "train_loss = tf.keras.losses.mae(reconstructions, normal_train_data)\n",
        "\n",
        "plt.hist(train_loss, bins=50)\n",
        "plt.xlabel(\"Train loss\")\n",
        "plt.ylabel(\"No of examples\")\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mh-3ChEF5hog"
      },
      "source": [
        "Choose a threshold value that is one standard deviations above the mean."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 15,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "82hkl0Chs3P_",
        "outputId": "72c3a0a0-560e-4e88-d7dc-6082fc7c2b69"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Threshold:  0.033812687\n"
          ]
        }
      ],
      "source": [
        "threshold = np.mean(train_loss) + np.std(train_loss)\n",
        "print(\"Threshold: \", threshold)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "uEGlA1Be50Nj"
      },
      "source": [
        "Note: There are other strategies you could use to select a threshold value above which test examples should be classified as anomalous, the correct approach will depend on your dataset. You will learn more later in the colab."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zpLSDAeb51D_"
      },
      "source": [
        "If you examine the recontruction error for the anomalous examples in the test set, you'll notice most have greater reconstruction error than the threshold. By varing the threshold, you can adjust the [precision](https://developers.google.com/machine-learning/glossary#precision) and [recall](https://developers.google.com/machine-learning/glossary#recall) of your classifier. "
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 16,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "id": "sKVwjQK955Wy",
        "outputId": "e2a1a1a9-f7db-4f3f-d686-f71b6bb5de31"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light",
            "tags": []
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "reconstructions = autoencoder.predict(anomalous_test_data)\n",
        "test_loss = tf.keras.losses.mae(reconstructions, anomalous_test_data)\n",
        "\n",
        "plt.hist(test_loss, bins=50)\n",
        "plt.xlabel(\"Test loss\")\n",
        "plt.ylabel(\"No of examples\")\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PFVk_XGE6AX2"
      },
      "source": [
        "Classify an ECG as an anomaly if the reconstruction error is greater than the threshold."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 17,
      "metadata": {
        "id": "mkgJZfhh6CHr"
      },
      "outputs": [],
      "source": [
        "def predict(model, data, threshold):\n",
        "  reconstructions = model(data)\n",
        "  loss = tf.keras.losses.mae(reconstructions, data)\n",
        "  return tf.math.less(loss, threshold), loss\n",
        "\n",
        "def print_stats(predictions, labels):\n",
        "  print(\"Accuracy = {}\".format(accuracy_score(labels, predictions)))\n",
        "  print(\"Precision = {}\".format(precision_score(labels, predictions)))\n",
        "  print(\"Recall = {}\".format(recall_score(labels, predictions)))"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 18,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "sOcfXfXq6FBd",
        "outputId": "6b7d9e1d-073d-4575-d00c-535e5f603a56"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Accuracy = 0.944\n",
            "Precision = 0.9921875\n",
            "Recall = 0.9071428571428571\n"
          ]
        }
      ],
      "source": [
        "preds, scores = predict(autoencoder, test_data, threshold)\n",
        "print_stats(preds, test_labels)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "N2sl_2z-uOv3"
      },
      "source": [
        "### ROC and AUC Metrics\n",
        "\n",
        "We've created a fairly accurate model for anomaly detection but our accuracy is highly dependant on the threshold we select. \n",
        "\n",
        "What if we wanted to evaluate how different thresholds impact our true positive and false positive rates?\n",
        "\n",
        "Enter Receiver Operating Characteristic (ROC) plots! This metric allows us to visualize the tradeoff between predicting anomalies as normal (false positives) and predicting normal data as an anomaly (false negative). Remember that normal rhythms are labeled as `1` in this dataset.  "
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 19,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "LketUNanyLte",
        "outputId": "9af9b6e9-bb92-40d7-c784-edddce7269fe"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light",
            "tags": []
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "fpr = []\n",
        "tpr = []\n",
        "#the test labels are flipped to match how the roc_curve function expects them.\n",
        "flipped_labels = 1-test_labels \n",
        "fpr, tpr, _ = roc_curve(flipped_labels, scores)\n",
        "plt.figure()\n",
        "lw = 2\n",
        "plt.plot(fpr, tpr, color='darkorange',\n",
        "         lw=lw, label='ROC curve ')\n",
        "plt.plot([0, 1], [0, 1], color='navy', lw=lw, linestyle='--')\n",
        "plt.xlim([0.0, 1.0])\n",
        "plt.ylim([0.0, 1.05])\n",
        "plt.xlabel('False Positive Rate')\n",
        "plt.ylabel('True Positive Rate')\n",
        "plt.title('Receiver operating characteristic example')\n",
        "plt.legend(loc=\"lower right\")\n",
        "plt.show()\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Hk6LfUhB1ohP"
      },
      "source": [
        "Since our model does a great job in diferentiating normal rythms from abnormal ones it seems easy to pick the threshold that would give us the high true positive rate (TPR) and low false positive rate (FPR) that is at the 'knee' of the curve.\n",
        "\n",
        "However, in some cases there may be an application constraint that requires a specific TPR or FPR, in which case we would have to move off of the 'knee' and sacrifice overall accuracy. In this case we might rather have false alarms than miss a potentially dangerous rythm."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6sH6Yv-kx17h"
      },
      "source": [
        "Now that we understand how to visualize the impact of the selected threshold, What if we wanted to compare the performance of models without factoring in the threshold? Simply comparing the accuracy won't work since that depends on the threshold you pick and that won't have the same impact across models.\n",
        "\n",
        "Instead we can measure the area under the curve (AUC) in the ROC plot. One way to interpret the AUC metric is as the probability that the model ranks a random positive example more highly than a random negative example.\n",
        "\n",
        "In general the AUC is a useful metic for comparison as it is threshold invariant *and* scale invariant"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 20,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "qDMEZfCvyoaP",
        "outputId": "f2fe4b99-6933-4fb5-f7e4-cd01cf2e9b29"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "0.9572930194805195\n"
          ]
        }
      ],
      "source": [
        "roc_auc = auc(fpr, tpr)\n",
        "print(roc_auc)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PPWFHwK0s83Q"
      },
      "source": [
        "## Links to Continue Learning\n",
        "\n",
        "If you would like to learn more about anomaly detection with autoencoders, check out this excellent [interactive example](https://anomagram.fastforwardlabs.com/#/) built with TensorFlow.js by Victor Dibia. For a real-world use case, you can learn how [Airbus Detects Anomalies in ISS Telemetry Data](https://blog.tensorflow.org/2020/04/how-airbus-detects-anomalies-iss-telemetry-data-tfx.html) using TensorFlow. \n",
        "\n",
        "To learn more about the basics of autoencoders, try the basis for this colab, [TensorFlow's Intro to Autoencoders](https://colab.research.google.com/github/tensorflow/docs/blob/master/site/en/tutorials/generative/autoencoder.ipynb#scrollTo=xfNT-mlFwxVM)\n",
        "\n",
        "For more info, consider reading this [blog post](https://blog.keras.io/building-autoencoders-in-keras.html) by François Chollet, and check out chapter 14 from [Deep Learning](https://www.deeplearningbook.org/) by Ian Goodfellow, Yoshua Bengio, and Aaron Courville.\n"
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "collapsed_sections": [],
      "name": "3-8-13-Autoencoders.ipynb",
      "provenance": [],
      "toc_visible": true
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}