hyperparams_analysis.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from collections import defaultdict\n",
    "import pandas\n",
    "from matplotlib import pyplot as plt\n",
    "import pandas_format\n",
    "pandas.set_option('display.max_columns', None)\n",
    "\n",
    "with open('/home/ict520c/Downloads/hparams_iteration1.json', 'r') as f:\n",
    "    import json\n",
    "    json_data = json.load(f)\n",
    "\n",
    "df_data = defaultdict(list)\n",
    "for iteration in json_data['sessionGroups']:\n",
    "    for param_name, param_value in iteration['hparams'].items():\n",
    "        df_data[param_name].append(param_value)\n",
    "\n",
    "    for metric in iteration['metricValues']:\n",
    "        df_data[metric['name']['tag']].append(metric['value'])\n",
    "\n",
    "df = pandas.DataFrame(df_data)\n",
    "# df.to_csv('hparams_iteration1.csv')\n",
    "\n",
    "metric_columns = df.columns[df.columns.str.contains('_')].drop('f1_score')\n",
    "hparam_columns = df.columns[~df.columns.str.contains('_')]\n",
    "m, h = metric_columns, hparam_columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "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>class_precision</th>\n",
       "      <th>class_recall</th>\n",
       "      <th>prec_15</th>\n",
       "      <th>prec_25</th>\n",
       "      <th>prec_5</th>\n",
       "      <th>recall_15</th>\n",
       "      <th>recall_25</th>\n",
       "      <th>recall_5</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Learning Rate</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0.00</th>\n",
       "      <td>0.70</td>\n",
       "      <td>0.81</td>\n",
       "      <td>0.65</td>\n",
       "      <td>0.78</td>\n",
       "      <td>0.31</td>\n",
       "      <td>0.76</td>\n",
       "      <td>0.87</td>\n",
       "      <td>0.40</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.00</th>\n",
       "      <td>0.78</td>\n",
       "      <td>0.68</td>\n",
       "      <td>0.51</td>\n",
       "      <td>0.71</td>\n",
       "      <td>0.21</td>\n",
       "      <td>0.39</td>\n",
       "      <td>0.48</td>\n",
       "      <td>0.19</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.00</th>\n",
       "      <td>0.86</td>\n",
       "      <td>0.63</td>\n",
       "      <td>0.42</td>\n",
       "      <td>0.63</td>\n",
       "      <td>0.14</td>\n",
       "      <td>0.18</td>\n",
       "      <td>0.23</td>\n",
       "      <td>0.08</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               class_precision  class_recall  prec_15  prec_25  prec_5  \\\n",
       "Learning Rate                                                            \n",
       "0.00                      0.70          0.81     0.65     0.78    0.31   \n",
       "0.00                      0.78          0.68     0.51     0.71    0.21   \n",
       "0.00                      0.86          0.63     0.42     0.63    0.14   \n",
       "\n",
       "               recall_15  recall_25  recall_5  \n",
       "Learning Rate                                  \n",
       "0.00                0.76       0.87      0.40  \n",
       "0.00                0.39       0.48      0.19  \n",
       "0.00                0.18       0.23      0.08  "
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.groupby('Learning Rate')[metric_columns].mean()\n",
    "# LR = 3e-4"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "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>Last Convolution</th>\n",
       "      <th># of GRUs</th>\n",
       "      <th>GRU size</th>\n",
       "      <th>Convolutional Filters</th>\n",
       "      <th>First Convolution</th>\n",
       "      <th>class_precision</th>\n",
       "      <th>class_recall</th>\n",
       "      <th>f1_score</th>\n",
       "      <th>prec_15</th>\n",
       "      <th>prec_25</th>\n",
       "      <th>prec_5</th>\n",
       "      <th>recall_15</th>\n",
       "      <th>recall_25</th>\n",
       "      <th>recall_5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>5.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>512.00</td>\n",
       "      <td>16.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.65</td>\n",
       "      <td>0.81</td>\n",
       "      <td>0.40</td>\n",
       "      <td>0.71</td>\n",
       "      <td>0.80</td>\n",
       "      <td>0.38</td>\n",
       "      <td>0.88</td>\n",
       "      <td>0.95</td>\n",
       "      <td>0.50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>4.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>128.00</td>\n",
       "      <td>16.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.74</td>\n",
       "      <td>0.86</td>\n",
       "      <td>0.36</td>\n",
       "      <td>0.70</td>\n",
       "      <td>0.81</td>\n",
       "      <td>0.32</td>\n",
       "      <td>0.86</td>\n",
       "      <td>0.95</td>\n",
       "      <td>0.43</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>5.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>128.00</td>\n",
       "      <td>16.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>0.72</td>\n",
       "      <td>0.84</td>\n",
       "      <td>0.42</td>\n",
       "      <td>0.68</td>\n",
       "      <td>0.80</td>\n",
       "      <td>0.32</td>\n",
       "      <td>0.83</td>\n",
       "      <td>0.94</td>\n",
       "      <td>0.42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>4.00</td>\n",
       "      <td>2.00</td>\n",
       "      <td>384.00</td>\n",
       "      <td>16.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.66</td>\n",
       "      <td>0.84</td>\n",
       "      <td>0.38</td>\n",
       "      <td>0.65</td>\n",
       "      <td>0.76</td>\n",
       "      <td>0.29</td>\n",
       "      <td>0.89</td>\n",
       "      <td>0.97</td>\n",
       "      <td>0.46</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>4.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>384.00</td>\n",
       "      <td>16.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.68</td>\n",
       "      <td>0.85</td>\n",
       "      <td>0.39</td>\n",
       "      <td>0.69</td>\n",
       "      <td>0.80</td>\n",
       "      <td>0.33</td>\n",
       "      <td>0.88</td>\n",
       "      <td>0.97</td>\n",
       "      <td>0.47</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>460</th>\n",
       "      <td>5.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>256.00</td>\n",
       "      <td>72.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.71</td>\n",
       "      <td>0.69</td>\n",
       "      <td>0.26</td>\n",
       "      <td>0.47</td>\n",
       "      <td>0.65</td>\n",
       "      <td>0.22</td>\n",
       "      <td>0.64</td>\n",
       "      <td>0.79</td>\n",
       "      <td>0.35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>462</th>\n",
       "      <td>4.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>256.00</td>\n",
       "      <td>72.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>0.68</td>\n",
       "      <td>0.82</td>\n",
       "      <td>0.38</td>\n",
       "      <td>0.65</td>\n",
       "      <td>0.76</td>\n",
       "      <td>0.31</td>\n",
       "      <td>0.85</td>\n",
       "      <td>0.94</td>\n",
       "      <td>0.45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>466</th>\n",
       "      <td>4.00</td>\n",
       "      <td>2.00</td>\n",
       "      <td>64.00</td>\n",
       "      <td>72.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>0.72</td>\n",
       "      <td>0.81</td>\n",
       "      <td>0.36</td>\n",
       "      <td>0.68</td>\n",
       "      <td>0.80</td>\n",
       "      <td>0.32</td>\n",
       "      <td>0.85</td>\n",
       "      <td>0.94</td>\n",
       "      <td>0.46</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>475</th>\n",
       "      <td>5.00</td>\n",
       "      <td>2.00</td>\n",
       "      <td>384.00</td>\n",
       "      <td>72.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.68</td>\n",
       "      <td>0.72</td>\n",
       "      <td>0.26</td>\n",
       "      <td>0.57</td>\n",
       "      <td>0.76</td>\n",
       "      <td>0.24</td>\n",
       "      <td>0.63</td>\n",
       "      <td>0.80</td>\n",
       "      <td>0.28</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>476</th>\n",
       "      <td>4.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>64.00</td>\n",
       "      <td>72.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>0.68</td>\n",
       "      <td>0.81</td>\n",
       "      <td>0.34</td>\n",
       "      <td>0.60</td>\n",
       "      <td>0.73</td>\n",
       "      <td>0.27</td>\n",
       "      <td>0.80</td>\n",
       "      <td>0.91</td>\n",
       "      <td>0.41</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>160 rows × 14 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     Last Convolution  # of GRUs  GRU size  Convolutional Filters  \\\n",
       "0                5.00       1.00    512.00                  16.00   \n",
       "2                4.00       1.00    128.00                  16.00   \n",
       "3                5.00       1.00    128.00                  16.00   \n",
       "5                4.00       2.00    384.00                  16.00   \n",
       "7                4.00       1.00    384.00                  16.00   \n",
       "..                ...        ...       ...                    ...   \n",
       "460              5.00       1.00    256.00                  72.00   \n",
       "462              4.00       1.00    256.00                  72.00   \n",
       "466              4.00       2.00     64.00                  72.00   \n",
       "475              5.00       2.00    384.00                  72.00   \n",
       "476              4.00       1.00     64.00                  72.00   \n",
       "\n",
       "     First Convolution  class_precision  class_recall  f1_score  prec_15  \\\n",
       "0                 0.00             0.65          0.81      0.40     0.71   \n",
       "2                 0.00             0.74          0.86      0.36     0.70   \n",
       "3                 1.00             0.72          0.84      0.42     0.68   \n",
       "5                 0.00             0.66          0.84      0.38     0.65   \n",
       "7                 0.00             0.68          0.85      0.39     0.69   \n",
       "..                 ...              ...           ...       ...      ...   \n",
       "460               0.00             0.71          0.69      0.26     0.47   \n",
       "462               1.00             0.68          0.82      0.38     0.65   \n",
       "466               1.00             0.72          0.81      0.36     0.68   \n",
       "475               0.00             0.68          0.72      0.26     0.57   \n",
       "476               1.00             0.68          0.81      0.34     0.60   \n",
       "\n",
       "     prec_25  prec_5  recall_15  recall_25  recall_5  \n",
       "0       0.80    0.38       0.88       0.95      0.50  \n",
       "2       0.81    0.32       0.86       0.95      0.43  \n",
       "3       0.80    0.32       0.83       0.94      0.42  \n",
       "5       0.76    0.29       0.89       0.97      0.46  \n",
       "7       0.80    0.33       0.88       0.97      0.47  \n",
       "..       ...     ...        ...        ...       ...  \n",
       "460     0.65    0.22       0.64       0.79      0.35  \n",
       "462     0.76    0.31       0.85       0.94      0.45  \n",
       "466     0.80    0.32       0.85       0.94      0.46  \n",
       "475     0.76    0.24       0.63       0.80      0.28  \n",
       "476     0.73    0.27       0.80       0.91      0.41  \n",
       "\n",
       "[160 rows x 14 columns]"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "h = h.drop(['Learning Rate'])\n",
    "df2 = df.loc[df['Learning Rate'] == 3e-4].drop('Learning Rate', axis=1)\n",
    "df2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 288x460.8 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "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>Last Convolution</th>\n",
       "      <th># of GRUs</th>\n",
       "      <th>GRU size</th>\n",
       "      <th>Convolutional Filters</th>\n",
       "      <th>First Convolution</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>class_precision</th>\n",
       "      <td>-0.103</td>\n",
       "      <td>0.278</td>\n",
       "      <td>0.108</td>\n",
       "      <td>-0.197</td>\n",
       "      <td>0.010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>class_recall</th>\n",
       "      <td>-0.471</td>\n",
       "      <td>0.051</td>\n",
       "      <td>-0.065</td>\n",
       "      <td>-0.503</td>\n",
       "      <td>0.201</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>prec_15</th>\n",
       "      <td>-0.404</td>\n",
       "      <td>0.181</td>\n",
       "      <td>0.142</td>\n",
       "      <td>-0.608</td>\n",
       "      <td>0.208</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>prec_25</th>\n",
       "      <td>-0.199</td>\n",
       "      <td>0.251</td>\n",
       "      <td>0.138</td>\n",
       "      <td>-0.513</td>\n",
       "      <td>0.247</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>prec_5</th>\n",
       "      <td>-0.354</td>\n",
       "      <td>0.159</td>\n",
       "      <td>0.167</td>\n",
       "      <td>-0.626</td>\n",
       "      <td>0.189</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>recall_15</th>\n",
       "      <td>-0.464</td>\n",
       "      <td>-0.108</td>\n",
       "      <td>-0.119</td>\n",
       "      <td>-0.436</td>\n",
       "      <td>0.136</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>recall_25</th>\n",
       "      <td>-0.413</td>\n",
       "      <td>-0.097</td>\n",
       "      <td>-0.162</td>\n",
       "      <td>-0.364</td>\n",
       "      <td>0.109</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>recall_5</th>\n",
       "      <td>-0.468</td>\n",
       "      <td>-0.105</td>\n",
       "      <td>-0.083</td>\n",
       "      <td>-0.483</td>\n",
       "      <td>0.153</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 Last Convolution  # of GRUs  GRU size  Convolutional Filters  \\\n",
       "class_precision            -0.103      0.278     0.108                 -0.197   \n",
       "class_recall               -0.471      0.051    -0.065                 -0.503   \n",
       "prec_15                    -0.404      0.181     0.142                 -0.608   \n",
       "prec_25                    -0.199      0.251     0.138                 -0.513   \n",
       "prec_5                     -0.354      0.159     0.167                 -0.626   \n",
       "recall_15                  -0.464     -0.108    -0.119                 -0.436   \n",
       "recall_25                  -0.413     -0.097    -0.162                 -0.364   \n",
       "recall_5                   -0.468     -0.105    -0.083                 -0.483   \n",
       "\n",
       "                 First Convolution  \n",
       "class_precision              0.010  \n",
       "class_recall                 0.201  \n",
       "prec_15                      0.208  \n",
       "prec_25                      0.247  \n",
       "prec_5                       0.189  \n",
       "recall_15                    0.136  \n",
       "recall_25                    0.109  \n",
       "recall_5                     0.153  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "corr = df2.corr()\n",
    "plt.matshow(corr.loc[m, h].abs())\n",
    "plt.xticks(range(len(h)), h, rotation='vertical')\n",
    "plt.yticks(range(len(m)), m)\n",
    "plt.show()\n",
    "with pandas_format.PandasFloatFormatter('{:,.3f}'):\n",
    "    display(corr.loc[m, h])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "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>Last Convolution</th>\n",
       "      <th># of GRUs</th>\n",
       "      <th>GRU size</th>\n",
       "      <th>Convolutional Filters</th>\n",
       "      <th>First Convolution</th>\n",
       "      <th>class_precision</th>\n",
       "      <th>class_recall</th>\n",
       "      <th>prec_15</th>\n",
       "      <th>prec_25</th>\n",
       "      <th>prec_5</th>\n",
       "      <th>recall_15</th>\n",
       "      <th>recall_25</th>\n",
       "      <th>recall_5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>45</th>\n",
       "      <td>4.0000</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>384.0000</td>\n",
       "      <td>16.0000</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>0.7930</td>\n",
       "      <td>0.8848</td>\n",
       "      <td>0.8071</td>\n",
       "      <td>0.8711</td>\n",
       "      <td>0.4773</td>\n",
       "      <td>0.9087</td>\n",
       "      <td>0.9702</td>\n",
       "      <td>0.5732</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>101</th>\n",
       "      <td>4.0000</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>512.0000</td>\n",
       "      <td>16.0000</td>\n",
       "      <td>0.0000</td>\n",
       "      <td>0.7495</td>\n",
       "      <td>0.8657</td>\n",
       "      <td>0.7974</td>\n",
       "      <td>0.8706</td>\n",
       "      <td>0.4309</td>\n",
       "      <td>0.9214</td>\n",
       "      <td>0.9780</td>\n",
       "      <td>0.5308</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>56</th>\n",
       "      <td>4.0000</td>\n",
       "      <td>2.0000</td>\n",
       "      <td>256.0000</td>\n",
       "      <td>16.0000</td>\n",
       "      <td>0.0000</td>\n",
       "      <td>0.8267</td>\n",
       "      <td>0.9199</td>\n",
       "      <td>0.7642</td>\n",
       "      <td>0.8521</td>\n",
       "      <td>0.4172</td>\n",
       "      <td>0.8823</td>\n",
       "      <td>0.9673</td>\n",
       "      <td>0.5103</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>51</th>\n",
       "      <td>4.0000</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>256.0000</td>\n",
       "      <td>16.0000</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>0.7090</td>\n",
       "      <td>0.8794</td>\n",
       "      <td>0.7983</td>\n",
       "      <td>0.8750</td>\n",
       "      <td>0.4299</td>\n",
       "      <td>0.9048</td>\n",
       "      <td>0.9766</td>\n",
       "      <td>0.5347</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36</th>\n",
       "      <td>5.0000</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>384.0000</td>\n",
       "      <td>16.0000</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>0.7651</td>\n",
       "      <td>0.8804</td>\n",
       "      <td>0.7935</td>\n",
       "      <td>0.8745</td>\n",
       "      <td>0.4072</td>\n",
       "      <td>0.9009</td>\n",
       "      <td>0.9653</td>\n",
       "      <td>0.5073</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>43</th>\n",
       "      <td>4.0000</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>256.0000</td>\n",
       "      <td>16.0000</td>\n",
       "      <td>0.0000</td>\n",
       "      <td>0.7554</td>\n",
       "      <td>0.8652</td>\n",
       "      <td>0.7969</td>\n",
       "      <td>0.8647</td>\n",
       "      <td>0.4014</td>\n",
       "      <td>0.9116</td>\n",
       "      <td>0.9702</td>\n",
       "      <td>0.5103</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>198</th>\n",
       "      <td>4.0000</td>\n",
       "      <td>2.0000</td>\n",
       "      <td>256.0000</td>\n",
       "      <td>32.0000</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>0.7617</td>\n",
       "      <td>0.8784</td>\n",
       "      <td>0.7915</td>\n",
       "      <td>0.8696</td>\n",
       "      <td>0.3965</td>\n",
       "      <td>0.8926</td>\n",
       "      <td>0.9702</td>\n",
       "      <td>0.5073</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>59</th>\n",
       "      <td>4.0000</td>\n",
       "      <td>2.0000</td>\n",
       "      <td>128.0000</td>\n",
       "      <td>16.0000</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>0.7290</td>\n",
       "      <td>0.8770</td>\n",
       "      <td>0.7935</td>\n",
       "      <td>0.8745</td>\n",
       "      <td>0.4258</td>\n",
       "      <td>0.8711</td>\n",
       "      <td>0.9551</td>\n",
       "      <td>0.4988</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>165</th>\n",
       "      <td>4.0000</td>\n",
       "      <td>2.0000</td>\n",
       "      <td>128.0000</td>\n",
       "      <td>32.0000</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>0.7339</td>\n",
       "      <td>0.8638</td>\n",
       "      <td>0.7886</td>\n",
       "      <td>0.8765</td>\n",
       "      <td>0.4016</td>\n",
       "      <td>0.8823</td>\n",
       "      <td>0.9707</td>\n",
       "      <td>0.4900</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>4.0000</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>512.0000</td>\n",
       "      <td>16.0000</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>0.7085</td>\n",
       "      <td>0.8618</td>\n",
       "      <td>0.7739</td>\n",
       "      <td>0.8516</td>\n",
       "      <td>0.3945</td>\n",
       "      <td>0.9238</td>\n",
       "      <td>0.9712</td>\n",
       "      <td>0.5044</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     Last Convolution  # of GRUs  GRU size  Convolutional Filters  \\\n",
       "45             4.0000     1.0000  384.0000                16.0000   \n",
       "101            4.0000     1.0000  512.0000                16.0000   \n",
       "56             4.0000     2.0000  256.0000                16.0000   \n",
       "51             4.0000     1.0000  256.0000                16.0000   \n",
       "36             5.0000     1.0000  384.0000                16.0000   \n",
       "43             4.0000     1.0000  256.0000                16.0000   \n",
       "198            4.0000     2.0000  256.0000                32.0000   \n",
       "59             4.0000     2.0000  128.0000                16.0000   \n",
       "165            4.0000     2.0000  128.0000                32.0000   \n",
       "35             4.0000     1.0000  512.0000                16.0000   \n",
       "\n",
       "     First Convolution  class_precision  class_recall  prec_15  prec_25  \\\n",
       "45              1.0000           0.7930        0.8848   0.8071   0.8711   \n",
       "101             0.0000           0.7495        0.8657   0.7974   0.8706   \n",
       "56              0.0000           0.8267        0.9199   0.7642   0.8521   \n",
       "51              1.0000           0.7090        0.8794   0.7983   0.8750   \n",
       "36              1.0000           0.7651        0.8804   0.7935   0.8745   \n",
       "43              0.0000           0.7554        0.8652   0.7969   0.8647   \n",
       "198             1.0000           0.7617        0.8784   0.7915   0.8696   \n",
       "59              1.0000           0.7290        0.8770   0.7935   0.8745   \n",
       "165             1.0000           0.7339        0.8638   0.7886   0.8765   \n",
       "35              1.0000           0.7085        0.8618   0.7739   0.8516   \n",
       "\n",
       "     prec_5  recall_15  recall_25  recall_5  \n",
       "45   0.4773     0.9087     0.9702    0.5732  \n",
       "101  0.4309     0.9214     0.9780    0.5308  \n",
       "56   0.4172     0.8823     0.9673    0.5103  \n",
       "51   0.4299     0.9048     0.9766    0.5347  \n",
       "36   0.4072     0.9009     0.9653    0.5073  \n",
       "43   0.4014     0.9116     0.9702    0.5103  \n",
       "198  0.3965     0.8926     0.9702    0.5073  \n",
       "59   0.4258     0.8711     0.9551    0.4988  \n",
       "165  0.4016     0.8823     0.9707    0.4900  \n",
       "35   0.3945     0.9238     0.9712    0.5044  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df2['sum_measures'] = df2[m].sum(axis=1)\n",
    "df2 = df2.sort_values('sum_measures', ascending=False)\n",
    "with pandas_format.PandasFloatFormatter('{:,.4f}'):\n",
    "    display(df2[h.append(m)].iloc[:10])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['class_precision', 'class_recall', 'prec_15', 'prec_25', 'prec_5',\n",
       "       'recall_15', 'recall_25', 'recall_5'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "m"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}