01_gridsearch_baseline_bn.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "01_gridsearch_baseline_bn.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true,
      "machine_shape": "hm",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/CrowdSalat/ann-dynamical-systems/blob/master/01_gridsearch_baseline_bn.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Pa9Mju9lc4Ll",
        "colab_type": "text"
      },
      "source": [
        "# overview\n",
        "\n",
        "Uses actions and an input state to predict the output state."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ijDD21X-gnli",
        "colab_type": "text"
      },
      "source": [
        "# imports"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_Z4pV678a1zv",
        "colab_type": "code",
        "outputId": "9d838c25-d4cd-48f0-bc48-9d0917f59aaf",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "try:\n",
        "  # %tensorflow_version only exists in Colab.\n",
        "  %tensorflow_version 2.x\n",
        "except Exception:\n",
        "  pass"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "TensorFlow 2.x selected.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2p4YwEK_f7kq",
        "colab_type": "code",
        "outputId": "65523e53-85d3-45af-df15-a372540190e3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "import pandas as pd\n",
        "import numpy as np\n",
        "import os\n",
        "import sys\n",
        "import datetime\n",
        "import math\n",
        "import pickle\n",
        "import matplotlib.pyplot as plt\n",
        "from sklearn.preprocessing import MinMaxScaler\n",
        "from sklearn.metrics import mean_squared_error\n",
        "from sklearn.metrics import mean_absolute_error\n",
        "\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.preprocessing import MinMaxScaler\n",
        "\n",
        "\n",
        "import tensorflow as tf\n",
        "print(tf.__version__)\n",
        "from tensorflow import keras\n",
        "\n",
        "from tensorflow.keras.models import Sequential\n",
        "from tensorflow.keras.layers import Dense\n",
        "from tensorflow.keras.layers import *\n",
        "from tensorflow.keras import metrics"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "2.0.0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RPDma-JqgyaM",
        "colab_type": "code",
        "outputId": "c6b950b0-0748-46e6-d1e9-621db88b8279",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "base_folder = './results'\n",
        "notebook_folder = '01_gridsearch_baseline_bn'\n",
        "results_folder = os.path.join(base_folder, notebook_folder)\n",
        "os.makedirs(results_folder, exist_ok=True)\n",
        "print('results_folder : ', results_folder)\n",
        "\n",
        "saved_model_weight = 'saved_weights'\n",
        "saved_model_weight_path = os.path.join(results_folder, saved_model_weight)\n",
        "os.makedirs(saved_model_weight_path, exist_ok=True)\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "results_folder :  ./results/01_gridsearch_baseline_bn\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rmj8c18xllpT",
        "colab_type": "text"
      },
      "source": [
        "# mount drive"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nUEqcpzvltdo",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "file = 'dataset_all_01.csv'"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TuGWEMS1loFz",
        "colab_type": "code",
        "outputId": "128d3972-59df-422f-ce37-0ae02ffefa67",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 53
        }
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')\n",
        "\n",
        "root_dir = '/content/drive/My Drive/xplane_data' \n",
        "os.chdir(root_dir)\n",
        "print ('working dir: ' + os.getcwd())"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n",
            "working dir: /content/drive/My Drive/xplane_data\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EF4tx110HmOj",
        "colab_type": "text"
      },
      "source": [
        "# load data"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EOM6RVAIduGr",
        "colab_type": "text"
      },
      "source": [
        "## define features (actions & states)\n",
        "actions and states at time t (t-1, t-2 ...) are the input for the network.\n",
        "\n",
        "the states at time t + 1 are the desired output of the network.\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YH9bVQ3YDyTh",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "col_m_time      = '   missn,_time '\n",
        "col_t_time      = '   _totl,_time '\n",
        "col_yaw_v       = '   ____Q,rad/s '\n",
        "\n",
        "col_yaw_deg     = '   _beta,__deg '\n",
        "col_roll_deg    = '   _roll,__deg '\n",
        "col_roll_v      = '   ____R,rad/s '\n",
        "col_pitch_deg   = '   pitch,__deg '\n",
        "col_pitch_v     = '   ____P,rad/s '\n",
        "col_alt         = '   __alt,ftmsl '\n",
        "col_vv          = '   __VVI,__fpm '\n",
        "col_v           = '   Vtrue,_ktas '\n",
        "col_rpm_prop    = '   rpm_1,_prop '\n",
        "col_throttle     = '   thro1,_part '\n",
        "\n",
        "cols_state = [ \n",
        "    col_yaw_v,\n",
        "    col_yaw_deg,\n",
        "    col_roll_deg,\n",
        "    col_roll_v,\n",
        "    col_pitch_deg,\n",
        "    col_pitch_v,\n",
        "    col_vv,\n",
        "    col_alt,\n",
        "    col_v,\n",
        "\n",
        "]\n",
        "\n",
        "cols_time = [\n",
        "    col_m_time,\n",
        "    col_t_time,\n",
        "]\n",
        "\n",
        "col_elev_servo                  = '   _elev,servo '\n",
        "col_ailrn_servo                  = '   ailrn,servo '\n",
        "col_ruddr_servo                  = '   ruddr,servo '\n",
        "\n",
        "cols_action = [\n",
        "    col_elev_servo,\n",
        "    col_ailrn_servo,\n",
        "    col_ruddr_servo,\n",
        "    col_rpm_prop,\n",
        "    col_throttle,\n",
        "]\n",
        "\n",
        "# all input columns\n",
        "cols_features = []\n",
        "cols_features.extend(cols_state)\n",
        "cols_features.extend(cols_action)\n",
        "\n",
        "# orga column\n",
        "col_dataset_index = 'dataset_index'\n",
        "col_maneuver_index = 'maneuver_index'\n",
        "col_maneuver_name = 'maneuver_name'\n",
        "col_maneuver_datapoint_index = 'maneuver_datapoint_index'\n",
        "col_maneuver_type = 'maneuver_type'\n",
        "cols_org = [col_dataset_index, col_maneuver_index, col_maneuver_name,\n",
        "            col_maneuver_datapoint_index,col_maneuver_type]\n",
        "\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gDSNkVBaDv2V",
        "colab_type": "text"
      },
      "source": [
        "## load data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uByPEHX3fS2G",
        "colab_type": "code",
        "outputId": "dd61080c-c901-4f92-aba0-3458fd9f471b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        }
      },
      "source": [
        "def load_xplane_csv(csv_path, columns):\n",
        "    return pd.read_csv(csv_path, delimiter=',', usecols = columns, \n",
        "                       float_precision='high')\n",
        "\n",
        "df_data = load_xplane_csv(file, cols_org + cols_state + cols_action)\n",
        "\n",
        "print(len(cols_org))\n",
        "print(len(cols_state))\n",
        "print(len(cols_action))\n",
        "print(df_data.shape)\n",
        "df_data.head(2)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "5\n",
            "9\n",
            "5\n",
            "(477893, 19)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "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>dataset_index</th>\n",
              "      <th>maneuver_index</th>\n",
              "      <th>maneuver_name</th>\n",
              "      <th>maneuver_type</th>\n",
              "      <th>maneuver_datapoint_index</th>\n",
              "      <th>Vtrue,_ktas</th>\n",
              "      <th>__VVI,__fpm</th>\n",
              "      <th>____Q,rad/s</th>\n",
              "      <th>____P,rad/s</th>\n",
              "      <th>____R,rad/s</th>\n",
              "      <th>pitch,__deg</th>\n",
              "      <th>_roll,__deg</th>\n",
              "      <th>_beta,__deg</th>\n",
              "      <th>__alt,ftmsl</th>\n",
              "      <th>thro1,_part</th>\n",
              "      <th>rpm_1,_prop</th>\n",
              "      <th>_elev,servo</th>\n",
              "      <th>ailrn,servo</th>\n",
              "      <th>ruddr,servo</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>115.86147</td>\n",
              "      <td>665.39697</td>\n",
              "      <td>-0.03945</td>\n",
              "      <td>-0.00235</td>\n",
              "      <td>0.00008</td>\n",
              "      <td>-0.03584</td>\n",
              "      <td>-0.10129</td>\n",
              "      <td>0.00793</td>\n",
              "      <td>953.06787</td>\n",
              "      <td>0.79608</td>\n",
              "      <td>2443.89453</td>\n",
              "      <td>-0.00489</td>\n",
              "      <td>-0.01818</td>\n",
              "      <td>0.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>115.85403</td>\n",
              "      <td>650.10065</td>\n",
              "      <td>-0.03868</td>\n",
              "      <td>-0.00234</td>\n",
              "      <td>0.00007</td>\n",
              "      <td>-0.14285</td>\n",
              "      <td>-0.10772</td>\n",
              "      <td>0.00770</td>\n",
              "      <td>953.30768</td>\n",
              "      <td>0.79608</td>\n",
              "      <td>2443.77563</td>\n",
              "      <td>0.00134</td>\n",
              "      <td>-0.01815</td>\n",
              "      <td>0.0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "   dataset_index  maneuver_index  ...     ailrn,servo      ruddr,servo \n",
              "0              0               0  ...         -0.01818              0.0\n",
              "1              0               0  ...         -0.01815              0.0\n",
              "\n",
              "[2 rows x 19 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "47xDI8rfLJTv",
        "colab_type": "text"
      },
      "source": [
        "# define data preperation functions"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-EBCtY9f7KnC",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def series_to_supervised(data, n_in=1, n_out=1, dropnan=True, cols_label=None):\n",
        "\tdata_in = data.drop(cols_org, axis=1)\n",
        "\tif cols_label == None:\n",
        "\t\tdata_out = data_in.drop(cols_action, axis=1)\n",
        "\telse:\n",
        "\t\tif type(cols_label) != list:\n",
        "\t\t\tcols_label = [cols_label]\n",
        "\t\tdata_out = data_in[cols_label]\n",
        "\n",
        "\tcols_in = data_in.columns\n",
        "\tcols_out = data_out.columns\n",
        "\tcols, names = list(), list()\n",
        "\t# input sequence (t-n, ... t-1, t)\n",
        "\tfor i in range(n_in, 0, -1):\n",
        "\t\tcols.append(data_in.shift(i))\n",
        "\t\tif i == 1:\n",
        "\t\t\tnames += [('%s(t)' % (col_in)) for col_in in cols_in]\n",
        "\t\telse:\n",
        "\t\t\tnames += [('%s(t-%d)' % (col_in, i-1)) for col_in in cols_in]\n",
        "\n",
        "\t# forecast sequence (t+1, ... t+n)\n",
        "\tfor i in range(0, n_out):\n",
        "\t\tcols.append(data_out.shift(-i))\n",
        "\t\tnames += [('%s(t+%d)' % (col_out, i+1)) for col_out in cols_out]\n",
        "\t# put it all together\n",
        "\tagg = pd.concat(cols, axis=1)\n",
        "\tagg.columns = names\n",
        "\t# drop rows with NaN values\n",
        "\tif dropnan:\n",
        "\t\tagg.dropna(inplace=True)\n",
        "\treturn agg, cols_out"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nOzBV1rYVGCn",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def split_feature_label(data_labeled, n_input_steps):\n",
        "  data_X = data_labeled.values[:, 0:n_input_steps*len(cols_features)]\n",
        "  data_Y = data_labeled.values[:, n_input_steps*len(cols_features):]\n",
        "  return data_X, data_Y"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "otFsVggWMbpz",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def scale_data(data_X, data_Y):\n",
        "  feature_scaler = MinMaxScaler()\n",
        "  data_X = feature_scaler.fit_transform(data_X)\n",
        "  label_scaler = MinMaxScaler()\n",
        "  data_Y = label_scaler.fit_transform(data_Y)\n",
        "  return feature_scaler, label_scaler, data_X, data_Y"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bYXja56FWJnk",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def prepare_datasets(data, n_in=1, n_out=1):\n",
        "  #train-test-split\n",
        "  test = data.query('dataset_index==4')\n",
        "  train = data.query('dataset_index!=4')\n",
        "  \n",
        "  #add-label\n",
        "  test_labeled, labels = series_to_supervised(test, n_in=n_in, n_out=n_out,dropnan=True)\n",
        "  train_labeled, labels = series_to_supervised(train, n_in=n_in, n_out=n_out,dropnan=True)\n",
        "  \n",
        "\n",
        "  #split-feature-label\n",
        "  train_X, train_Y = split_feature_label(train_labeled, n_in)\n",
        "  test_X, test_Y = split_feature_label(test_labeled, n_in)\n",
        "  #scale-feat\n",
        "  feature_scaler, label_scaler, train_X, train_Y = scale_data(train_X, train_Y)\n",
        "  #scale-label\n",
        "  test_X, test_Y = feature_scaler.transform(test_X), label_scaler.transform(test_Y)\n",
        "\n",
        "  return train_X, train_Y, test_X, test_Y, feature_scaler, label_scaler, labels\n",
        "\n",
        "\n",
        "#TODO für RNN muss noch reshape stattfinden"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "S_kDrS1qH7kc",
        "colab_type": "text"
      },
      "source": [
        "# train models"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rjgpogwABFhX",
        "colab_type": "text"
      },
      "source": [
        "## define callbacks functions"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "18w_KQkpBFEY",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def prepare_tensorboard_callback(model_name=''):\n",
        "  logs_base_dir = \"./tensorbaord_logs\"\n",
        "  model_log_dir = datetime.datetime.now().strftime(\"%Y%m%d-%H%M%S\") + '_' + model_name\n",
        "  target_log_dir = os.path.join(logs_base_dir, model_log_dir)\n",
        "  os.makedirs(target_log_dir, exist_ok=True)\n",
        "  tensorboard_callback = tf.keras.callbacks.TensorBoard(target_log_dir, histogram_freq=1)\n",
        "  return tensorboard_callback\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MYl_6s7sBMpy",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def prepare_checkpoint_callback(model_name):\n",
        "  checkpoint_base_dir = './checkpoints'\n",
        "  os.makedirs(checkpoint_base_dir, exist_ok=True)\n",
        "  filepath = os.path.join(checkpoint_base_dir, 'cp_'+ model_name)\n",
        "  cp_callback = tf.keras.callbacks.ModelCheckpoint(filepath=filepath,\n",
        "                                                  save_weights_only=True,\n",
        "                                                  verbose=1)\n",
        "  return cp_callback "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3y07Q50xkLvG",
        "colab_type": "text"
      },
      "source": [
        "## define plot_and_save functions"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "l8uMddMPkSc1",
        "colab_type": "text"
      },
      "source": [
        "### loss historie"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "I9UHTxwnkWyp",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def save_plots_before_show(diagramm_name):\n",
        "  if not os.path.exists(results_folder):\n",
        "      os.makedirs(diagramm_folder)\n",
        "\n",
        "  plt.savefig(results_folder + '/' + diagramm_name +'.pdf')\n",
        "  plt.savefig(results_folder + '/' + diagramm_name +'.svg')\n",
        "  plt.savefig(results_folder + '/' + diagramm_name +'.png')\n",
        "\n",
        "def pas_history(model_tupels, histories, key='mean_squared_error'):\n",
        "  model_names = [model_tupel[0] for model_tupel in model_tupels]\n",
        "  plt.figure(figsize=(16,10))\n",
        "  for i, history in enumerate(histories):\n",
        "    history_epoch= range(0, len(history['loss']) )\n",
        "\n",
        "    val = plt.plot(history_epoch, history['val_'+key],\n",
        "                   '--', label=model_names[i]+' Val')\n",
        "    plt.plot(history_epoch, history[key], color=val[0].get_color(),\n",
        "             label=model_names[i]+' Train')\n",
        "\n",
        "  plt.xlabel('Epochs')\n",
        "  plt.yscale(\"log\")\n",
        "  plt.ylabel(key.replace('_',' ').title())\n",
        "  plt.legend()\n",
        "\n",
        "  plt.xlim([0,max(history_epoch)])\n",
        "  save_plots_before_show('history_loss'+'_'+key)\n",
        "  plt.show()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VJNfegiglg59",
        "colab_type": "text"
      },
      "source": [
        "### metrics (overall and for all single targets )\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "be-UQiKHli16",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def pas_metrics_scaled(model_tupel):\n",
        "  model_name = model_tupel[0]\n",
        "  df = _df_prediction_metrics(model_tupel, scaled=True)\n",
        "  file_path = os.path.join(results_folder, model_name + '_scaled.csv')\n",
        "  df.to_csv(path_or_buf=file_path)\n",
        "  return df\n",
        "  \n",
        "def pas_metrics(model_tupel):\n",
        "  model_name = model_tupel[0]\n",
        "  df = _df_prediction_metrics(model_tupel, scaled=False)\n",
        "  file_path = os.path.join(results_folder, model_name + '.csv')\n",
        "  df.to_csv(path_or_buf=file_path)\n",
        "  return df\n",
        "\n",
        "\n",
        "def _df_prediction_metrics(model_tupel, scaled):\n",
        "  '''\n",
        "  creates describe dataframe of labels and \n",
        "  connect them with mse,rmse and mae per target and overall \n",
        "  '''\n",
        "  \n",
        "  model = model_tupel[1]\n",
        "  train_X, train_Y, test_X, test_Y = model_tupel[2]\n",
        "  cols_label = model_tupel[4]\n",
        "\n",
        "\n",
        "  pred_Y = model.predict(test_X)\n",
        "  \n",
        "  if scaled:\n",
        "    truth_Y = test_Y\n",
        "  else:\n",
        "    label_scaler = model_tupel[3]\n",
        "    pred_Y = label_scaler.inverse_transform(pred_Y)\n",
        "    truth_Y = label_scaler.inverse_transform(test_Y)\n",
        "\n",
        "  df_describe = pd.DataFrame(truth_Y,columns=cols_label).describe()\n",
        "  mses = mean_squared_error(truth_Y, pred_Y, multioutput='raw_values')\n",
        "  rmses = np.array([math.sqrt(mse) for mse in mses])\n",
        "  maes = mean_absolute_error(truth_Y, pred_Y, multioutput='raw_values')\n",
        "  # average over list should do the job as well\n",
        "  mse = mean_squared_error(truth_Y, pred_Y)\n",
        "  rmse = math.sqrt(mse)\n",
        "  mae = mean_absolute_error(truth_Y, pred_Y)\n",
        "\n",
        "\n",
        "  df_mse_per_feature = pd.DataFrame(data=mses.reshape(1,-1),    # values\n",
        "            index=['mses'],     # index\n",
        "            columns=cols_label)  #  column names\n",
        "  df_rmse_per_feature = pd.DataFrame(data=rmses.reshape(1,-1),    # values\n",
        "        index=['rmes'],     # index\n",
        "        columns=cols_label)  #  column names\n",
        "  df_mae_per_feature = pd.DataFrame(data=maes.reshape(1,-1),    # values\n",
        "        index=['maes'],     # index\n",
        "        columns=cols_label)  #  column names\n",
        "  \n",
        "  \n",
        "  df_mse_all = pd.DataFrame(data=mse,    # values\n",
        "            index=['mse'],     # index\n",
        "            columns=cols_label)  #  column names\n",
        "  df_rmse_all = pd.DataFrame(data=rmse,    # values\n",
        "        index=['rme'],     # index\n",
        "        columns=cols_label)  #  column names\n",
        "  df_mae_all = pd.DataFrame(data=mae,    # values\n",
        "        index=['mae'],     # index\n",
        "        columns=cols_label)  #  column names\n",
        "  \n",
        "\n",
        "  metrics = [df_mse_per_feature, df_rmse_per_feature, df_mae_per_feature, \n",
        "             df_mse_all,df_rmse_all,df_mae_all]\n",
        "  return df_describe.append(metrics)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4gaR7IV326E_",
        "colab_type": "text"
      },
      "source": [
        "## define models"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Fhc2M4VD2_YP",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def create_model(model_name, layers, keras_activation, kernel_initializer, \n",
        "                 loss, metrics,  n_in=1, n_out=1):\n",
        "  if not layers:\n",
        "    raise Exception('No layers defined. Expected in format [100,100]') \n",
        "\n",
        " #prepare data\n",
        "  train_X, train_Y, test_X, test_Y, feature_scaler, label_scaler, labels = prepare_datasets(df_data, n_in, n_out)\n",
        "  features_nr = train_X.shape[1]\n",
        "  target_nr = train_Y.shape[1]\n",
        "    \n",
        "  model = keras.Sequential(name=model_name)\n",
        "  \n",
        "  model.add(Dense(layers[0],input_shape=(features_nr,), \n",
        "                          kernel_initializer=kernel_initializer))\n",
        "  model.add(BatchNormalization())\n",
        "  model.add(keras_activation())\n",
        "  for i in range(1, len(layers)):\n",
        "    layer_val = layers[i]\n",
        "    model.add(Dense(layer_val, kernel_initializer=kernel_initializer))\n",
        "    model.add(BatchNormalization())\n",
        "    model.add(keras_activation())\n",
        "\n",
        "  model.add(Dense(target_nr ))\n",
        "\n",
        "  model.compile(optimizer='adam',\n",
        "                        loss=loss,\n",
        "                        metrics=metrics)\n",
        "\n",
        "  model.summary()\n",
        "  print(model.to_json())\n",
        "  print(train_X.shape, train_Y.shape, test_X.shape, test_Y.shape)\n",
        "  return (model_name, model, (train_X, train_Y, test_X, test_Y), label_scaler, labels)\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KwMj6E7mBL9m",
        "colab_type": "code",
        "outputId": "d2fa5ff0-1aad-4fe4-e84e-5b23ee0a084b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "loss='mean_squared_error',\n",
        "metrics=['mean_squared_error', 'mean_absolute_error']\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "baseline = create_model('baseline', [1200, 600, 600], ReLU, 'he_uniform', \n",
        "             loss, metrics, 1,1)\n",
        "\n",
        "\n",
        "bigger = create_model('bigger', [1200, 1000, 800], ReLU, 'he_uniform', \n",
        "             loss, metrics, 1,1)\n",
        "\n",
        "\n",
        "smaller = create_model('smaller', [600, 400, 400], ReLU, 'he_uniform', \n",
        "             loss, metrics, 1,1)\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "models = [baseline,\n",
        "          bigger,\n",
        "          smaller \n",
        "          ]"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"baseline\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "dense (Dense)                (None, 1200)              18000     \n",
            "_________________________________________________________________\n",
            "batch_normalization (BatchNo (None, 1200)              4800      \n",
            "_________________________________________________________________\n",
            "re_lu (ReLU)                 (None, 1200)              0         \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 600)               720600    \n",
            "_________________________________________________________________\n",
            "batch_normalization_1 (Batch (None, 600)               2400      \n",
            "_________________________________________________________________\n",
            "re_lu_1 (ReLU)               (None, 600)               0         \n",
            "_________________________________________________________________\n",
            "dense_2 (Dense)              (None, 600)               360600    \n",
            "_________________________________________________________________\n",
            "batch_normalization_2 (Batch (None, 600)               2400      \n",
            "_________________________________________________________________\n",
            "re_lu_2 (ReLU)               (None, 600)               0         \n",
            "_________________________________________________________________\n",
            "dense_3 (Dense)              (None, 9)                 5409      \n",
            "=================================================================\n",
            "Total params: 1,114,209\n",
            "Trainable params: 1,109,409\n",
            "Non-trainable params: 4,800\n",
            "_________________________________________________________________\n",
            "{\"class_name\": \"Sequential\", \"config\": {\"name\": \"baseline\", \"layers\": [{\"class_name\": \"Dense\", \"config\": {\"name\": \"dense\", \"trainable\": true, \"batch_input_shape\": [null, 14], \"dtype\": \"float32\", \"units\": 1200, \"activation\": \"linear\", \"use_bias\": true, \"kernel_initializer\": {\"class_name\": \"VarianceScaling\", \"config\": {\"scale\": 2.0, \"mode\": \"fan_in\", \"distribution\": \"uniform\", \"seed\": null}}, \"bias_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"kernel_regularizer\": null, \"bias_regularizer\": null, \"activity_regularizer\": null, \"kernel_constraint\": null, \"bias_constraint\": null}}, {\"class_name\": \"BatchNormalization\", \"config\": {\"name\": \"batch_normalization\", \"trainable\": true, \"dtype\": \"float32\", \"axis\": [1], \"momentum\": 0.99, \"epsilon\": 0.001, \"center\": true, \"scale\": true, \"beta_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"gamma_initializer\": {\"class_name\": \"Ones\", \"config\": {}}, \"moving_mean_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"moving_variance_initializer\": {\"class_name\": \"Ones\", \"config\": {}}, \"beta_regularizer\": null, \"gamma_regularizer\": null, \"beta_constraint\": null, \"gamma_constraint\": null}}, {\"class_name\": \"ReLU\", \"config\": {\"name\": \"re_lu\", \"trainable\": true, \"dtype\": \"float32\", \"max_value\": null, \"negative_slope\": 0.0, \"threshold\": 0.0}}, {\"class_name\": \"Dense\", \"config\": {\"name\": \"dense_1\", \"trainable\": true, \"dtype\": \"float32\", \"units\": 600, \"activation\": \"linear\", \"use_bias\": true, \"kernel_initializer\": {\"class_name\": \"VarianceScaling\", \"config\": {\"scale\": 2.0, \"mode\": \"fan_in\", \"distribution\": \"uniform\", \"seed\": null}}, \"bias_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"kernel_regularizer\": null, \"bias_regularizer\": null, \"activity_regularizer\": null, \"kernel_constraint\": null, \"bias_constraint\": null}}, {\"class_name\": \"BatchNormalization\", \"config\": {\"name\": \"batch_normalization_1\", \"trainable\": true, \"dtype\": \"float32\", \"axis\": [1], \"momentum\": 0.99, \"epsilon\": 0.001, \"center\": true, \"scale\": true, \"beta_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"gamma_initializer\": {\"class_name\": \"Ones\", \"config\": {}}, \"moving_mean_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"moving_variance_initializer\": {\"class_name\": \"Ones\", \"config\": {}}, \"beta_regularizer\": null, \"gamma_regularizer\": null, \"beta_constraint\": null, \"gamma_constraint\": null}}, {\"class_name\": \"ReLU\", \"config\": {\"name\": \"re_lu_1\", \"trainable\": true, \"dtype\": \"float32\", \"max_value\": null, \"negative_slope\": 0.0, \"threshold\": 0.0}}, {\"class_name\": \"Dense\", \"config\": {\"name\": \"dense_2\", \"trainable\": true, \"dtype\": \"float32\", \"units\": 600, \"activation\": \"linear\", \"use_bias\": true, \"kernel_initializer\": {\"class_name\": \"VarianceScaling\", \"config\": {\"scale\": 2.0, \"mode\": \"fan_in\", \"distribution\": \"uniform\", \"seed\": null}}, \"bias_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"kernel_regularizer\": null, \"bias_regularizer\": null, \"activity_regularizer\": null, \"kernel_constraint\": null, \"bias_constraint\": null}}, {\"class_name\": \"BatchNormalization\", \"config\": {\"name\": \"batch_normalization_2\", \"trainable\": true, \"dtype\": \"float32\", \"axis\": [1], \"momentum\": 0.99, \"epsilon\": 0.001, \"center\": true, \"scale\": true, \"beta_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"gamma_initializer\": {\"class_name\": \"Ones\", \"config\": {}}, \"moving_mean_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"moving_variance_initializer\": {\"class_name\": \"Ones\", \"config\": {}}, \"beta_regularizer\": null, \"gamma_regularizer\": null, \"beta_constraint\": null, \"gamma_constraint\": null}}, {\"class_name\": \"ReLU\", \"config\": {\"name\": \"re_lu_2\", \"trainable\": true, \"dtype\": \"float32\", \"max_value\": null, \"negative_slope\": 0.0, \"threshold\": 0.0}}, {\"class_name\": \"Dense\", \"config\": {\"name\": \"dense_3\", \"trainable\": true, \"dtype\": \"float32\", \"units\": 9, \"activation\": \"linear\", \"use_bias\": true, \"kernel_initializer\": {\"class_name\": \"GlorotUniform\", \"config\": {\"seed\": null}}, \"bias_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"kernel_regularizer\": null, \"bias_regularizer\": null, \"activity_regularizer\": null, \"kernel_constraint\": null, \"bias_constraint\": null}}]}, \"keras_version\": \"2.2.4-tf\", \"backend\": \"tensorflow\"}\n",
            "(330984, 14) (330984, 9) (146907, 14) (146907, 9)\n",
            "Model: \"bigger\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "dense_4 (Dense)              (None, 1200)              18000     \n",
            "_________________________________________________________________\n",
            "batch_normalization_3 (Batch (None, 1200)              4800      \n",
            "_________________________________________________________________\n",
            "re_lu_3 (ReLU)               (None, 1200)              0         \n",
            "_________________________________________________________________\n",
            "dense_5 (Dense)              (None, 1000)              1201000   \n",
            "_________________________________________________________________\n",
            "batch_normalization_4 (Batch (None, 1000)              4000      \n",
            "_________________________________________________________________\n",
            "re_lu_4 (ReLU)               (None, 1000)              0         \n",
            "_________________________________________________________________\n",
            "dense_6 (Dense)              (None, 800)               800800    \n",
            "_________________________________________________________________\n",
            "batch_normalization_5 (Batch (None, 800)               3200      \n",
            "_________________________________________________________________\n",
            "re_lu_5 (ReLU)               (None, 800)               0         \n",
            "_________________________________________________________________\n",
            "dense_7 (Dense)              (None, 9)                 7209      \n",
            "=================================================================\n",
            "Total params: 2,039,009\n",
            "Trainable params: 2,033,009\n",
            "Non-trainable params: 6,000\n",
            "_________________________________________________________________\n",
            "{\"class_name\": \"Sequential\", \"config\": {\"name\": \"bigger\", \"layers\": [{\"class_name\": \"Dense\", \"config\": {\"name\": \"dense_4\", \"trainable\": true, \"batch_input_shape\": [null, 14], \"dtype\": \"float32\", \"units\": 1200, \"activation\": \"linear\", \"use_bias\": true, \"kernel_initializer\": {\"class_name\": \"VarianceScaling\", \"config\": {\"scale\": 2.0, \"mode\": \"fan_in\", \"distribution\": \"uniform\", \"seed\": null}}, \"bias_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"kernel_regularizer\": null, \"bias_regularizer\": null, \"activity_regularizer\": null, \"kernel_constraint\": null, \"bias_constraint\": null}}, {\"class_name\": \"BatchNormalization\", \"config\": {\"name\": \"batch_normalization_3\", \"trainable\": true, \"dtype\": \"float32\", \"axis\": [1], \"momentum\": 0.99, \"epsilon\": 0.001, \"center\": true, \"scale\": true, \"beta_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"gamma_initializer\": {\"class_name\": \"Ones\", \"config\": {}}, \"moving_mean_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"moving_variance_initializer\": {\"class_name\": \"Ones\", \"config\": {}}, \"beta_regularizer\": null, \"gamma_regularizer\": null, \"beta_constraint\": null, \"gamma_constraint\": null}}, {\"class_name\": \"ReLU\", \"config\": {\"name\": \"re_lu_3\", \"trainable\": true, \"dtype\": \"float32\", \"max_value\": null, \"negative_slope\": 0.0, \"threshold\": 0.0}}, {\"class_name\": \"Dense\", \"config\": {\"name\": \"dense_5\", \"trainable\": true, \"dtype\": \"float32\", \"units\": 1000, \"activation\": \"linear\", \"use_bias\": true, \"kernel_initializer\": {\"class_name\": \"VarianceScaling\", \"config\": {\"scale\": 2.0, \"mode\": \"fan_in\", \"distribution\": \"uniform\", \"seed\": null}}, \"bias_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"kernel_regularizer\": null, \"bias_regularizer\": null, \"activity_regularizer\": null, \"kernel_constraint\": null, \"bias_constraint\": null}}, {\"class_name\": \"BatchNormalization\", \"config\": {\"name\": \"batch_normalization_4\", \"trainable\": true, \"dtype\": \"float32\", \"axis\": [1], \"momentum\": 0.99, \"epsilon\": 0.001, \"center\": true, \"scale\": true, \"beta_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"gamma_initializer\": {\"class_name\": \"Ones\", \"config\": {}}, \"moving_mean_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"moving_variance_initializer\": {\"class_name\": \"Ones\", \"config\": {}}, \"beta_regularizer\": null, \"gamma_regularizer\": null, \"beta_constraint\": null, \"gamma_constraint\": null}}, {\"class_name\": \"ReLU\", \"config\": {\"name\": \"re_lu_4\", \"trainable\": true, \"dtype\": \"float32\", \"max_value\": null, \"negative_slope\": 0.0, \"threshold\": 0.0}}, {\"class_name\": \"Dense\", \"config\": {\"name\": \"dense_6\", \"trainable\": true, \"dtype\": \"float32\", \"units\": 800, \"activation\": \"linear\", \"use_bias\": true, \"kernel_initializer\": {\"class_name\": \"VarianceScaling\", \"config\": {\"scale\": 2.0, \"mode\": \"fan_in\", \"distribution\": \"uniform\", \"seed\": null}}, \"bias_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"kernel_regularizer\": null, \"bias_regularizer\": null, \"activity_regularizer\": null, \"kernel_constraint\": null, \"bias_constraint\": null}}, {\"class_name\": \"BatchNormalization\", \"config\": {\"name\": \"batch_normalization_5\", \"trainable\": true, \"dtype\": \"float32\", \"axis\": [1], \"momentum\": 0.99, \"epsilon\": 0.001, \"center\": true, \"scale\": true, \"beta_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"gamma_initializer\": {\"class_name\": \"Ones\", \"config\": {}}, \"moving_mean_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"moving_variance_initializer\": {\"class_name\": \"Ones\", \"config\": {}}, \"beta_regularizer\": null, \"gamma_regularizer\": null, \"beta_constraint\": null, \"gamma_constraint\": null}}, {\"class_name\": \"ReLU\", \"config\": {\"name\": \"re_lu_5\", \"trainable\": true, \"dtype\": \"float32\", \"max_value\": null, \"negative_slope\": 0.0, \"threshold\": 0.0}}, {\"class_name\": \"Dense\", \"config\": {\"name\": \"dense_7\", \"trainable\": true, \"dtype\": \"float32\", \"units\": 9, \"activation\": \"linear\", \"use_bias\": true, \"kernel_initializer\": {\"class_name\": \"GlorotUniform\", \"config\": {\"seed\": null}}, \"bias_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"kernel_regularizer\": null, \"bias_regularizer\": null, \"activity_regularizer\": null, \"kernel_constraint\": null, \"bias_constraint\": null}}]}, \"keras_version\": \"2.2.4-tf\", \"backend\": \"tensorflow\"}\n",
            "(330984, 14) (330984, 9) (146907, 14) (146907, 9)\n",
            "Model: \"smaller\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "dense_8 (Dense)              (None, 600)               9000      \n",
            "_________________________________________________________________\n",
            "batch_normalization_6 (Batch (None, 600)               2400      \n",
            "_________________________________________________________________\n",
            "re_lu_6 (ReLU)               (None, 600)               0         \n",
            "_________________________________________________________________\n",
            "dense_9 (Dense)              (None, 400)               240400    \n",
            "_________________________________________________________________\n",
            "batch_normalization_7 (Batch (None, 400)               1600      \n",
            "_________________________________________________________________\n",
            "re_lu_7 (ReLU)               (None, 400)               0         \n",
            "_________________________________________________________________\n",
            "dense_10 (Dense)             (None, 400)               160400    \n",
            "_________________________________________________________________\n",
            "batch_normalization_8 (Batch (None, 400)               1600      \n",
            "_________________________________________________________________\n",
            "re_lu_8 (ReLU)               (None, 400)               0         \n",
            "_________________________________________________________________\n",
            "dense_11 (Dense)             (None, 9)                 3609      \n",
            "=================================================================\n",
            "Total params: 419,009\n",
            "Trainable params: 416,209\n",
            "Non-trainable params: 2,800\n",
            "_________________________________________________________________\n",
            "{\"class_name\": \"Sequential\", \"config\": {\"name\": \"smaller\", \"layers\": [{\"class_name\": \"Dense\", \"config\": {\"name\": \"dense_8\", \"trainable\": true, \"batch_input_shape\": [null, 14], \"dtype\": \"float32\", \"units\": 600, \"activation\": \"linear\", \"use_bias\": true, \"kernel_initializer\": {\"class_name\": \"VarianceScaling\", \"config\": {\"scale\": 2.0, \"mode\": \"fan_in\", \"distribution\": \"uniform\", \"seed\": null}}, \"bias_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"kernel_regularizer\": null, \"bias_regularizer\": null, \"activity_regularizer\": null, \"kernel_constraint\": null, \"bias_constraint\": null}}, {\"class_name\": \"BatchNormalization\", \"config\": {\"name\": \"batch_normalization_6\", \"trainable\": true, \"dtype\": \"float32\", \"axis\": [1], \"momentum\": 0.99, \"epsilon\": 0.001, \"center\": true, \"scale\": true, \"beta_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"gamma_initializer\": {\"class_name\": \"Ones\", \"config\": {}}, \"moving_mean_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"moving_variance_initializer\": {\"class_name\": \"Ones\", \"config\": {}}, \"beta_regularizer\": null, \"gamma_regularizer\": null, \"beta_constraint\": null, \"gamma_constraint\": null}}, {\"class_name\": \"ReLU\", \"config\": {\"name\": \"re_lu_6\", \"trainable\": true, \"dtype\": \"float32\", \"max_value\": null, \"negative_slope\": 0.0, \"threshold\": 0.0}}, {\"class_name\": \"Dense\", \"config\": {\"name\": \"dense_9\", \"trainable\": true, \"dtype\": \"float32\", \"units\": 400, \"activation\": \"linear\", \"use_bias\": true, \"kernel_initializer\": {\"class_name\": \"VarianceScaling\", \"config\": {\"scale\": 2.0, \"mode\": \"fan_in\", \"distribution\": \"uniform\", \"seed\": null}}, \"bias_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"kernel_regularizer\": null, \"bias_regularizer\": null, \"activity_regularizer\": null, \"kernel_constraint\": null, \"bias_constraint\": null}}, {\"class_name\": \"BatchNormalization\", \"config\": {\"name\": \"batch_normalization_7\", \"trainable\": true, \"dtype\": \"float32\", \"axis\": [1], \"momentum\": 0.99, \"epsilon\": 0.001, \"center\": true, \"scale\": true, \"beta_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"gamma_initializer\": {\"class_name\": \"Ones\", \"config\": {}}, \"moving_mean_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"moving_variance_initializer\": {\"class_name\": \"Ones\", \"config\": {}}, \"beta_regularizer\": null, \"gamma_regularizer\": null, \"beta_constraint\": null, \"gamma_constraint\": null}}, {\"class_name\": \"ReLU\", \"config\": {\"name\": \"re_lu_7\", \"trainable\": true, \"dtype\": \"float32\", \"max_value\": null, \"negative_slope\": 0.0, \"threshold\": 0.0}}, {\"class_name\": \"Dense\", \"config\": {\"name\": \"dense_10\", \"trainable\": true, \"dtype\": \"float32\", \"units\": 400, \"activation\": \"linear\", \"use_bias\": true, \"kernel_initializer\": {\"class_name\": \"VarianceScaling\", \"config\": {\"scale\": 2.0, \"mode\": \"fan_in\", \"distribution\": \"uniform\", \"seed\": null}}, \"bias_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"kernel_regularizer\": null, \"bias_regularizer\": null, \"activity_regularizer\": null, \"kernel_constraint\": null, \"bias_constraint\": null}}, {\"class_name\": \"BatchNormalization\", \"config\": {\"name\": \"batch_normalization_8\", \"trainable\": true, \"dtype\": \"float32\", \"axis\": [1], \"momentum\": 0.99, \"epsilon\": 0.001, \"center\": true, \"scale\": true, \"beta_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"gamma_initializer\": {\"class_name\": \"Ones\", \"config\": {}}, \"moving_mean_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"moving_variance_initializer\": {\"class_name\": \"Ones\", \"config\": {}}, \"beta_regularizer\": null, \"gamma_regularizer\": null, \"beta_constraint\": null, \"gamma_constraint\": null}}, {\"class_name\": \"ReLU\", \"config\": {\"name\": \"re_lu_8\", \"trainable\": true, \"dtype\": \"float32\", \"max_value\": null, \"negative_slope\": 0.0, \"threshold\": 0.0}}, {\"class_name\": \"Dense\", \"config\": {\"name\": \"dense_11\", \"trainable\": true, \"dtype\": \"float32\", \"units\": 9, \"activation\": \"linear\", \"use_bias\": true, \"kernel_initializer\": {\"class_name\": \"GlorotUniform\", \"config\": {\"seed\": null}}, \"bias_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"kernel_regularizer\": null, \"bias_regularizer\": null, \"activity_regularizer\": null, \"kernel_constraint\": null, \"bias_constraint\": null}}]}, \"keras_version\": \"2.2.4-tf\", \"backend\": \"tensorflow\"}\n",
            "(330984, 14) (330984, 9) (146907, 14) (146907, 9)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HtFN-OP2HlqQ",
        "colab_type": "text"
      },
      "source": [
        "## train, save and load model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DDmtdBRxJKJd",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def train_model(model_tupel, epochs, batch_size):\n",
        "  model_name = model_tupel[0]\n",
        "  keras_model = model_tupel[1]\n",
        "  train_X, train_Y, test_X, test_Y = model_tupel[2]\n",
        "\n",
        "  callbacks = [\n",
        "              #prepare_checkpoint_callback(model_name=model_name),\n",
        "              #prepare_tensorboard_callback(model_name=model_name)\n",
        "              ]\n",
        "\n",
        "  history = keras_model.fit(train_X,\n",
        "                            train_Y,\n",
        "                            epochs=epochs,\n",
        "                            batch_size=batch_size,\n",
        "                            validation_data=(test_X, test_Y),\n",
        "                            verbose=2,\n",
        "                            callbacks=callbacks\n",
        "                            )\n",
        "  return history  \n",
        "\n",
        "def _format_model_path(model_name):\n",
        "  file_name = '{}_end'.format(model_name)\n",
        "  return os.path.join(saved_model_weight_path, file_name)\n",
        "\n",
        "def save_model(model_tupel):\n",
        "  model_name = model_tupel[0]\n",
        "  keras_model = model_tupel[1]\n",
        "  file_path = _format_model_path(model_name)\n",
        "  keras_model.save_weights(file_path)\n",
        "\n",
        "def load_model(model_tupel):\n",
        "  model_name = model_tupel[0]\n",
        "  keras_model = model_tupel[1]\n",
        "  file_path = _format_model_path(model_name)\n",
        "  exact_file_path = file_path + '.index'\n",
        "  if os.path.isfile(exact_file_path):\n",
        "    print('load model {}'.format(model_name))\n",
        "    keras_model.load_weights(file_path)\n",
        "    return True\n",
        "  else:\n",
        "    return False\n",
        "\n",
        "def _format_history_path(model_name):\n",
        "  file_name = '{}_history_dict'.format(model_name)\n",
        "  return os.path.join(results_folder, file_name)\n",
        "  \n",
        "\n",
        "def save_history(model_tupel, history):\n",
        "  file_path = _format_history_path(model_tupel[0])\n",
        "  with open(file_path, 'wb') as file_hist:\n",
        "        pickle.dump(history.history, file_hist)\n",
        "  \n",
        "\n",
        "def load_history(model_tupel):\n",
        "  file_path = _format_history_path(model_tupel[0])\n",
        "  if not os.path.isfile(file_path):\n",
        "    return dict()\n",
        "  with open(file_path, 'rb') as file_hist:\n",
        "    return pickle.load(file_hist)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DxQg934JhE_H",
        "colab_type": "code",
        "outputId": "db50237d-991a-4759-fd34-00722325369c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "histories = []\n",
        "for model in models:\n",
        "  if not load_model(model):\n",
        "    print(model[0])\n",
        "    history = train_model(model, epochs=30, batch_size=64)\n",
        "    save_model(model)\n",
        "    save_history(model, history)\n",
        "    histories.append(history.history)\n",
        "    print('*******************')\n",
        "  else:\n",
        "    history = load_history(model)\n",
        "    histories.append(history)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "baseline\n",
            "Train on 330984 samples, validate on 146907 samples\n",
            "Epoch 1/30\n",
            "330984/330984 - 40s - loss: 0.0033 - mean_squared_error: 0.0033 - mean_absolute_error: 0.0347 - val_loss: 0.0044 - val_mean_squared_error: 0.0044 - val_mean_absolute_error: 0.0535\n",
            "Epoch 2/30\n",
            "330984/330984 - 37s - loss: 0.0011 - mean_squared_error: 0.0011 - mean_absolute_error: 0.0254 - val_loss: 0.0020 - val_mean_squared_error: 0.0020 - val_mean_absolute_error: 0.0368\n",
            "Epoch 3/30\n",
            "330984/330984 - 37s - loss: 7.9329e-04 - mean_squared_error: 7.9329e-04 - mean_absolute_error: 0.0215 - val_loss: 0.0012 - val_mean_squared_error: 0.0012 - val_mean_absolute_error: 0.0300\n",
            "Epoch 4/30\n",
            "330984/330984 - 37s - loss: 5.7949e-04 - mean_squared_error: 5.7949e-04 - mean_absolute_error: 0.0181 - val_loss: 5.3728e-04 - val_mean_squared_error: 5.3728e-04 - val_mean_absolute_error: 0.0191\n",
            "Epoch 5/30\n",
            "330984/330984 - 37s - loss: 4.5170e-04 - mean_squared_error: 4.5170e-04 - mean_absolute_error: 0.0157 - val_loss: 7.6081e-04 - val_mean_squared_error: 7.6081e-04 - val_mean_absolute_error: 0.0229\n",
            "Epoch 6/30\n",
            "330984/330984 - 37s - loss: 3.4290e-04 - mean_squared_error: 3.4290e-04 - mean_absolute_error: 0.0133 - val_loss: 2.4203e-04 - val_mean_squared_error: 2.4203e-04 - val_mean_absolute_error: 0.0117\n",
            "Epoch 7/30\n",
            "330984/330984 - 37s - loss: 2.6779e-04 - mean_squared_error: 2.6779e-04 - mean_absolute_error: 0.0116 - val_loss: 1.9266e-04 - val_mean_squared_error: 1.9266e-04 - val_mean_absolute_error: 0.0104\n",
            "Epoch 8/30\n",
            "330984/330984 - 37s - loss: 2.2109e-04 - mean_squared_error: 2.2109e-04 - mean_absolute_error: 0.0104 - val_loss: 1.8095e-04 - val_mean_squared_error: 1.8095e-04 - val_mean_absolute_error: 0.0099\n",
            "Epoch 9/30\n",
            "330984/330984 - 38s - loss: 1.7782e-04 - mean_squared_error: 1.7782e-04 - mean_absolute_error: 0.0094 - val_loss: 2.0259e-04 - val_mean_squared_error: 2.0259e-04 - val_mean_absolute_error: 0.0100\n",
            "Epoch 10/30\n",
            "330984/330984 - 38s - loss: 1.4998e-04 - mean_squared_error: 1.4998e-04 - mean_absolute_error: 0.0087 - val_loss: 1.8662e-04 - val_mean_squared_error: 1.8662e-04 - val_mean_absolute_error: 0.0101\n",
            "Epoch 11/30\n",
            "330984/330984 - 37s - loss: 1.1774e-04 - mean_squared_error: 1.1774e-04 - mean_absolute_error: 0.0078 - val_loss: 1.6895e-04 - val_mean_squared_error: 1.6895e-04 - val_mean_absolute_error: 0.0088\n",
            "Epoch 12/30\n",
            "330984/330984 - 37s - loss: 8.7643e-05 - mean_squared_error: 8.7643e-05 - mean_absolute_error: 0.0068 - val_loss: 1.4685e-04 - val_mean_squared_error: 1.4685e-04 - val_mean_absolute_error: 0.0091\n",
            "Epoch 13/30\n",
            "330984/330984 - 37s - loss: 6.8768e-05 - mean_squared_error: 6.8768e-05 - mean_absolute_error: 0.0060 - val_loss: 1.1299e-04 - val_mean_squared_error: 1.1299e-04 - val_mean_absolute_error: 0.0075\n",
            "Epoch 14/30\n",
            "330984/330984 - 37s - loss: 6.0304e-05 - mean_squared_error: 6.0304e-05 - mean_absolute_error: 0.0056 - val_loss: 1.3400e-04 - val_mean_squared_error: 1.3400e-04 - val_mean_absolute_error: 0.0085\n",
            "Epoch 15/30\n",
            "330984/330984 - 37s - loss: 5.1875e-05 - mean_squared_error: 5.1875e-05 - mean_absolute_error: 0.0052 - val_loss: 1.0995e-04 - val_mean_squared_error: 1.0995e-04 - val_mean_absolute_error: 0.0079\n",
            "Epoch 16/30\n",
            "330984/330984 - 37s - loss: 4.6153e-05 - mean_squared_error: 4.6153e-05 - mean_absolute_error: 0.0049 - val_loss: 9.8306e-05 - val_mean_squared_error: 9.8306e-05 - val_mean_absolute_error: 0.0063\n",
            "Epoch 17/30\n",
            "330984/330984 - 38s - loss: 4.2757e-05 - mean_squared_error: 4.2757e-05 - mean_absolute_error: 0.0046 - val_loss: 7.2744e-05 - val_mean_squared_error: 7.2744e-05 - val_mean_absolute_error: 0.0056\n",
            "Epoch 18/30\n",
            "330984/330984 - 37s - loss: 3.7431e-05 - mean_squared_error: 3.7431e-05 - mean_absolute_error: 0.0043 - val_loss: 1.1172e-04 - val_mean_squared_error: 1.1172e-04 - val_mean_absolute_error: 0.0069\n",
            "Epoch 19/30\n",
            "330984/330984 - 37s - loss: 3.6263e-05 - mean_squared_error: 3.6263e-05 - mean_absolute_error: 0.0042 - val_loss: 7.0976e-05 - val_mean_squared_error: 7.0976e-05 - val_mean_absolute_error: 0.0053\n",
            "Epoch 20/30\n",
            "330984/330984 - 37s - loss: 3.3786e-05 - mean_squared_error: 3.3786e-05 - mean_absolute_error: 0.0041 - val_loss: 8.2221e-05 - val_mean_squared_error: 8.2221e-05 - val_mean_absolute_error: 0.0057\n",
            "Epoch 21/30\n",
            "330984/330984 - 37s - loss: 3.0853e-05 - mean_squared_error: 3.0853e-05 - mean_absolute_error: 0.0039 - val_loss: 1.0097e-04 - val_mean_squared_error: 1.0097e-04 - val_mean_absolute_error: 0.0067\n",
            "Epoch 22/30\n",
            "330984/330984 - 37s - loss: 3.0505e-05 - mean_squared_error: 3.0505e-05 - mean_absolute_error: 0.0039 - val_loss: 1.0930e-04 - val_mean_squared_error: 1.0930e-04 - val_mean_absolute_error: 0.0074\n",
            "Epoch 23/30\n",
            "330984/330984 - 37s - loss: 2.7815e-05 - mean_squared_error: 2.7814e-05 - mean_absolute_error: 0.0037 - val_loss: 7.6263e-05 - val_mean_squared_error: 7.6263e-05 - val_mean_absolute_error: 0.0057\n",
            "Epoch 24/30\n",
            "330984/330984 - 37s - loss: 2.8293e-05 - mean_squared_error: 2.8293e-05 - mean_absolute_error: 0.0037 - val_loss: 7.0571e-05 - val_mean_squared_error: 7.0571e-05 - val_mean_absolute_error: 0.0052\n",
            "Epoch 25/30\n",
            "330984/330984 - 37s - loss: 2.6460e-05 - mean_squared_error: 2.6460e-05 - mean_absolute_error: 0.0036 - val_loss: 6.2075e-05 - val_mean_squared_error: 6.2075e-05 - val_mean_absolute_error: 0.0048\n",
            "Epoch 26/30\n",
            "330984/330984 - 38s - loss: 2.5139e-05 - mean_squared_error: 2.5139e-05 - mean_absolute_error: 0.0035 - val_loss: 7.5105e-05 - val_mean_squared_error: 7.5105e-05 - val_mean_absolute_error: 0.0054\n",
            "Epoch 27/30\n",
            "330984/330984 - 37s - loss: 2.3951e-05 - mean_squared_error: 2.3951e-05 - mean_absolute_error: 0.0034 - val_loss: 6.1700e-05 - val_mean_squared_error: 6.1700e-05 - val_mean_absolute_error: 0.0046\n",
            "Epoch 28/30\n",
            "330984/330984 - 37s - loss: 2.3840e-05 - mean_squared_error: 2.3840e-05 - mean_absolute_error: 0.0034 - val_loss: 7.8350e-05 - val_mean_squared_error: 7.8350e-05 - val_mean_absolute_error: 0.0059\n",
            "Epoch 29/30\n",
            "330984/330984 - 37s - loss: 2.2435e-05 - mean_squared_error: 2.2435e-05 - mean_absolute_error: 0.0033 - val_loss: 8.3627e-05 - val_mean_squared_error: 8.3627e-05 - val_mean_absolute_error: 0.0063\n",
            "Epoch 30/30\n",
            "330984/330984 - 37s - loss: 2.1738e-05 - mean_squared_error: 2.1738e-05 - mean_absolute_error: 0.0032 - val_loss: 7.1771e-05 - val_mean_squared_error: 7.1771e-05 - val_mean_absolute_error: 0.0055\n",
            "*******************\n",
            "bigger\n",
            "Train on 330984 samples, validate on 146907 samples\n",
            "Epoch 1/30\n",
            "330984/330984 - 38s - loss: 0.0041 - mean_squared_error: 0.0041 - mean_absolute_error: 0.0371 - val_loss: 0.0021 - val_mean_squared_error: 0.0021 - val_mean_absolute_error: 0.0366\n",
            "Epoch 2/30\n",
            "330984/330984 - 37s - loss: 0.0014 - mean_squared_error: 0.0014 - mean_absolute_error: 0.0289 - val_loss: 0.0017 - val_mean_squared_error: 0.0017 - val_mean_absolute_error: 0.0340\n",
            "Epoch 3/30\n",
            "330984/330984 - 36s - loss: 0.0010 - mean_squared_error: 0.0010 - mean_absolute_error: 0.0243 - val_loss: 0.0022 - val_mean_squared_error: 0.0022 - val_mean_absolute_error: 0.0388\n",
            "Epoch 4/30\n",
            "330984/330984 - 37s - loss: 7.2928e-04 - mean_squared_error: 7.2928e-04 - mean_absolute_error: 0.0200 - val_loss: 7.0525e-04 - val_mean_squared_error: 7.0525e-04 - val_mean_absolute_error: 0.0224\n",
            "Epoch 5/30\n",
            "330984/330984 - 37s - loss: 5.0316e-04 - mean_squared_error: 5.0316e-04 - mean_absolute_error: 0.0159 - val_loss: 3.2911e-04 - val_mean_squared_error: 3.2911e-04 - val_mean_absolute_error: 0.0146\n",
            "Epoch 6/30\n",
            "330984/330984 - 37s - loss: 3.7624e-04 - mean_squared_error: 3.7624e-04 - mean_absolute_error: 0.0133 - val_loss: 2.0070e-04 - val_mean_squared_error: 2.0071e-04 - val_mean_absolute_error: 0.0102\n",
            "Epoch 7/30\n",
            "330984/330984 - 36s - loss: 3.1114e-04 - mean_squared_error: 3.1114e-04 - mean_absolute_error: 0.0120 - val_loss: 2.1377e-04 - val_mean_squared_error: 2.1377e-04 - val_mean_absolute_error: 0.0110\n",
            "Epoch 8/30\n",
            "330984/330984 - 37s - loss: 2.7034e-04 - mean_squared_error: 2.7034e-04 - mean_absolute_error: 0.0113 - val_loss: 1.5740e-04 - val_mean_squared_error: 1.5740e-04 - val_mean_absolute_error: 0.0087\n",
            "Epoch 9/30\n",
            "330984/330984 - 37s - loss: 1.9222e-04 - mean_squared_error: 1.9222e-04 - mean_absolute_error: 0.0099 - val_loss: 1.2924e-04 - val_mean_squared_error: 1.2924e-04 - val_mean_absolute_error: 0.0078\n",
            "Epoch 10/30\n",
            "330984/330984 - 37s - loss: 1.2341e-04 - mean_squared_error: 1.2341e-04 - mean_absolute_error: 0.0081 - val_loss: 2.3286e-04 - val_mean_squared_error: 2.3286e-04 - val_mean_absolute_error: 0.0118\n",
            "Epoch 11/30\n",
            "330984/330984 - 37s - loss: 8.8571e-05 - mean_squared_error: 8.8571e-05 - mean_absolute_error: 0.0069 - val_loss: 1.3642e-04 - val_mean_squared_error: 1.3642e-04 - val_mean_absolute_error: 0.0073\n",
            "Epoch 12/30\n",
            "330984/330984 - 36s - loss: 7.2572e-05 - mean_squared_error: 7.2572e-05 - mean_absolute_error: 0.0061 - val_loss: 1.3442e-04 - val_mean_squared_error: 1.3442e-04 - val_mean_absolute_error: 0.0077\n",
            "Epoch 13/30\n",
            "330984/330984 - 38s - loss: 5.7606e-05 - mean_squared_error: 5.7606e-05 - mean_absolute_error: 0.0055 - val_loss: 1.1951e-04 - val_mean_squared_error: 1.1951e-04 - val_mean_absolute_error: 0.0072\n",
            "Epoch 14/30\n",
            "330984/330984 - 36s - loss: 5.1205e-05 - mean_squared_error: 5.1205e-05 - mean_absolute_error: 0.0051 - val_loss: 9.2106e-05 - val_mean_squared_error: 9.2106e-05 - val_mean_absolute_error: 0.0067\n",
            "Epoch 15/30\n",
            "330984/330984 - 37s - loss: 4.4811e-05 - mean_squared_error: 4.4811e-05 - mean_absolute_error: 0.0048 - val_loss: 1.0868e-04 - val_mean_squared_error: 1.0868e-04 - val_mean_absolute_error: 0.0072\n",
            "Epoch 16/30\n",
            "330984/330984 - 36s - loss: 4.0885e-05 - mean_squared_error: 4.0885e-05 - mean_absolute_error: 0.0046 - val_loss: 8.9922e-05 - val_mean_squared_error: 8.9922e-05 - val_mean_absolute_error: 0.0061\n",
            "Epoch 17/30\n",
            "330984/330984 - 37s - loss: 3.8272e-05 - mean_squared_error: 3.8272e-05 - mean_absolute_error: 0.0044 - val_loss: 7.9245e-05 - val_mean_squared_error: 7.9245e-05 - val_mean_absolute_error: 0.0056\n",
            "Epoch 18/30\n",
            "330984/330984 - 36s - loss: 3.5478e-05 - mean_squared_error: 3.5478e-05 - mean_absolute_error: 0.0042 - val_loss: 8.8572e-05 - val_mean_squared_error: 8.8572e-05 - val_mean_absolute_error: 0.0061\n",
            "Epoch 19/30\n",
            "330984/330984 - 36s - loss: 3.2390e-05 - mean_squared_error: 3.2390e-05 - mean_absolute_error: 0.0040 - val_loss: 6.6455e-05 - val_mean_squared_error: 6.6456e-05 - val_mean_absolute_error: 0.0047\n",
            "Epoch 20/30\n",
            "330984/330984 - 36s - loss: 3.0630e-05 - mean_squared_error: 3.0630e-05 - mean_absolute_error: 0.0039 - val_loss: 8.6238e-05 - val_mean_squared_error: 8.6238e-05 - val_mean_absolute_error: 0.0056\n",
            "Epoch 21/30\n",
            "330984/330984 - 37s - loss: 2.9592e-05 - mean_squared_error: 2.9592e-05 - mean_absolute_error: 0.0038 - val_loss: 8.7868e-05 - val_mean_squared_error: 8.7868e-05 - val_mean_absolute_error: 0.0063\n",
            "Epoch 22/30\n",
            "330984/330984 - 36s - loss: 2.8831e-05 - mean_squared_error: 2.8831e-05 - mean_absolute_error: 0.0038 - val_loss: 7.8507e-05 - val_mean_squared_error: 7.8507e-05 - val_mean_absolute_error: 0.0056\n",
            "Epoch 23/30\n",
            "330984/330984 - 36s - loss: 2.6471e-05 - mean_squared_error: 2.6471e-05 - mean_absolute_error: 0.0036 - val_loss: 7.1359e-05 - val_mean_squared_error: 7.1359e-05 - val_mean_absolute_error: 0.0056\n",
            "Epoch 24/30\n",
            "330984/330984 - 36s - loss: 2.6520e-05 - mean_squared_error: 2.6520e-05 - mean_absolute_error: 0.0036 - val_loss: 7.1771e-05 - val_mean_squared_error: 7.1771e-05 - val_mean_absolute_error: 0.0054\n",
            "Epoch 25/30\n",
            "330984/330984 - 36s - loss: 2.5249e-05 - mean_squared_error: 2.5249e-05 - mean_absolute_error: 0.0035 - val_loss: 6.4300e-05 - val_mean_squared_error: 6.4300e-05 - val_mean_absolute_error: 0.0047\n",
            "Epoch 26/30\n",
            "330984/330984 - 36s - loss: 2.3476e-05 - mean_squared_error: 2.3476e-05 - mean_absolute_error: 0.0034 - val_loss: 7.2054e-05 - val_mean_squared_error: 7.2054e-05 - val_mean_absolute_error: 0.0050\n",
            "Epoch 27/30\n",
            "330984/330984 - 36s - loss: 2.3636e-05 - mean_squared_error: 2.3636e-05 - mean_absolute_error: 0.0034 - val_loss: 7.1735e-05 - val_mean_squared_error: 7.1735e-05 - val_mean_absolute_error: 0.0055\n",
            "Epoch 28/30\n",
            "330984/330984 - 36s - loss: 2.2877e-05 - mean_squared_error: 2.2877e-05 - mean_absolute_error: 0.0033 - val_loss: 6.0308e-05 - val_mean_squared_error: 6.0308e-05 - val_mean_absolute_error: 0.0050\n",
            "Epoch 29/30\n",
            "330984/330984 - 36s - loss: 2.2647e-05 - mean_squared_error: 2.2647e-05 - mean_absolute_error: 0.0033 - val_loss: 6.6279e-05 - val_mean_squared_error: 6.6279e-05 - val_mean_absolute_error: 0.0049\n",
            "Epoch 30/30\n",
            "330984/330984 - 37s - loss: 2.1076e-05 - mean_squared_error: 2.1076e-05 - mean_absolute_error: 0.0032 - val_loss: 5.5615e-05 - val_mean_squared_error: 5.5615e-05 - val_mean_absolute_error: 0.0047\n",
            "*******************\n",
            "smaller\n",
            "Train on 330984 samples, validate on 146907 samples\n",
            "Epoch 1/30\n",
            "330984/330984 - 39s - loss: 0.0028 - mean_squared_error: 0.0028 - mean_absolute_error: 0.0323 - val_loss: 0.0019 - val_mean_squared_error: 0.0019 - val_mean_absolute_error: 0.0334\n",
            "Epoch 2/30\n",
            "330984/330984 - 37s - loss: 8.4876e-04 - mean_squared_error: 8.4876e-04 - mean_absolute_error: 0.0224 - val_loss: 4.2993e-04 - val_mean_squared_error: 4.2993e-04 - val_mean_absolute_error: 0.0159\n",
            "Epoch 3/30\n",
            "330984/330984 - 37s - loss: 5.8124e-04 - mean_squared_error: 5.8124e-04 - mean_absolute_error: 0.0184 - val_loss: 7.8010e-04 - val_mean_squared_error: 7.8010e-04 - val_mean_absolute_error: 0.0227\n",
            "Epoch 4/30\n",
            "330984/330984 - 37s - loss: 4.1796e-04 - mean_squared_error: 4.1796e-04 - mean_absolute_error: 0.0155 - val_loss: 3.8869e-04 - val_mean_squared_error: 3.8870e-04 - val_mean_absolute_error: 0.0156\n",
            "Epoch 5/30\n",
            "330984/330984 - 37s - loss: 3.2539e-04 - mean_squared_error: 3.2539e-04 - mean_absolute_error: 0.0134 - val_loss: 1.9085e-04 - val_mean_squared_error: 1.9085e-04 - val_mean_absolute_error: 0.0099\n",
            "Epoch 6/30\n",
            "330984/330984 - 37s - loss: 2.5131e-04 - mean_squared_error: 2.5131e-04 - mean_absolute_error: 0.0116 - val_loss: 2.1244e-04 - val_mean_squared_error: 2.1244e-04 - val_mean_absolute_error: 0.0103\n",
            "Epoch 7/30\n",
            "330984/330984 - 37s - loss: 2.0575e-04 - mean_squared_error: 2.0575e-04 - mean_absolute_error: 0.0104 - val_loss: 2.9792e-04 - val_mean_squared_error: 2.9792e-04 - val_mean_absolute_error: 0.0127\n",
            "Epoch 8/30\n",
            "330984/330984 - 38s - loss: 1.7231e-04 - mean_squared_error: 1.7231e-04 - mean_absolute_error: 0.0094 - val_loss: 1.9275e-04 - val_mean_squared_error: 1.9275e-04 - val_mean_absolute_error: 0.0103\n",
            "Epoch 9/30\n",
            "330984/330984 - 37s - loss: 1.4087e-04 - mean_squared_error: 1.4087e-04 - mean_absolute_error: 0.0085 - val_loss: 2.5680e-04 - val_mean_squared_error: 2.5680e-04 - val_mean_absolute_error: 0.0112\n",
            "Epoch 10/30\n",
            "330984/330984 - 37s - loss: 1.2901e-04 - mean_squared_error: 1.2901e-04 - mean_absolute_error: 0.0082 - val_loss: 1.6316e-04 - val_mean_squared_error: 1.6316e-04 - val_mean_absolute_error: 0.0087\n",
            "Epoch 11/30\n",
            "330984/330984 - 37s - loss: 1.0805e-04 - mean_squared_error: 1.0805e-04 - mean_absolute_error: 0.0075 - val_loss: 1.3142e-04 - val_mean_squared_error: 1.3142e-04 - val_mean_absolute_error: 0.0076\n",
            "Epoch 12/30\n",
            "330984/330984 - 37s - loss: 9.6531e-05 - mean_squared_error: 9.6531e-05 - mean_absolute_error: 0.0071 - val_loss: 1.3058e-04 - val_mean_squared_error: 1.3058e-04 - val_mean_absolute_error: 0.0078\n",
            "Epoch 13/30\n",
            "330984/330984 - 37s - loss: 8.4241e-05 - mean_squared_error: 8.4241e-05 - mean_absolute_error: 0.0066 - val_loss: 1.3849e-04 - val_mean_squared_error: 1.3849e-04 - val_mean_absolute_error: 0.0078\n",
            "Epoch 14/30\n",
            "330984/330984 - 37s - loss: 7.4029e-05 - mean_squared_error: 7.4029e-05 - mean_absolute_error: 0.0062 - val_loss: 1.0857e-04 - val_mean_squared_error: 1.0857e-04 - val_mean_absolute_error: 0.0067\n",
            "Epoch 15/30\n",
            "330984/330984 - 37s - loss: 6.6206e-05 - mean_squared_error: 6.6206e-05 - mean_absolute_error: 0.0059 - val_loss: 2.0423e-04 - val_mean_squared_error: 2.0423e-04 - val_mean_absolute_error: 0.0098\n",
            "Epoch 16/30\n",
            "330984/330984 - 38s - loss: 5.7824e-05 - mean_squared_error: 5.7824e-05 - mean_absolute_error: 0.0055 - val_loss: 1.0307e-04 - val_mean_squared_error: 1.0307e-04 - val_mean_absolute_error: 0.0062\n",
            "Epoch 17/30\n",
            "330984/330984 - 37s - loss: 5.2278e-05 - mean_squared_error: 5.2278e-05 - mean_absolute_error: 0.0052 - val_loss: 1.1690e-04 - val_mean_squared_error: 1.1690e-04 - val_mean_absolute_error: 0.0075\n",
            "Epoch 18/30\n",
            "330984/330984 - 37s - loss: 4.6796e-05 - mean_squared_error: 4.6796e-05 - mean_absolute_error: 0.0049 - val_loss: 1.2690e-04 - val_mean_squared_error: 1.2690e-04 - val_mean_absolute_error: 0.0068\n",
            "Epoch 19/30\n",
            "330984/330984 - 37s - loss: 4.5049e-05 - mean_squared_error: 4.5049e-05 - mean_absolute_error: 0.0048 - val_loss: 1.1115e-04 - val_mean_squared_error: 1.1115e-04 - val_mean_absolute_error: 0.0064\n",
            "Epoch 20/30\n",
            "330984/330984 - 37s - loss: 4.0510e-05 - mean_squared_error: 4.0510e-05 - mean_absolute_error: 0.0046 - val_loss: 1.2693e-04 - val_mean_squared_error: 1.2693e-04 - val_mean_absolute_error: 0.0074\n",
            "Epoch 21/30\n",
            "330984/330984 - 37s - loss: 3.7606e-05 - mean_squared_error: 3.7606e-05 - mean_absolute_error: 0.0044 - val_loss: 1.3223e-04 - val_mean_squared_error: 1.3223e-04 - val_mean_absolute_error: 0.0072\n",
            "Epoch 22/30\n",
            "330984/330984 - 37s - loss: 3.7064e-05 - mean_squared_error: 3.7064e-05 - mean_absolute_error: 0.0044 - val_loss: 9.4098e-05 - val_mean_squared_error: 9.4098e-05 - val_mean_absolute_error: 0.0058\n",
            "Epoch 23/30\n",
            "330984/330984 - 37s - loss: 3.4695e-05 - mean_squared_error: 3.4695e-05 - mean_absolute_error: 0.0042 - val_loss: 1.2301e-04 - val_mean_squared_error: 1.2301e-04 - val_mean_absolute_error: 0.0067\n",
            "Epoch 24/30\n",
            "330984/330984 - 37s - loss: 3.3148e-05 - mean_squared_error: 3.3148e-05 - mean_absolute_error: 0.0041 - val_loss: 9.1240e-05 - val_mean_squared_error: 9.1240e-05 - val_mean_absolute_error: 0.0062\n",
            "Epoch 25/30\n",
            "330984/330984 - 37s - loss: 3.1752e-05 - mean_squared_error: 3.1752e-05 - mean_absolute_error: 0.0040 - val_loss: 1.1011e-04 - val_mean_squared_error: 1.1011e-04 - val_mean_absolute_error: 0.0063\n",
            "Epoch 26/30\n",
            "330984/330984 - 37s - loss: 2.9699e-05 - mean_squared_error: 2.9699e-05 - mean_absolute_error: 0.0039 - val_loss: 1.0732e-04 - val_mean_squared_error: 1.0732e-04 - val_mean_absolute_error: 0.0063\n",
            "Epoch 27/30\n",
            "330984/330984 - 37s - loss: 2.9265e-05 - mean_squared_error: 2.9265e-05 - mean_absolute_error: 0.0038 - val_loss: 1.1353e-04 - val_mean_squared_error: 1.1353e-04 - val_mean_absolute_error: 0.0064\n",
            "Epoch 28/30\n",
            "330984/330984 - 37s - loss: 2.8651e-05 - mean_squared_error: 2.8651e-05 - mean_absolute_error: 0.0038 - val_loss: 1.0829e-04 - val_mean_squared_error: 1.0829e-04 - val_mean_absolute_error: 0.0065\n",
            "Epoch 29/30\n",
            "330984/330984 - 37s - loss: 2.7179e-05 - mean_squared_error: 2.7179e-05 - mean_absolute_error: 0.0037 - val_loss: 9.8890e-05 - val_mean_squared_error: 9.8890e-05 - val_mean_absolute_error: 0.0060\n",
            "Epoch 30/30\n",
            "330984/330984 - 37s - loss: 2.6203e-05 - mean_squared_error: 2.6203e-05 - mean_absolute_error: 0.0036 - val_loss: 7.8605e-05 - val_mean_squared_error: 7.8605e-05 - val_mean_absolute_error: 0.0051\n",
            "*******************\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XRLpUIJnZ-eq",
        "colab_type": "text"
      },
      "source": [
        "## visualize"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8X1nsxWEO59n",
        "colab_type": "text"
      },
      "source": [
        "### Test & Training Loss"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "q7eOZ0lAe1CP",
        "colab_type": "code",
        "outputId": "3ccd46f5-c151-412d-a212-bc7468dc8846",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 178
        }
      },
      "source": [
        "for i, history in enumerate(histories):\n",
        "  last_loss = history['loss'][-1]\n",
        "  last_val_loss = history['val_loss'][-1]\n",
        "  model_name = models[i][0]\n",
        "  print('{} model last loss: {}'.format(model_name,last_loss))\n",
        "  print('{} model last val loss: {}'.format(model_name,last_val_loss))\n",
        "  print('*******')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "baseline model last loss: 2.1737792101892417e-05\n",
            "baseline model last val loss: 7.177138656315764e-05\n",
            "*******\n",
            "bigger model last loss: 2.1076007996481955e-05\n",
            "bigger model last val loss: 5.5615317633272435e-05\n",
            "*******\n",
            "smaller model last loss: 2.6203176058980836e-05\n",
            "smaller model last val loss: 7.860530174445859e-05\n",
            "*******\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gbFhTgXGITKV",
        "colab_type": "code",
        "outputId": "d809073f-c748-4fd8-8124-f00ac491df2b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 606
        }
      },
      "source": [
        "pas_history(models, histories=histories)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAA7gAAAJNCAYAAAAbEdlFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nZBmxfgAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 1152x720 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4vT7MLiJPSQ2",
        "colab_type": "text"
      },
      "source": [
        "\n",
        "### MSE per target (scaled)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "o03Cq_x5nhRe",
        "colab_type": "code",
        "outputId": "e080e470-c683-4de2-8c5d-e9dc66e0f044",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 928
        }
      },
      "source": [
        "for model_tupel in models:\n",
        "  df = pas_metrics_scaled(model_tupel)\n",
        "  print(model_tupel[0])\n",
        "  with pd.option_context('display.max_rows', None, 'display.max_columns', None,\n",
        "                         'display.width', 1000): \n",
        "    print(df)\n",
        "  print('*********************************************************')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "baseline\n",
            "          Vtrue,_ktas      __VVI,__fpm      ____Q,rad/s      ____P,rad/s      ____R,rad/s      pitch,__deg      _roll,__deg      _beta,__deg      __alt,ftmsl \n",
            "count    146907.000000    146907.000000    146907.000000    146907.000000    146907.000000    146907.000000    146907.000000    146907.000000    146907.000000\n",
            "mean          0.504412         0.737227         0.570341         0.485060         0.485233         0.531587         0.464654         0.620080         0.541116\n",
            "std           0.106511         0.106115         0.096593         0.137143         0.297851         0.109146         0.243538         0.211064         0.313282\n",
            "min          -0.145712         0.315198         0.075314        -0.114067         0.001815         0.194934         0.019929        -0.224448         0.020174\n",
            "25%           0.491325         0.719309         0.483381         0.478066         0.141444         0.498651         0.178618         0.402814         0.258517\n",
            "50%           0.516065         0.736735         0.612474         0.485824         0.495373         0.523474         0.475117         0.619407         0.502333\n",
            "75%           0.540668         0.757760         0.622681         0.493745         0.817728         0.538931         0.733206         0.857215         0.743406\n",
            "max           0.863996         1.112528         0.958919         1.093909         0.990564         1.288604         0.928746         1.008269         0.980654\n",
            "mses          0.000075         0.000111         0.000053         0.000040         0.000092         0.000103         0.000080         0.000045         0.000046\n",
            "rmes          0.008685         0.010543         0.007286         0.006350         0.009582         0.010161         0.008967         0.006683         0.006768\n",
            "maes          0.005608         0.006680         0.004668         0.004180         0.006997         0.005957         0.006553         0.004482         0.004596\n",
            "mse           0.000072         0.000072         0.000072         0.000072         0.000072         0.000072         0.000072         0.000072         0.000072\n",
            "rme           0.008472         0.008472         0.008472         0.008472         0.008472         0.008472         0.008472         0.008472         0.008472\n",
            "mae           0.005524         0.005524         0.005524         0.005524         0.005524         0.005524         0.005524         0.005524         0.005524\n",
            "*********************************************************\n",
            "bigger\n",
            "          Vtrue,_ktas      __VVI,__fpm      ____Q,rad/s      ____P,rad/s      ____R,rad/s      pitch,__deg      _roll,__deg      _beta,__deg      __alt,ftmsl \n",
            "count    146907.000000    146907.000000    146907.000000    146907.000000    146907.000000    146907.000000    146907.000000    146907.000000    146907.000000\n",
            "mean          0.504412         0.737227         0.570341         0.485060         0.485233         0.531587         0.464654         0.620080         0.541116\n",
            "std           0.106511         0.106115         0.096593         0.137143         0.297851         0.109146         0.243538         0.211064         0.313282\n",
            "min          -0.145712         0.315198         0.075314        -0.114067         0.001815         0.194934         0.019929        -0.224448         0.020174\n",
            "25%           0.491325         0.719309         0.483381         0.478066         0.141444         0.498651         0.178618         0.402814         0.258517\n",
            "50%           0.516065         0.736735         0.612474         0.485824         0.495373         0.523474         0.475117         0.619407         0.502333\n",
            "75%           0.540668         0.757760         0.622681         0.493745         0.817728         0.538931         0.733206         0.857215         0.743406\n",
            "max           0.863996         1.112528         0.958919         1.093909         0.990564         1.288604         0.928746         1.008269         0.980654\n",
            "mses          0.000075         0.000108         0.000042         0.000046         0.000034         0.000075         0.000041         0.000038         0.000040\n",
            "rmes          0.008680         0.010400         0.006493         0.006783         0.005870         0.008671         0.006430         0.006162         0.006318\n",
            "maes          0.005634         0.005529         0.003841         0.005434         0.004145         0.005117         0.004306         0.003244         0.004643\n",
            "mse           0.000056         0.000056         0.000056         0.000056         0.000056         0.000056         0.000056         0.000056         0.000056\n",
            "rme           0.007458         0.007458         0.007458         0.007458         0.007458         0.007458         0.007458         0.007458         0.007458\n",
            "mae           0.004655         0.004655         0.004655         0.004655         0.004655         0.004655         0.004655         0.004655         0.004655\n",
            "*********************************************************\n",
            "smaller\n",
            "          Vtrue,_ktas      __VVI,__fpm      ____Q,rad/s      ____P,rad/s      ____R,rad/s      pitch,__deg      _roll,__deg      _beta,__deg      __alt,ftmsl \n",
            "count    146907.000000    146907.000000    146907.000000    146907.000000    146907.000000    146907.000000    146907.000000    146907.000000    146907.000000\n",
            "mean          0.504412         0.737227         0.570341         0.485060         0.485233         0.531587         0.464654         0.620080         0.541116\n",
            "std           0.106511         0.106115         0.096593         0.137143         0.297851         0.109146         0.243538         0.211064         0.313282\n",
            "min          -0.145712         0.315198         0.075314        -0.114067         0.001815         0.194934         0.019929        -0.224448         0.020174\n",
            "25%           0.491325         0.719309         0.483381         0.478066         0.141444         0.498651         0.178618         0.402814         0.258517\n",
            "50%           0.516065         0.736735         0.612474         0.485824         0.495373         0.523474         0.475117         0.619407         0.502333\n",
            "75%           0.540668         0.757760         0.622681         0.493745         0.817728         0.538931         0.733206         0.857215         0.743406\n",
            "max           0.863996         1.112528         0.958919         1.093909         0.990564         1.288604         0.928746         1.008269         0.980654\n",
            "mses          0.000141         0.000108         0.000078         0.000028         0.000091         0.000092         0.000059         0.000074         0.000034\n",
            "rmes          0.011886         0.010398         0.008856         0.005303         0.009556         0.009603         0.007703         0.008629         0.005847\n",
            "maes          0.007725         0.005851         0.006301         0.003347         0.005107         0.005184         0.004570         0.004038         0.003879\n",
            "mse           0.000079         0.000079         0.000079         0.000079         0.000079         0.000079         0.000079         0.000079         0.000079\n",
            "rme           0.008866         0.008866         0.008866         0.008866         0.008866         0.008866         0.008866         0.008866         0.008866\n",
            "mae           0.005111         0.005111         0.005111         0.005111         0.005111         0.005111         0.005111         0.005111         0.005111\n",
            "*********************************************************\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "bVhouDRRpewh",
        "colab_type": "text"
      },
      "source": [
        "### MSE per target (unscaled)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "crfMWfnKuAXz",
        "colab_type": "code",
        "outputId": "65254af6-84b7-4411-fd2d-876e7d64ac9b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "models[1][0]"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'bigger'"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 23
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oTTBfKRgRnTP",
        "colab_type": "code",
        "outputId": "77662b72-c9c8-4a4a-c30c-a4a2f1fa4c13",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 928
        }
      },
      "source": [
        "for model_tupel in models:\n",
        "  df = pas_metrics(model_tupel)\n",
        "  print(model_tupel[0])\n",
        "  with pd.option_context('display.max_rows', None, 'display.max_columns', None,\n",
        "                        'display.width', 1000):  \n",
        "    print(df)\n",
        "  print('*********************************************************')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "baseline\n",
            "          Vtrue,_ktas      __VVI,__fpm      ____Q,rad/s      ____P,rad/s      ____R,rad/s      pitch,__deg      _roll,__deg      _beta,__deg      __alt,ftmsl \n",
            "count    146907.000000    146907.000000     1.469070e+05    146907.000000    146907.000000    146907.000000    146907.000000    146907.000000    146907.000000\n",
            "mean        116.933735         1.454978     7.135510e-03        -0.000140        -0.001338        -0.605152        -0.722674         0.024940      3118.794644\n",
            "std           8.275017       276.265852     7.759326e-03         0.025104         0.032829         1.788004        12.122463         0.178670      1339.530659\n",
            "min          66.424780     -1097.281130    -3.263000e-02        -0.109810        -0.054620        -6.120130       -22.859510        -0.689970       891.350710\n",
            "25%         115.917000       -45.193610     1.500000e-04        -0.001420        -0.039230        -1.144705       -14.960540        -0.158980      1910.458070\n",
            "50%         117.839070         0.174230     1.052000e-02         0.000000        -0.000220        -0.738060        -0.201880         0.024370      2952.964840\n",
            "75%         119.750560        54.913025     1.134000e-02         0.001450         0.035310        -0.484845        12.644855         0.225680      3983.743410\n",
            "max         144.870320       978.538330     3.835000e-02         0.111310         0.054360        11.796170        22.378150         0.353550      4998.167480\n",
            "mses          0.455289       753.354787     3.425999e-07         0.000001         0.000001         0.027708         0.199230         0.000032       837.538498\n",
            "rmes          0.674751        27.447309     5.853203e-04         0.001162         0.001056         0.166458         0.446352         0.005657        28.940257\n",
            "maes          0.435707        17.390710     3.749511e-04         0.000765         0.000771         0.097585         0.326206         0.003794        19.650583\n",
            "mse         176.841727       176.841727     1.768417e+02       176.841727       176.841727       176.841727       176.841727       176.841727       176.841727\n",
            "rme          13.298185        13.298185     1.329819e+01        13.298185        13.298185        13.298185        13.298185        13.298185        13.298185\n",
            "mae           4.211833         4.211833     4.211833e+00         4.211833         4.211833         4.211833         4.211833         4.211833         4.211833\n",
            "*********************************************************\n",
            "bigger\n",
            "          Vtrue,_ktas      __VVI,__fpm      ____Q,rad/s      ____P,rad/s      ____R,rad/s      pitch,__deg      _roll,__deg      _beta,__deg      __alt,ftmsl \n",
            "count    146907.000000    146907.000000     1.469070e+05    146907.000000     1.469070e+05    146907.000000    146907.000000    146907.000000    146907.000000\n",
            "mean        116.933735         1.454978     7.135510e-03        -0.000140    -1.337600e-03        -0.605152        -0.722674         0.024940      3118.794644\n",
            "std           8.275017       276.265852     7.759326e-03         0.025104     3.282916e-02         1.788004        12.122463         0.178670      1339.530659\n",
            "min          66.424780     -1097.281130    -3.263000e-02        -0.109810    -5.462000e-02        -6.120130       -22.859510        -0.689970       891.350710\n",
            "25%         115.917000       -45.193610     1.500000e-04        -0.001420    -3.923000e-02        -1.144705       -14.960540        -0.158980      1910.458070\n",
            "50%         117.839070         0.174230     1.052000e-02         0.000000    -2.200000e-04        -0.738060        -0.201880         0.024370      2952.964840\n",
            "75%         119.750560        54.913025     1.134000e-02         0.001450     3.531000e-02        -0.484845        12.644855         0.225680      3983.743410\n",
            "max         144.870320       978.538330     3.835000e-02         0.111310     5.436000e-02        11.796170        22.378150         0.353550      4998.167480\n",
            "mses          0.454715       733.119314     2.720288e-07         0.000002     4.186467e-07         0.020178         0.102425         0.000027       729.747139\n",
            "rmes          0.674325        27.076176     5.215638e-04         0.001242     6.470292e-04         0.142048         0.320039         0.005217        27.013832\n",
            "maes          0.437700        14.393475     3.085295e-04         0.000995     4.568384e-04         0.083818         0.214316         0.002746        19.853147\n",
            "mse         162.604867       162.604867     1.626049e+02       162.604867     1.626049e+02       162.604867       162.604867       162.604867       162.604867\n",
            "rme          12.751661        12.751661     1.275166e+01        12.751661     1.275166e+01        12.751661        12.751661        12.751661        12.751661\n",
            "mae           3.887440         3.887440     3.887440e+00         3.887440     3.887440e+00         3.887440         3.887440         3.887440         3.887440\n",
            "*********************************************************\n",
            "smaller\n",
            "          Vtrue,_ktas      __VVI,__fpm      ____Q,rad/s      ____P,rad/s      ____R,rad/s      pitch,__deg      _roll,__deg      _beta,__deg      __alt,ftmsl \n",
            "count    146907.000000    146907.000000     1.469070e+05     1.469070e+05    146907.000000    146907.000000    146907.000000    146907.000000    146907.000000\n",
            "mean        116.933735         1.454978     7.135510e-03    -1.397062e-04        -0.001338        -0.605152        -0.722674         0.024940      3118.794644\n",
            "std           8.275017       276.265852     7.759326e-03     2.510405e-02         0.032829         1.788004        12.122463         0.178670      1339.530659\n",
            "min          66.424780     -1097.281130    -3.263000e-02    -1.098100e-01        -0.054620        -6.120130       -22.859510        -0.689970       891.350710\n",
            "25%         115.917000       -45.193610     1.500000e-04    -1.420000e-03        -0.039230        -1.144705       -14.960540        -0.158980      1910.458070\n",
            "50%         117.839070         0.174230     1.052000e-02     0.000000e+00        -0.000220        -0.738060        -0.201880         0.024370      2952.964840\n",
            "75%         119.750560        54.913025     1.134000e-02     1.450000e-03         0.035310        -0.484845        12.644855         0.225680      3983.743410\n",
            "max         144.870320       978.538330     3.835000e-02     1.113100e-01         0.054360        11.796170        22.378150         0.353550      4998.167480\n",
            "mses          0.852692       732.809785     5.060991e-07     9.423869e-07         0.000001         0.024745         0.147032         0.000053       624.956447\n",
            "rmes          0.923413        27.070460     7.114065e-04     9.707661e-04         0.001053         0.157307         0.383447         0.007305        24.999129\n",
            "maes          0.600168        15.233447     5.061811e-04     6.126378e-04         0.000563         0.084929         0.227490         0.003418        16.584744\n",
            "mse         150.976751       150.976751     1.509768e+02     1.509768e+02       150.976751       150.976751       150.976751       150.976751       150.976751\n",
            "rme          12.287260        12.287260     1.228726e+01     1.228726e+01        12.287260        12.287260        12.287260        12.287260        12.287260\n",
            "mae           3.637320         3.637320     3.637320e+00     3.637320e+00         3.637320         3.637320         3.637320         3.637320         3.637320\n",
            "*********************************************************\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mKtWiYSYjKJp",
        "colab_type": "text"
      },
      "source": [
        "#latex formatted\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "x372S3zt0XoH",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def _df_prediction_metrics_all(model_tupel, scaled):\n",
        "  \n",
        "  model_name =  model_tupel[0]\n",
        "  model = model_tupel[1]\n",
        "  train_X, train_Y, test_X, test_Y = model_tupel[2]\n",
        "  cols_label = model_tupel[4]\n",
        "\n",
        "\n",
        "  pred_Y = model.predict(test_X)\n",
        "  \n",
        "  if scaled:\n",
        "    truth_Y = test_Y\n",
        "  else:\n",
        "    label_scaler = model_tupel[3]\n",
        "    pred_Y = label_scaler.inverse_transform(pred_Y)\n",
        "    truth_Y = label_scaler.inverse_transform(test_Y)\n",
        "\n",
        "  df_describe = pd.DataFrame(truth_Y,columns=cols_label).describe()\n",
        "  mses = mean_squared_error(truth_Y, pred_Y, multioutput='raw_values')\n",
        "  #rmses = np.array([math.sqrt(mse) for mse in mses])\n",
        "  #maes = mean_absolute_error(truth_Y, pred_Y, multioutput='raw_values')\n",
        "  # average over list should do the job as well\n",
        "  mse = mean_squared_error(truth_Y, pred_Y)\n",
        "  rmse = math.sqrt(mse)\n",
        "  mae = mean_absolute_error(truth_Y, pred_Y)\n",
        "\n",
        "\n",
        "  df_mse_per_feature = pd.DataFrame(data=mses.reshape(1,-1),    # values\n",
        "            index=[model_name],     # index\n",
        "            columns=cols_label)  #  column names\n",
        "  \n",
        "  df_mse_per_feature['MSE'] = mse\n",
        "  df_mse_per_feature['RMSE'] = rmse\n",
        "  df_mse_per_feature['MAE'] = mae\n",
        "\n",
        "\n",
        "  #df_rmse_per_feature = pd.DataFrame(data=rmses.reshape(1,-1),    # values\n",
        "   #     index=['rmes'],     # index\n",
        "  #      columns=cols_label)  #  column names\n",
        "  #df_mae_per_feature = pd.DataFrame(data=maes.reshape(1,-1),    # values\n",
        "    #    index=['maes'],     # index\n",
        "   #     columns=cols_label)  #  column names\n",
        "  \n",
        "  return df_mse_per_feature\n",
        "\n",
        "\n",
        "def pas_metrics_all(models):\n",
        "  cols_trans = ['MSE', 'RMSE', 'MAE', col_vv,col_alt, col_v]\n",
        "  cols_rot = [col_roll_deg, col_pitch_deg, col_yaw_deg, col_pitch_v, col_yaw_v, col_roll_v]\n",
        "  rows_trans = []\n",
        "  rows_rot = []\n",
        "\n",
        "  for model_tupel in models:\n",
        "    model_row = _df_prediction_metrics_all(model_tupel, scaled=False)\n",
        "    rows_trans.append(model_row[cols_trans])\n",
        "    rows_rot.append(model_row[cols_rot])\n",
        "\n",
        "  df_trans = pd.concat(rows_trans).round(2)\n",
        "  file_path = os.path.join(results_folder, 'all_mse_trans' + '.csv')\n",
        "  df_trans.to_csv(path_or_buf=file_path)\n",
        "  \n",
        "  df_rot = pd.concat(rows_rot).round(10)\n",
        "  file_path = os.path.join(results_folder, 'all_mse_rot' + '.csv')\n",
        "  df_rot.to_csv(path_or_buf=file_path)\n",
        "  return df_trans, df_rot\n",
        "\n",
        "\n",
        "df_trans, df_rot = pas_metrics_all(models)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "srUx1ixdEobz",
        "colab_type": "code",
        "outputId": "bfb9859b-5fe9-4c92-95de-7a267df0ae53",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 143
        }
      },
      "source": [
        "df_trans"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "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>MSE</th>\n",
              "      <th>RMSE</th>\n",
              "      <th>MAE</th>\n",
              "      <th>__VVI,__fpm</th>\n",
              "      <th>__alt,ftmsl</th>\n",
              "      <th>Vtrue,_ktas</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>baseline</th>\n",
              "      <td>176.84</td>\n",
              "      <td>13.30</td>\n",
              "      <td>4.21</td>\n",
              "      <td>753.35</td>\n",
              "      <td>837.54</td>\n",
              "      <td>0.46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>bigger</th>\n",
              "      <td>162.60</td>\n",
              "      <td>12.75</td>\n",
              "      <td>3.89</td>\n",
              "      <td>733.12</td>\n",
              "      <td>729.75</td>\n",
              "      <td>0.45</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>smaller</th>\n",
              "      <td>150.98</td>\n",
              "      <td>12.29</td>\n",
              "      <td>3.64</td>\n",
              "      <td>732.81</td>\n",
              "      <td>624.96</td>\n",
              "      <td>0.85</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "             MSE   RMSE  ...     __alt,ftmsl      Vtrue,_ktas \n",
              "baseline  176.84  13.30  ...           837.54             0.46\n",
              "bigger    162.60  12.75  ...           729.75             0.45\n",
              "smaller   150.98  12.29  ...           624.96             0.85\n",
              "\n",
              "[3 rows x 6 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 132
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ktZVFuJ2EoPN",
        "colab_type": "code",
        "outputId": "6c1af28f-f4ca-4903-eb01-abda7d10edfb",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 143
        }
      },
      "source": [
        "df_rot.round(10)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "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>_roll,__deg</th>\n",
              "      <th>pitch,__deg</th>\n",
              "      <th>_beta,__deg</th>\n",
              "      <th>____P,rad/s</th>\n",
              "      <th>____Q,rad/s</th>\n",
              "      <th>____R,rad/s</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>baseline</th>\n",
              "      <td>0.19923</td>\n",
              "      <td>0.02771</td>\n",
              "      <td>0.00003</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>bigger</th>\n",
              "      <td>0.10242</td>\n",
              "      <td>0.02018</td>\n",
              "      <td>0.00003</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>smaller</th>\n",
              "      <td>0.14703</td>\n",
              "      <td>0.02475</td>\n",
              "      <td>0.00005</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "             _roll,__deg      pitch,__deg   ...     ____Q,rad/s      ____R,rad/s \n",
              "baseline          0.19923          0.02771  ...          0.00000          0.00000\n",
              "bigger            0.10242          0.02018  ...          0.00000          0.00000\n",
              "smaller           0.14703          0.02475  ...          0.00000          0.00000\n",
              "\n",
              "[3 rows x 6 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 141
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WgK03QWILdSm",
        "colab_type": "code",
        "outputId": "e410bbb4-26d0-4fe0-cf25-963cfd51abb9",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 143
        }
      },
      "source": [
        "df_rot"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "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>_roll,__deg</th>\n",
              "      <th>pitch,__deg</th>\n",
              "      <th>_beta,__deg</th>\n",
              "      <th>____P,rad/s</th>\n",
              "      <th>____Q,rad/s</th>\n",
              "      <th>____R,rad/s</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>baseline</th>\n",
              "      <td>0.19923</td>\n",
              "      <td>0.02771</td>\n",
              "      <td>0.00003</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>bigger</th>\n",
              "      <td>0.10242</td>\n",
              "      <td>0.02018</td>\n",
              "      <td>0.00003</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>smaller</th>\n",
              "      <td>0.14703</td>\n",
              "      <td>0.02475</td>\n",
              "      <td>0.00005</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "             _roll,__deg      pitch,__deg   ...     ____Q,rad/s      ____R,rad/s \n",
              "baseline          0.19923          0.02771  ...          0.00000          0.00000\n",
              "bigger            0.10242          0.02018  ...          0.00000          0.00000\n",
              "smaller           0.14703          0.02475  ...          0.00000          0.00000\n",
              "\n",
              "[3 rows x 6 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 140
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HVeNtx--F9sE",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}