gecko_team_17_thursday.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "gecko_team_17_thursday.ipynb",
      "provenance": [],
      "toc_visible": true,
      "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/ShutingZhai/ai4ess-hackathon-2020/blob/master/gecko_team_17_thursday.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sF05XQGN3QHT"
      },
      "source": [
        "## Previous code for data preprocessing"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kEorFUGITvoY",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 357
        },
        "outputId": "f61d5a4f-a105-48ae-f53b-be87a62249ac"
      },
      "source": [
        "gpu_info = !nvidia-smi\n",
        "gpu_info = '\\n'.join(gpu_info)\n",
        "if gpu_info.find('failed') >= 0:\n",
        "  print('Select the Runtime → \"Change runtime type\" menu to enable a GPU accelerator, ')\n",
        "  print('and then re-execute this cell.')\n",
        "else:\n",
        "  print(gpu_info)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Thu Jun 25 21:18:28 2020       \n",
            "+-----------------------------------------------------------------------------+\n",
            "| NVIDIA-SMI 450.36.06    Driver Version: 418.67       CUDA Version: 10.1     |\n",
            "|-------------------------------+----------------------+----------------------+\n",
            "| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |\n",
            "| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |\n",
            "|                               |                      |               MIG M. |\n",
            "|===============================+======================+======================|\n",
            "|   0  Tesla K80           Off  | 00000000:00:04.0 Off |                    0 |\n",
            "| N/A   65C    P8    34W / 149W |      0MiB / 11441MiB |      0%      Default |\n",
            "|                               |                      |                 ERR! |\n",
            "+-------------------------------+----------------------+----------------------+\n",
            "                                                                               \n",
            "+-----------------------------------------------------------------------------+\n",
            "| Processes:                                                                  |\n",
            "|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |\n",
            "|        ID   ID                                                   Usage      |\n",
            "|=============================================================================|\n",
            "|  No running processes found                                                 |\n",
            "+-----------------------------------------------------------------------------+\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "M6P0FHlnpD9L",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 986
        },
        "outputId": "c19bb3bc-eda7-4081-ca9e-7bb55b8b1f04"
      },
      "source": [
        "# Setting up the environment\n",
        "! pip install numpy scipy matplotlib xarray pandas netcdf4 tqdm tensorflow scikit-learn seaborn"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Requirement already satisfied: numpy in /usr/local/lib/python3.6/dist-packages (1.18.5)\n",
            "Requirement already satisfied: scipy in /usr/local/lib/python3.6/dist-packages (1.4.1)\n",
            "Requirement already satisfied: matplotlib in /usr/local/lib/python3.6/dist-packages (3.2.2)\n",
            "Requirement already satisfied: xarray in /usr/local/lib/python3.6/dist-packages (0.15.1)\n",
            "Requirement already satisfied: pandas in /usr/local/lib/python3.6/dist-packages (1.0.5)\n",
            "Collecting netcdf4\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/35/4f/d49fe0c65dea4d2ebfdc602d3e3d2a45a172255c151f4497c43f6d94a5f6/netCDF4-1.5.3-cp36-cp36m-manylinux1_x86_64.whl (4.1MB)\n",
            "\u001b[K     |████████████████████████████████| 4.1MB 7.4MB/s \n",
            "\u001b[?25hRequirement already satisfied: tqdm in /usr/local/lib/python3.6/dist-packages (4.41.1)\n",
            "Requirement already satisfied: tensorflow in /usr/local/lib/python3.6/dist-packages (2.2.0)\n",
            "Requirement already satisfied: scikit-learn in /usr/local/lib/python3.6/dist-packages (0.22.2.post1)\n",
            "Requirement already satisfied: seaborn in /usr/local/lib/python3.6/dist-packages (0.10.1)\n",
            "Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib) (1.2.0)\n",
            "Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.6/dist-packages (from matplotlib) (0.10.0)\n",
            "Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib) (2.4.7)\n",
            "Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib) (2.8.1)\n",
            "Requirement already satisfied: setuptools>=41.2 in /usr/local/lib/python3.6/dist-packages (from xarray) (47.3.1)\n",
            "Requirement already satisfied: pytz>=2017.2 in /usr/local/lib/python3.6/dist-packages (from pandas) (2018.9)\n",
            "Collecting cftime\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/0f/5e/ee154e2aabb0beea0c4c7dc3d93c6f64f96a2a2019bbd05afc905439d042/cftime-1.1.3-cp36-cp36m-manylinux1_x86_64.whl (322kB)\n",
            "\u001b[K     |████████████████████████████████| 327kB 41.2MB/s \n",
            "\u001b[?25hRequirement already satisfied: termcolor>=1.1.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow) (1.1.0)\n",
            "Requirement already satisfied: grpcio>=1.8.6 in /usr/local/lib/python3.6/dist-packages (from tensorflow) (1.29.0)\n",
            "Requirement already satisfied: tensorboard<2.3.0,>=2.2.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow) (2.2.2)\n",
            "Requirement already satisfied: google-pasta>=0.1.8 in /usr/local/lib/python3.6/dist-packages (from tensorflow) (0.2.0)\n",
            "Requirement already satisfied: keras-preprocessing>=1.1.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow) (1.1.2)\n",
            "Requirement already satisfied: h5py<2.11.0,>=2.10.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow) (2.10.0)\n",
            "Requirement already satisfied: absl-py>=0.7.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow) (0.9.0)\n",
            "Requirement already satisfied: astunparse==1.6.3 in /usr/local/lib/python3.6/dist-packages (from tensorflow) (1.6.3)\n",
            "Requirement already satisfied: tensorflow-estimator<2.3.0,>=2.2.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow) (2.2.0)\n",
            "Requirement already satisfied: protobuf>=3.8.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow) (3.10.0)\n",
            "Requirement already satisfied: six>=1.12.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow) (1.12.0)\n",
            "Requirement already satisfied: wheel>=0.26; python_version >= \"3\" in /usr/local/lib/python3.6/dist-packages (from tensorflow) (0.34.2)\n",
            "Requirement already satisfied: wrapt>=1.11.1 in /usr/local/lib/python3.6/dist-packages (from tensorflow) (1.12.1)\n",
            "Requirement already satisfied: gast==0.3.3 in /usr/local/lib/python3.6/dist-packages (from tensorflow) (0.3.3)\n",
            "Requirement already satisfied: opt-einsum>=2.3.2 in /usr/local/lib/python3.6/dist-packages (from tensorflow) (3.2.1)\n",
            "Requirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.6/dist-packages (from scikit-learn) (0.15.1)\n",
            "Requirement already satisfied: google-auth<2,>=1.6.3 in /usr/local/lib/python3.6/dist-packages (from tensorboard<2.3.0,>=2.2.0->tensorflow) (1.17.2)\n",
            "Requirement already satisfied: google-auth-oauthlib<0.5,>=0.4.1 in /usr/local/lib/python3.6/dist-packages (from tensorboard<2.3.0,>=2.2.0->tensorflow) (0.4.1)\n",
            "Requirement already satisfied: tensorboard-plugin-wit>=1.6.0 in /usr/local/lib/python3.6/dist-packages (from tensorboard<2.3.0,>=2.2.0->tensorflow) (1.6.0.post3)\n",
            "Requirement already satisfied: markdown>=2.6.8 in /usr/local/lib/python3.6/dist-packages (from tensorboard<2.3.0,>=2.2.0->tensorflow) (3.2.2)\n",
            "Requirement already satisfied: requests<3,>=2.21.0 in /usr/local/lib/python3.6/dist-packages (from tensorboard<2.3.0,>=2.2.0->tensorflow) (2.23.0)\n",
            "Requirement already satisfied: werkzeug>=0.11.15 in /usr/local/lib/python3.6/dist-packages (from tensorboard<2.3.0,>=2.2.0->tensorflow) (1.0.1)\n",
            "Requirement already satisfied: cachetools<5.0,>=2.0.0 in /usr/local/lib/python3.6/dist-packages (from google-auth<2,>=1.6.3->tensorboard<2.3.0,>=2.2.0->tensorflow) (4.1.0)\n",
            "Requirement already satisfied: rsa<5,>=3.1.4; python_version >= \"3\" in /usr/local/lib/python3.6/dist-packages (from google-auth<2,>=1.6.3->tensorboard<2.3.0,>=2.2.0->tensorflow) (4.6)\n",
            "Requirement already satisfied: pyasn1-modules>=0.2.1 in /usr/local/lib/python3.6/dist-packages (from google-auth<2,>=1.6.3->tensorboard<2.3.0,>=2.2.0->tensorflow) (0.2.8)\n",
            "Requirement already satisfied: requests-oauthlib>=0.7.0 in /usr/local/lib/python3.6/dist-packages (from google-auth-oauthlib<0.5,>=0.4.1->tensorboard<2.3.0,>=2.2.0->tensorflow) (1.3.0)\n",
            "Requirement already satisfied: importlib-metadata; python_version < \"3.8\" in /usr/local/lib/python3.6/dist-packages (from markdown>=2.6.8->tensorboard<2.3.0,>=2.2.0->tensorflow) (1.6.1)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.6/dist-packages (from requests<3,>=2.21.0->tensorboard<2.3.0,>=2.2.0->tensorflow) (2020.4.5.2)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.6/dist-packages (from requests<3,>=2.21.0->tensorboard<2.3.0,>=2.2.0->tensorflow) (3.0.4)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.6/dist-packages (from requests<3,>=2.21.0->tensorboard<2.3.0,>=2.2.0->tensorflow) (1.24.3)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.6/dist-packages (from requests<3,>=2.21.0->tensorboard<2.3.0,>=2.2.0->tensorflow) (2.9)\n",
            "Requirement already satisfied: pyasn1>=0.1.3 in /usr/local/lib/python3.6/dist-packages (from rsa<5,>=3.1.4; python_version >= \"3\"->google-auth<2,>=1.6.3->tensorboard<2.3.0,>=2.2.0->tensorflow) (0.4.8)\n",
            "Requirement already satisfied: oauthlib>=3.0.0 in /usr/local/lib/python3.6/dist-packages (from requests-oauthlib>=0.7.0->google-auth-oauthlib<0.5,>=0.4.1->tensorboard<2.3.0,>=2.2.0->tensorflow) (3.1.0)\n",
            "Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.6/dist-packages (from importlib-metadata; python_version < \"3.8\"->markdown>=2.6.8->tensorboard<2.3.0,>=2.2.0->tensorflow) (3.1.0)\n",
            "Installing collected packages: cftime, netcdf4\n",
            "Successfully installed cftime-1.1.3 netcdf4-1.5.3\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VAoJD5AipD9Q",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        },
        "outputId": "606e8941-c17a-4456-b83e-b2d6f02c9e75"
      },
      "source": [
        "import s3fs\n",
        "import seaborn as sns\n",
        "import pandas as pd\n",
        "import pyarrow\n",
        "import re\n",
        "import random\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "from glob import glob\n",
        "from os.path import join, exists\n",
        "from matplotlib.colors import LogNorm\n",
        "import tensorflow as tf\n",
        "from sklearn.metrics import mean_squared_error\n",
        "from sklearn.preprocessing import StandardScaler, MinMaxScaler\n",
        "from tensorflow.keras.layers import Dense, Input\n",
        "from tensorflow.keras.models import Model\n",
        "from sklearn.model_selection import GroupKFold\n",
        "from tensorflow.keras.optimizers import Adam\n",
        "%matplotlib inline\n",
        "\n",
        "seed = 8886\n",
        "np.random.seed(seed)\n",
        "random.seed(seed)\n",
        "tf.random.set_seed(seed)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/statsmodels/tools/_testing.py:19: FutureWarning: pandas.util.testing is deprecated. Use the functions in the public API at pandas.testing instead.\n",
            "  import pandas.util.testing as tm\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "W389QzhRpD9U"
      },
      "source": [
        "# Load functions. Copied from the instruction notebook.\n",
        "# -----------------------------------------------\n",
        "\"\"\" Data Processing \"\"\"\n",
        "\n",
        "def fetch_data(f):\n",
        "    \"\"\" Load data directly from Amazon S3 storage \"\"\"\n",
        "    \n",
        "    obj = fs.open(f)\n",
        "    df = pd.read_parquet(obj)\n",
        "    return df\n",
        "\n",
        "# def prepare_data(data, input_columns, output_columns):\n",
        "#     \"\"\" Splits data to input/output with corresponding lag \"\"\"\n",
        "    \n",
        "#     data['id'] = data['id'].apply(lambda x: int(x[3:])) # change expnumber to int for sorting\n",
        "#     data.index.name = 'indx'\n",
        "#     data = data.sort_values(['id','indx'])\n",
        "#     x_data = data[input_columns].iloc[:-1, :]\n",
        "#     y_data = data[output_columns].iloc[1:, :]\n",
        "    \n",
        "#     return x_data, y_data\n",
        "\n",
        "def prepare_data(data, input_columns, output_columns):\n",
        "    \"\"\" Splits data to input/output with corresponding lag \"\"\"\n",
        "    data['id'] = data['id'].apply(lambda x: int(x[3:])) # change expnumber to int for sorting\n",
        "    data.index.name = 'indx'\n",
        "    data = data.sort_values(['id','indx'])\n",
        "    # lag in/out by experiment\n",
        "    mask = data['Time [s]'] != data['Time [s]'].max()\n",
        "    x_data = data.loc[mask, input_columns]\n",
        "    mask = data['Time [s]'] != data['Time [s]'].min()\n",
        "    y_data = data.loc[mask, output_columns]\n",
        "    return x_data, y_data\n",
        "\n",
        "def get_starting_conds(data, input_vars, exp, starting_ts=0):\n",
        "    \"\"\" Take data and expiriment number to gather initial starting condition for box emulator\"\"\"\n",
        "    exp_data = data[data['id'] == exp]\n",
        "    \n",
        "    if exp_data.shape[0] == 0:\n",
        "        raise ValueError('That expiriment number does not exist in this dataset.')\n",
        "    \n",
        "    ts_data = exp_data.iloc[starting_ts:starting_ts+1, :]\n",
        "\n",
        "    return ts_data\n",
        "# -----------------------------------------------\n",
        "\"\"\" Evaluation \"\"\"\n",
        "\n",
        "def calc_pdf_hist(x, x_bins):\n",
        "    return np.histogram(x, x_bins, density=True)[0]\n",
        "\n",
        "def rmse(y_true, y_pred):\n",
        "    return np.sqrt(mean_squared_error(y_true, y_pred))\n",
        "\n",
        "def hellinger(x, pdf_p, pdf_q):\n",
        "    pdf_distances = (np.sqrt(pdf_p) - np.sqrt(pdf_q)) ** 2\n",
        "    return np.trapz(pdf_distances, x) / 2\n",
        "\n",
        "def hellinger_distance(y_true, y_pred, bins=50):\n",
        "    bin_points = np.linspace(np.minimum(y_true.min(), y_pred.min()),\n",
        "                       np.maximum(y_true.max(), y_pred.max()),\n",
        "                       bins)\n",
        "    bin_centers = 0.5 * (bin_points[:-1] + bin_points[1:])\n",
        "    y_true_pdf = calc_pdf_hist(y_true, bin_points)\n",
        "    y_pred_pdf = calc_pdf_hist(y_pred, bin_points)\n",
        "    return hellinger(bin_centers, y_true_pdf, y_pred_pdf)\n",
        "\n",
        "def r2_corr(y_true, y_pred):\n",
        "    return np.corrcoef(y_true, y_pred)[0, 1] ** 2\n",
        "\n",
        "def evaluate_mod(true_output, model_results):\n",
        "    \n",
        "    print('RMSE: Precursor: {:.5f}, Gas: {:.5f}, Aerosols: {:.5f}'.format(\n",
        "        rmse(true_output.iloc[:,0], model_results.iloc[:,0]),\n",
        "        rmse(true_output.iloc[:,1], model_results.iloc[:,1]),\n",
        "        rmse(true_output.iloc[:,2], model_results.iloc[:,2])))\n",
        "    print('R2: Precursor: {:.5f}, Gas: {:.5f}, Aerosols: {:.5f}'.format(\n",
        "        r2_corr(true_output.iloc[:,0], model_results.iloc[:,0]),\n",
        "        r2_corr(true_output.iloc[:,1], model_results.iloc[:,1]),\n",
        "        r2_corr(true_output.iloc[:,2], model_results.iloc[:,2])))\n",
        "    print('Hellenger Distance: Precursor: {:.5f}, Gas: {:.5f}, Aerosols: {:.5f}'.format(\n",
        "        hellinger_distance(true_output.iloc[:,0], model_results.iloc[:,0]),\n",
        "        hellinger_distance(true_output.iloc[:,1], model_results.iloc[:,1]),\n",
        "        hellinger_distance(true_output.iloc[:,2], model_results.iloc[:,2])))\n",
        "    \n",
        "    return\n",
        "# -----------------------------------------------\n",
        "\"\"\" Models \"\"\"\n",
        "def dense_neural_net(num_layers=2, num_neurons=100, activation=\"relu\", learning_rate=0.001,\n",
        "                     num_input_columns=9, num_output_columns=3):\n",
        "    \"\"\" Build and return dense neural network with specified hyperparameters\"\"\"\n",
        "    \n",
        "    input_layer = Input(num_input_columns)\n",
        "    n_net = input_layer\n",
        "    for l in range(num_layers):\n",
        "        n_net= Dense(num_neurons, activation=activation)(n_net)\n",
        "    out = Dense(num_output_columns)(n_net)\n",
        "    mod = Model(input_layer, out)\n",
        "    mod.compile(Adam(learning_rate=learning_rate), \"mse\")\n",
        "    \n",
        "    return mod\n",
        "\n",
        "def box_emulator(neural_network, starting_conds, input_scaler, output_scaler, num_timesteps=1439):\n",
        "    \"\"\" Forward box emulator\n",
        "    Args:\n",
        "        neural_netwrok: NN model object that predits output at t+1\n",
        "        starting_conds: Initial input conditions used for first prediction\n",
        "        input_scaler: Input scaler object\n",
        "        onput_scaler: Onput scaler object\n",
        "        num_timesteps: how many timesteps forward to run emulator\n",
        "    Returns:\n",
        "        results: Pandas DataFrame of box emulator output\n",
        "    \"\"\"\n",
        "    mod = neural_network\n",
        "    scaled_input = input_scaler.transform(starting_conds.iloc[:,1:-1])\n",
        "    static_input = scaled_input[:,-6:]\n",
        "\n",
        "    for i in range(num_timesteps):\n",
        "\n",
        "        if i == 0:\n",
        "\n",
        "            pred = mod.predict(scaled_input)\n",
        "            new_input = np.concatenate([pred,static_input], axis=1)\n",
        "            pred_array = pred\n",
        "\n",
        "        else:\n",
        "\n",
        "            pred = mod.predict(new_input)\n",
        "            new_input = np.concatenate([pred,static_input], axis=1)\n",
        "            pred_array = np.concatenate([pred_array, pred], axis=0)\n",
        "            \n",
        "    results = pd.DataFrame(output_scaler.inverse_transform(pred_array))\n",
        "    results['id'] = starting_conds.iloc[0,-1]\n",
        "    results.columns = starting_conds.columns[[1,2,3,-1]]\n",
        "    return results\n",
        "\n",
        "# def run_emulator_ensemble(input_data, mod, num_expiriments, input_vars, input_scaler, \n",
        "#                           output_scaler, run_length=1439):\n",
        "#     \"\"\" Run an ensemble of emulators on n amount of random expiriments\n",
        "#     Args:\n",
        "#         input_data: unscaled dataframe collection of expiriment data\n",
        "#         mod: NN model object that predits output at t+1\n",
        "#         num_expiriments: How many expiriments to run\n",
        "#         input_vars: Input variables for model\n",
        "#         output_vars: Output variables for model\n",
        "#         run_length: number of timesteps to run each emulator forward\n",
        "#     Returns:\n",
        "#         pred_df: Aggregated Pandas DataFrame of predictions for each expiriment\n",
        "#         true_df: Corresponding Dataframe with raw expiriment data\n",
        "#     \"\"\"\n",
        "    output_cols = input_data.columns[[1,2,3, -1]]\n",
        "    exps = sorted(random.sample(list(input_data['id'].unique()), num_expiriments))\n",
        "    pred_df  = pd.DataFrame(columns = output_cols)\n",
        "    true_df = input_data.loc[input_data['id'].isin(exps)][output_cols].reset_index(drop=True)\n",
        "    true_df =  true_df.groupby('id').apply(lambda x: x.iloc[1:,:]).reset_index(drop=True)  \n",
        "    \n",
        "    for i in range(num_expiriments):\n",
        "        \n",
        "        init_conds = get_starting_conds(input_data, input_vars, exps[i])\n",
        "        preds = box_emulator(mod, init_conds, input_scaler, output_scaler, run_length)\n",
        "        preds['id'] = exps[i]\n",
        "        preds.columns = pred_df.columns\n",
        "        pred_df = pred_df.append(preds, ignore_index=True)\n",
        "\n",
        "    return pred_df, true_df\n",
        "\n",
        "def run_emulator_ensemble(input_data, output_data, mod, num_expiriments, input_vars, input_scaler, \n",
        "                          output_scaler, run_length=1439):\n",
        "    \"\"\" Run an ensemble of emulators on n amount of random expiriments\n",
        "    Args:\n",
        "        input_data: unscaled dataframe collection of expiriment data\n",
        "        mod: NN model object that predits output at t+1\n",
        "        num_expiriments: How many expiriments to run\n",
        "        input_vars: Input variables for model\n",
        "        output_vars: Output variables for model\n",
        "        run_length: number of timesteps to run each emulator forward\n",
        "    Returns:\n",
        "        pred_df: Aggregated Pandas DataFrame of predictions for each expiriment\n",
        "        true_df: Corresponding Dataframe with raw expiriment data\n",
        "    \"\"\"\n",
        "    output_cols = input_data.columns[[1,2,3, -1]]\n",
        "    exps = sorted(random.sample(list(input_data['id'].unique()), num_expiriments))\n",
        "    pred_df  = pd.DataFrame(columns = output_cols)\n",
        "    true_df = output_data.loc[output_data['id'].isin(exps)].iloc[:,1:-1].reset_index(drop=True)\n",
        "     \n",
        "    for i in range(num_expiriments):\n",
        "        init_conds = get_starting_conds(input_data, input_vars, exps[i])\n",
        "        preds = box_emulator(mod, init_conds, input_scaler, output_scaler, run_length)\n",
        "        preds['id'] = exps[i]\n",
        "        preds.columns = pred_df.columns\n",
        "        pred_df = pred_df.append(preds, ignore_index=True)\n",
        "    return pred_df, true_df"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hSrnsvGqDDH0"
      },
      "source": [
        "Load the data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "isnQ-mikpD9Z"
      },
      "source": [
        "# load data\n",
        "fs = s3fs.S3FileSystem(anon=True)\n",
        "gecko_files = fs.ls(\"ncar-aiml-data-commons/gecko/\")[1:]\n",
        "test, train, val = map(fetch_data, gecko_files)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2w-G4vZ8pD9c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "outputId": "3c98d6b8-938d-4b81-b14d-efd9c66abbc9"
      },
      "source": [
        "train.head()"
      ],
      "execution_count": null,
      "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>Time [s]</th>\n",
              "      <th>Precursor [ug/m3]</th>\n",
              "      <th>Gas [ug/m3]</th>\n",
              "      <th>Aerosol [ug_m3]</th>\n",
              "      <th>temperature (K)</th>\n",
              "      <th>solar zenith angle (degree)</th>\n",
              "      <th>pre-existing aerosols (ug/m3)</th>\n",
              "      <th>o3 (ppb)</th>\n",
              "      <th>nox (ppb)</th>\n",
              "      <th>oh (10^6 molec/cm3)</th>\n",
              "      <th>id</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>1.000000</td>\n",
              "      <td>0.069675</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>264.135037</td>\n",
              "      <td>78.052738</td>\n",
              "      <td>0.022782</td>\n",
              "      <td>38.768707</td>\n",
              "      <td>0.206029</td>\n",
              "      <td>7.360145</td>\n",
              "      <td>Exp1666</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>301.207794</td>\n",
              "      <td>0.067673</td>\n",
              "      <td>0.002783</td>\n",
              "      <td>0.000001</td>\n",
              "      <td>264.135037</td>\n",
              "      <td>78.052738</td>\n",
              "      <td>0.022782</td>\n",
              "      <td>38.768707</td>\n",
              "      <td>0.206029</td>\n",
              "      <td>7.360145</td>\n",
              "      <td>Exp1666</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>601.415588</td>\n",
              "      <td>0.065728</td>\n",
              "      <td>0.005749</td>\n",
              "      <td>0.000004</td>\n",
              "      <td>264.135037</td>\n",
              "      <td>78.052738</td>\n",
              "      <td>0.022782</td>\n",
              "      <td>38.768707</td>\n",
              "      <td>0.206029</td>\n",
              "      <td>7.360145</td>\n",
              "      <td>Exp1666</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>901.623352</td>\n",
              "      <td>0.063838</td>\n",
              "      <td>0.008627</td>\n",
              "      <td>0.000008</td>\n",
              "      <td>264.135037</td>\n",
              "      <td>78.052738</td>\n",
              "      <td>0.022782</td>\n",
              "      <td>38.768707</td>\n",
              "      <td>0.206029</td>\n",
              "      <td>7.360145</td>\n",
              "      <td>Exp1666</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>1201.831177</td>\n",
              "      <td>0.062003</td>\n",
              "      <td>0.011420</td>\n",
              "      <td>0.000013</td>\n",
              "      <td>264.135037</td>\n",
              "      <td>78.052738</td>\n",
              "      <td>0.022782</td>\n",
              "      <td>38.768707</td>\n",
              "      <td>0.206029</td>\n",
              "      <td>7.360145</td>\n",
              "      <td>Exp1666</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "      Time [s]  Precursor [ug/m3]  ...  oh (10^6 molec/cm3)       id\n",
              "0     1.000000           0.069675  ...             7.360145  Exp1666\n",
              "1   301.207794           0.067673  ...             7.360145  Exp1666\n",
              "2   601.415588           0.065728  ...             7.360145  Exp1666\n",
              "3   901.623352           0.063838  ...             7.360145  Exp1666\n",
              "4  1201.831177           0.062003  ...             7.360145  Exp1666\n",
              "\n",
              "[5 rows x 11 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "a_G4mSE_pD9f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 297
        },
        "outputId": "242c57ce-6419-4ea2-bd9d-1fdb886addb7"
      },
      "source": [
        "# get some summary statistics of the training data\n",
        "train.groupby('id').mean().describe()"
      ],
      "execution_count": null,
      "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>Time [s]</th>\n",
              "      <th>Precursor [ug/m3]</th>\n",
              "      <th>Gas [ug/m3]</th>\n",
              "      <th>Aerosol [ug_m3]</th>\n",
              "      <th>temperature (K)</th>\n",
              "      <th>solar zenith angle (degree)</th>\n",
              "      <th>pre-existing aerosols (ug/m3)</th>\n",
              "      <th>o3 (ppb)</th>\n",
              "      <th>nox (ppb)</th>\n",
              "      <th>oh (10^6 molec/cm3)</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>count</th>\n",
              "      <td>1.400000e+03</td>\n",
              "      <td>1400.000000</td>\n",
              "      <td>1400.000000</td>\n",
              "      <td>1400.000000</td>\n",
              "      <td>1400.000000</td>\n",
              "      <td>1400.000000</td>\n",
              "      <td>1400.000000</td>\n",
              "      <td>1400.000000</td>\n",
              "      <td>1400.000000</td>\n",
              "      <td>1400.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>mean</th>\n",
              "      <td>2.160005e+05</td>\n",
              "      <td>0.003125</td>\n",
              "      <td>0.027601</td>\n",
              "      <td>0.032703</td>\n",
              "      <td>277.413314</td>\n",
              "      <td>44.923263</td>\n",
              "      <td>11.388653</td>\n",
              "      <td>75.422052</td>\n",
              "      <td>2.124920</td>\n",
              "      <td>5.493831</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>std</th>\n",
              "      <td>2.329138e-10</td>\n",
              "      <td>0.002215</td>\n",
              "      <td>0.013710</td>\n",
              "      <td>0.020581</td>\n",
              "      <td>21.712817</td>\n",
              "      <td>26.064725</td>\n",
              "      <td>21.082846</td>\n",
              "      <td>43.495636</td>\n",
              "      <td>2.468413</td>\n",
              "      <td>2.582480</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>min</th>\n",
              "      <td>2.160005e+05</td>\n",
              "      <td>0.001244</td>\n",
              "      <td>0.004393</td>\n",
              "      <td>0.001704</td>\n",
              "      <td>240.024774</td>\n",
              "      <td>0.018612</td>\n",
              "      <td>0.010031</td>\n",
              "      <td>1.017131</td>\n",
              "      <td>0.100186</td>\n",
              "      <td>1.011878</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>25%</th>\n",
              "      <td>2.160005e+05</td>\n",
              "      <td>0.001598</td>\n",
              "      <td>0.016943</td>\n",
              "      <td>0.015331</td>\n",
              "      <td>258.824046</td>\n",
              "      <td>22.235231</td>\n",
              "      <td>0.099235</td>\n",
              "      <td>36.417272</td>\n",
              "      <td>0.314910</td>\n",
              "      <td>3.231017</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>50%</th>\n",
              "      <td>2.160005e+05</td>\n",
              "      <td>0.002253</td>\n",
              "      <td>0.024975</td>\n",
              "      <td>0.028837</td>\n",
              "      <td>277.096217</td>\n",
              "      <td>44.858750</td>\n",
              "      <td>1.085966</td>\n",
              "      <td>76.044879</td>\n",
              "      <td>1.004467</td>\n",
              "      <td>5.479992</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>75%</th>\n",
              "      <td>2.160005e+05</td>\n",
              "      <td>0.003803</td>\n",
              "      <td>0.035183</td>\n",
              "      <td>0.048352</td>\n",
              "      <td>296.098148</td>\n",
              "      <td>68.039551</td>\n",
              "      <td>10.946353</td>\n",
              "      <td>112.534705</td>\n",
              "      <td>3.059370</td>\n",
              "      <td>7.758317</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>max</th>\n",
              "      <td>2.160005e+05</td>\n",
              "      <td>0.012054</td>\n",
              "      <td>0.079415</td>\n",
              "      <td>0.086156</td>\n",
              "      <td>314.980647</td>\n",
              "      <td>89.966654</td>\n",
              "      <td>99.280288</td>\n",
              "      <td>149.927193</td>\n",
              "      <td>9.980638</td>\n",
              "      <td>9.999925</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "           Time [s]  Precursor [ug/m3]  ...    nox (ppb)  oh (10^6 molec/cm3)\n",
              "count  1.400000e+03        1400.000000  ...  1400.000000          1400.000000\n",
              "mean   2.160005e+05           0.003125  ...     2.124920             5.493831\n",
              "std    2.329138e-10           0.002215  ...     2.468413             2.582480\n",
              "min    2.160005e+05           0.001244  ...     0.100186             1.011878\n",
              "25%    2.160005e+05           0.001598  ...     0.314910             3.231017\n",
              "50%    2.160005e+05           0.002253  ...     1.004467             5.479992\n",
              "75%    2.160005e+05           0.003803  ...     3.059370             7.758317\n",
              "max    2.160005e+05           0.012054  ...     9.980638             9.999925\n",
              "\n",
              "[8 rows x 10 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nWKno6IFpD9o"
      },
      "source": [
        "# Data Preperation\n",
        "'''\n",
        "Copied from the instruction notebook\n",
        "\n",
        "Here we prepare the data for machine learning by taking the respective variables \n",
        "from each column, offsetting the output data by 1 timestep (this is done inside \n",
        "the `prepare_data()` function, and scale. Make sure to not re-fit the scaler on \n",
        "the validation/test data and only `transform()` it. \n",
        "'''\n",
        "\n",
        "input_vars, output_vars = train.columns, train.columns[[0,1,2,3,-1]]\n",
        "train_in, train_out = prepare_data(train, input_vars, output_vars)\n",
        "val_in, val_out = prepare_data(val, input_vars, output_vars)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "N_kGntYI3aGI"
      },
      "source": [
        "Data scaling"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "W28hlxCBRHSp"
      },
      "source": [
        "# Min max scaling\n",
        "input_scaler = MinMaxScaler()\n",
        "output_scaler = MinMaxScaler()\n",
        "\n",
        "scaled_train_in = input_scaler.fit_transform(train_in.iloc[:,1:-1])\n",
        "scaled_train_out = output_scaler.fit_transform(train_out.iloc[:,1:-1])\n",
        "scaled_val_in = input_scaler.transform(val_in.iloc[:,1:-1])\n",
        "scaled_val_out = output_scaler.transform(val_out.iloc[:,1:-1])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "G6zjJbYmpD9r",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "dc540c19-fcab-4ab3-c9b2-87c117462a82"
      },
      "source": [
        "scaled_train_in.shape, scaled_train_out.shape, scaled_val_in.shape, scaled_val_out.shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((2014600, 9), (2014600, 3), (287800, 9), (287800, 3))"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vNCpJdbHRYEn"
      },
      "source": [
        "# Standardizing\n",
        "input_std_scaler = StandardScaler()\n",
        "output_std_scaler = StandardScaler()\n",
        "\n",
        "scaled_std_train_in = input_std_scaler.fit_transform(train_in.iloc[:,1:-1])\n",
        "scaled_std_train_out = output_std_scaler.fit_transform(train_out.iloc[:,1:-1])\n",
        "scaled_std_val_in = input_std_scaler.transform(val_in.iloc[:,1:-1])\n",
        "scaled_std_val_out = output_std_scaler.transform(val_out.iloc[:,1:-1])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MGWhBlVjRq7w",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "14951692-4f53-4e51-d932-ae11efcefb38"
      },
      "source": [
        "scaled_std_train_in.shape, scaled_std_train_out.shape, scaled_std_val_in.shape, scaled_std_val_out.shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((2014600, 9), (2014600, 3), (287800, 9), (287800, 3))"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 12
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aFIWKg70VVbF"
      },
      "source": [
        "## Baseline model and previous ML model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0E_ch8ADpD9x",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 493
        },
        "outputId": "996fc040-c8bf-40f1-957d-72f375eada07"
      },
      "source": [
        "%%time\n",
        "tf.random.set_seed(seed)\n",
        "mod = dense_neural_net()\n",
        "history = mod.fit(scaled_train_in, scaled_train_out, validation_data=(scaled_val_in, scaled_val_out), \n",
        "                  batch_size=256, epochs=5, verbose=1)\n",
        "mod.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/5\n",
            "7870/7870 [==============================] - 22s 3ms/step - loss: 2.7214e-04 - val_loss: 4.8278e-06\n",
            "Epoch 2/5\n",
            "7870/7870 [==============================] - 22s 3ms/step - loss: 3.1891e-06 - val_loss: 2.6881e-06\n",
            "Epoch 3/5\n",
            "7870/7870 [==============================] - 22s 3ms/step - loss: 2.2360e-06 - val_loss: 9.9619e-07\n",
            "Epoch 4/5\n",
            "7870/7870 [==============================] - 22s 3ms/step - loss: 1.7544e-06 - val_loss: 2.2398e-06\n",
            "Epoch 5/5\n",
            "7870/7870 [==============================] - 22s 3ms/step - loss: 1.4408e-06 - val_loss: 1.2054e-06\n",
            "Model: \"model\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_1 (InputLayer)         [(None, 9)]               0         \n",
            "_________________________________________________________________\n",
            "dense_3 (Dense)              (None, 100)               1000      \n",
            "_________________________________________________________________\n",
            "dense_4 (Dense)              (None, 100)               10100     \n",
            "_________________________________________________________________\n",
            "dense_5 (Dense)              (None, 3)                 303       \n",
            "=================================================================\n",
            "Total params: 11,403\n",
            "Trainable params: 11,403\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "CPU times: user 2min 9s, sys: 14.7 s, total: 2min 23s\n",
            "Wall time: 1min 50s\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qJhqQmgJ8s1f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 85
        },
        "outputId": "b95e594e-9647-4259-b03c-59ed3c067847"
      },
      "source": [
        "# get predictions and truth\n",
        "\n",
        "# pred_nn = pd.DataFrame(output_scaler.inverse_transform(mod.predict(scaled_val_in))).iloc[1:,:]\n",
        "# true_val = val_out.iloc[:-1,1:4]\n",
        "\n",
        "pred_nn = pd.DataFrame(output_scaler.inverse_transform(mod.predict(scaled_val_in)))\n",
        "true_val = val_out.iloc[:,1:-1]\n",
        "\n",
        "print('Metrics for base model:')\n",
        "evaluate_mod(true_val, pred_nn)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Metrics for base model:\n",
            "RMSE: Precursor: 0.00003, Gas: 0.00018, Aerosols: 0.00009\n",
            "R2: Precursor: 1.00000, Gas: 0.99999, Aerosols: 0.99999\n",
            "Hellenger Distance: Precursor: 0.00000, Gas: 0.00009, Aerosols: 0.00005\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4jjtwYpfuoGk",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 71
        },
        "outputId": "c2ad1f8b-87be-4947-8d4f-88d60583d58d"
      },
      "source": [
        "output_columns = ['Precursor [ug/m3]', 'Gas [ug/m3]', 'Aerosol [ug_m3]']\n",
        "pred_nn.columns = output_columns\n",
        "\n",
        "pred_train_nn = pd.DataFrame(output_scaler.inverse_transform(mod.predict(scaled_train_in)))\n",
        "pred_train_nn.columns = output_columns"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:5: UserWarning: Pandas doesn't allow columns to be created via a new attribute name - see https://pandas.pydata.org/pandas-docs/stable/indexing.html#attribute-access\n",
            "  \"\"\"\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MNmhPoBiVjFr"
      },
      "source": [
        "Baseline model with standardized input."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JzTjznlf8g4N",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 493
        },
        "outputId": "36fd8d8a-75e7-4833-f08c-4dd53175ad5e"
      },
      "source": [
        "%%time\n",
        "tf.random.set_seed(seed)\n",
        "mod_std = dense_neural_net()\n",
        "history = mod_std.fit(scaled_train_in, scaled_train_out, validation_data=(scaled_std_val_in, scaled_std_val_out), \n",
        "                  batch_size=256, epochs=5, verbose=1)\n",
        "mod_std.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/5\n",
            "7870/7870 [==============================] - 27s 3ms/step - loss: 2.7206e-04 - val_loss: 0.1394\n",
            "Epoch 2/5\n",
            "7870/7870 [==============================] - 26s 3ms/step - loss: 3.2016e-06 - val_loss: 0.1326\n",
            "Epoch 3/5\n",
            "7870/7870 [==============================] - 26s 3ms/step - loss: 2.2764e-06 - val_loss: 0.1289\n",
            "Epoch 4/5\n",
            "7870/7870 [==============================] - 26s 3ms/step - loss: 1.7520e-06 - val_loss: 0.1266\n",
            "Epoch 5/5\n",
            "7870/7870 [==============================] - 27s 3ms/step - loss: 1.4404e-06 - val_loss: 0.1260\n",
            "Model: \"model_2\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_3 (InputLayer)         [(None, 9)]               0         \n",
            "_________________________________________________________________\n",
            "dense_43 (Dense)             (None, 100)               1000      \n",
            "_________________________________________________________________\n",
            "dense_44 (Dense)             (None, 100)               10100     \n",
            "_________________________________________________________________\n",
            "dense_45 (Dense)             (None, 3)                 303       \n",
            "=================================================================\n",
            "Total params: 11,403\n",
            "Trainable params: 11,403\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "CPU times: user 2min 32s, sys: 17.8 s, total: 2min 49s\n",
            "Wall time: 2min 12s\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Yr-sfUfa8wKn",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 85
        },
        "outputId": "79b5e368-5218-4dde-889b-6fbceb71cf43"
      },
      "source": [
        "# get predictions and truth\n",
        "\n",
        "# pred_nn = pd.DataFrame(output_std_scaler.inverse_transform(mod.predict(scaled_std_val_in))).iloc[1:,:]\n",
        "# true_val = val_out.iloc[:-1,1:4]\n",
        "\n",
        "pred_nn_std = pd.DataFrame(output_std_scaler.inverse_transform(mod_std.predict(scaled_std_val_in)))\n",
        "true_val = val_out.iloc[:,1:-1]\n",
        "\n",
        "print('Metrics for base model:')\n",
        "evaluate_mod(true_val, pred_nn_std)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Metrics for base model:\n",
            "RMSE: Precursor: 0.00246, Gas: 0.00952, Aerosols: 0.00884\n",
            "R2: Precursor: 0.98158, Gas: 0.95517, Aerosols: 0.95286\n",
            "Hellenger Distance: Precursor: 0.26690, Gas: 0.18135, Aerosols: 0.09445\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1ZBhp0pCpD-I"
      },
      "source": [
        "## Evaluate two interpretation methods for your machine learning solution"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nDs8lBeCDYW7"
      },
      "source": [
        "### Models to evaluate"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "u6bS7t_oiIrp"
      },
      "source": [
        "The RNN Model to evaluate:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KrpICY7zf4ak",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 493
        },
        "outputId": "f561064e-d7b6-45a7-fb90-654d7304ec37"
      },
      "source": [
        "from tensorflow.keras.layers import LSTM, SimpleRNN\n",
        "from tensorflow.keras.models import Sequential\n",
        "\n",
        "tf.random.set_seed(seed)\n",
        "\n",
        "# Transform the model for use in RNN\n",
        "# previously applied transformer: output_std_scaler\n",
        "rnn_std_train_in = scaled_std_train_in.reshape(-1,1,9)\n",
        "rnn_std_train_out = scaled_std_train_out.reshape(-1,1,3)\n",
        "rnn_std_val_in = scaled_std_val_in.reshape(-1,1,9)\n",
        "rnn_std_val_out = scaled_std_val_out.reshape(-1,1,3)\n",
        "\n",
        "# Set up the model\n",
        "mod_rnn_std_2 = Sequential()\n",
        "mod_rnn_std_2.add(SimpleRNN(128, input_shape=(1, 9), return_sequences=True))\n",
        "mod_rnn_std_2.add(LSTM(256, return_sequences=True))\n",
        "mod_rnn_std_2.add(Dense(3))\n",
        "mod_rnn_std_2.compile(Adam(learning_rate=0.001), \"mse\")\n",
        "\n",
        "# fit the model\n",
        "history = mod_rnn_std_2.fit(rnn_std_train_in, rnn_std_train_out, \n",
        "                      validation_data=(rnn_std_val_in, rnn_std_val_out), \n",
        "                      batch_size=256, epochs=5, verbose=1, workers=-1)\n",
        "mod_rnn_std_2.summary()\n",
        "\n",
        "# get predictions and truth\n",
        "pred_rnn_std_reshape_2 = mod_rnn_std_2.predict(rnn_std_val_in).reshape(scaled_std_val_out.shape)\n",
        "pred_rnn_std_2 = pd.DataFrame(output_std_scaler.inverse_transform(pred_rnn_std_reshape_2))\n",
        "true_val = val_out.iloc[:,1:-1]\n",
        "\n",
        "print('Metrics for LSTM model:')\n",
        "evaluate_mod(true_val, pred_rnn_std_2)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/5\n",
            "7870/7870 [==============================] - 55s 7ms/step - loss: 0.0013 - val_loss: 1.1885e-04\n",
            "Epoch 2/5\n",
            "7870/7870 [==============================] - 55s 7ms/step - loss: 2.4182e-05 - val_loss: 1.6885e-05\n",
            "Epoch 3/5\n",
            "7870/7870 [==============================] - 54s 7ms/step - loss: 1.9606e-05 - val_loss: 1.0075e-05\n",
            "Epoch 4/5\n",
            "7870/7870 [==============================] - 55s 7ms/step - loss: 1.4716e-05 - val_loss: 1.0276e-05\n",
            "Epoch 5/5\n",
            "7870/7870 [==============================] - 54s 7ms/step - loss: 1.3341e-05 - val_loss: 1.0726e-05\n",
            "Model: \"sequential_2\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "simple_rnn_2 (SimpleRNN)     (None, 1, 128)            17664     \n",
            "_________________________________________________________________\n",
            "lstm (LSTM)                  (None, 1, 256)            394240    \n",
            "_________________________________________________________________\n",
            "dense_2 (Dense)              (None, 1, 3)              771       \n",
            "=================================================================\n",
            "Total params: 412,675\n",
            "Trainable params: 412,675\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "Metrics for LSTM model:\n",
            "RMSE: Precursor: 0.00180, Gas: 0.00042, Aerosols: 0.00100\n",
            "R2: Precursor: 0.96539, Gas: 0.99971, Aerosols: 0.99795\n",
            "Hellenger Distance: Precursor: 0.00000, Gas: 0.00001, Aerosols: 0.00005\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BAUAvBfUtYDn"
      },
      "source": [
        "output_columns = ['Precursor [ug/m3]', 'Gas [ug/m3]', 'Aerosol [ug_m3]']\n",
        "pred_rnn_std_2.columns = output_columns\n",
        "\n",
        "pred_train_rnn_std_reshape_2 = mod_rnn_std_2.predict(rnn_std_train_in).reshape(scaled_std_train_out.shape)\n",
        "pred_train_rnn_std_2 = pd.DataFrame(output_std_scaler.inverse_transform(pred_train_rnn_std_reshape_2))\n",
        "\n",
        "pred_train_rnn_std_2.columns = output_columns"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Rk4ow6haixga"
      },
      "source": [
        "Linear model to evaluate"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LjSuLhSsc2MC",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 85
        },
        "outputId": "7b0cff05-9928-4d3f-d0a5-a18c9eb37182"
      },
      "source": [
        "# from sklearn import linear_model\n",
        "# lmod = linear_model.LinearRegression().fit(scaled_std_train_in, scaled_std_train_out)\n",
        "\n",
        "# # get predictions and truth\n",
        "# pred_lmod_raw = lmod.predict(scaled_std_val_in)\n",
        "# pred_lmod = pd.DataFrame(output_std_scaler.inverse_transform(pred_rnn_std_reshape_8))\n",
        "# true_val = val_out.iloc[:,1:-1]\n",
        "\n",
        "# print('Metrics for Linear model:')\n",
        "# evaluate_mod(true_val, pred_lmod)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Metrics for Linear model:\n",
            "RMSE: Precursor: 0.00084, Gas: 0.00069, Aerosols: 0.00092\n",
            "R2: Precursor: 0.99244, Gas: 0.99922, Aerosols: 0.99827\n",
            "Hellenger Distance: Precursor: 0.00097, Gas: 0.00068, Aerosols: 0.00232\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "phzogMrpjd6F"
      },
      "source": [
        "index_columns = ['Time [s]', 'id']\n",
        "output_columns = ['Precursor [ug/m3]', 'Gas [ug/m3]', 'Aerosol [ug_m3]']\n",
        "input_columns = list(set(train.columns) - set(index_columns))\n",
        "\n",
        "def train_model():\n",
        "    est= {}\n",
        "    msqe = {}\n",
        "    for j, col in enumerate(['Precursor [ug/m3]', 'Gas [ug/m3]', 'Aerosol [ug_m3]']):\n",
        "        e = linear_model.LinearRegression().fit(scaled_std_train_in, scaled_std_train_out[:,j])\n",
        "#         e = svm.LinearSVR(tol=0.9).fit(train_in_scaled, train_out_scaled[:,j])\n",
        "        msqe[col] = mean_squared_error(scaled_std_val_out[:,j], e.predict(scaled_std_val_in))\n",
        "        est[col] = e\n",
        "    return msqe, est"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ypzWdt7dpb3s"
      },
      "source": [
        "msqe, est = train_model()\n",
        "\n",
        "fit_train = {col: est[col].predict(scaled_std_train_in) for col in output_columns}\n",
        "fit_val = {col: est[col].predict(scaled_std_val_in) for col in output_columns}\n",
        "train_out_arr = np.concatenate([fit_train[col][:,np.newaxis] for col in output_columns], axis=-1)\n",
        "val_out_arr = np.concatenate([fit_val[col][:,np.newaxis] for col in output_columns], axis=-1)\n",
        "linear_mod_train = pd.DataFrame(output_std_scaler.inverse_transform(train_out_arr), columns=output_columns)\n",
        "linear_mod_val = pd.DataFrame(output_std_scaler.inverse_transform(val_out_arr), columns=output_columns)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "64dGg0_S5oTJ"
      },
      "source": [
        "Random Forest Model to evaluate"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sS5W2GwQ9eQU",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 136
        },
        "outputId": "a3fdf9e3-f821-403a-896f-5c582ea930b1"
      },
      "source": [
        "from sklearn.ensemble import RandomForestRegressor\n",
        "# Instantiate model \n",
        "model_rf = RandomForestRegressor(n_estimators=20, max_depth=5)\n",
        "# Train the model with training set\n",
        "model_rf.fit(scaled_std_train_in, scaled_std_train_out)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "RandomForestRegressor(bootstrap=True, ccp_alpha=0.0, criterion='mse',\n",
              "                      max_depth=5, max_features='auto', max_leaf_nodes=None,\n",
              "                      max_samples=None, min_impurity_decrease=0.0,\n",
              "                      min_impurity_split=None, min_samples_leaf=1,\n",
              "                      min_samples_split=2, min_weight_fraction_leaf=0.0,\n",
              "                      n_estimators=20, n_jobs=None, oob_score=False,\n",
              "                      random_state=None, verbose=0, warm_start=False)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 134
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8BYa1JX9a_50",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 139
        },
        "outputId": "00c01c5f-d531-4b1e-ec9d-4925ed77eb59"
      },
      "source": [
        "# Score\n",
        "print('Training data score: ' + str(model_rf.score(scaled_std_train_in, scaled_std_train_out)))\n",
        "print('Testing data score: ' + str(model_rf.score(scaled_std_val_in, scaled_std_val_out)))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/sklearn/base.py:434: FutureWarning: The default value of multioutput (not exposed in score method) will change from 'variance_weighted' to 'uniform_average' in 0.23 to keep consistent with 'metrics.r2_score'. To specify the default value manually and avoid the warning, please either call 'metrics.r2_score' directly or make a custom scorer with 'metrics.make_scorer' (the built-in scorer 'r2' uses multioutput='uniform_average').\n",
            "  \"multioutput='uniform_average').\", FutureWarning)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Training data score: 0.9558880993196023\n",
            "Testing data score: 0.9568028519353541\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/sklearn/base.py:434: FutureWarning: The default value of multioutput (not exposed in score method) will change from 'variance_weighted' to 'uniform_average' in 0.23 to keep consistent with 'metrics.r2_score'. To specify the default value manually and avoid the warning, please either call 'metrics.r2_score' directly or make a custom scorer with 'metrics.make_scorer' (the built-in scorer 'r2' uses multioutput='uniform_average').\n",
            "  \"multioutput='uniform_average').\", FutureWarning)\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "p3OgtLqrbaM-",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 85
        },
        "outputId": "64ab6841-0cb8-41e5-a1dc-84b311fde443"
      },
      "source": [
        "# get predictions and truth\n",
        "pred_rf = pd.DataFrame(output_std_scaler.inverse_transform(model_rf.predict(scaled_std_val_in)))\n",
        "true_val = val_out.iloc[:,1:4]\n",
        "\n",
        "print('Metrics for random forest model:')\n",
        "evaluate_mod(true_val, pred_rf)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Metrics for random forest model:\n",
            "RMSE: Precursor: 0.00135, Gas: 0.00564, Aerosols: 0.00518\n",
            "R2: Precursor: 0.98030, Gas: 0.94650, Aerosols: 0.94536\n",
            "Hellenger Distance: Precursor: 0.04319, Gas: 0.31254, Aerosols: 0.33035\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "g-mGkS145xiW"
      },
      "source": [
        "output_columns = ['Precursor [ug/m3]', 'Gas [ug/m3]', 'Aerosol [ug_m3]']\n",
        "pred_rf.columns = output_columns\n",
        "\n",
        "pred_train_rf = pd.DataFrame(output_std_scaler.inverse_transform(model_rf.predict(scaled_std_train_in)))\n",
        "pred_train_rf.columns = output_columns\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DbJOtN3l6ot7"
      },
      "source": [
        "Simple RNN"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CBdgaFWc6n7T",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 459
        },
        "outputId": "7e265962-8569-49c5-e0e6-535da36d90da"
      },
      "source": [
        "from tensorflow.keras.layers import LSTM, SimpleRNN\n",
        "from tensorflow.keras.models import Sequential\n",
        "\n",
        "tf.random.set_seed(seed)\n",
        "\n",
        "# Transform the model for use in RNN\n",
        "# previously applied transformer: output_std_scaler\n",
        "rnn_std_train_in = scaled_std_train_in.reshape(-1,1,9)\n",
        "rnn_std_train_out = scaled_std_train_out.reshape(-1,1,3)\n",
        "rnn_std_val_in = scaled_std_val_in.reshape(-1,1,9)\n",
        "rnn_std_val_out = scaled_std_val_out.reshape(-1,1,3)\n",
        "\n",
        "# Set up the model\n",
        "mod_rnn_std_simple = Sequential()\n",
        "mod_rnn_std_simple.add(SimpleRNN(256, input_shape=(1, 9), return_sequences=True))\n",
        "mod_rnn_std_simple.add(Dense(3))\n",
        "mod_rnn_std_simple.compile(Adam(learning_rate=0.001), \"mse\")\n",
        "\n",
        "# fit the model\n",
        "history = mod_rnn_std_simple.fit(rnn_std_train_in, rnn_std_train_out, \n",
        "                      validation_data=(rnn_std_val_in, rnn_std_val_out), \n",
        "                      batch_size=256, epochs=5, verbose=1, workers=-1)\n",
        "mod_rnn_std_simple.summary()\n",
        "\n",
        "# get predictions and truth\n",
        "pred_rnn_std_reshape_simple = mod_rnn_std_simple.predict(rnn_std_val_in).reshape(scaled_std_val_out.shape)\n",
        "pred_rnn_std_simple = pd.DataFrame(output_std_scaler.inverse_transform(pred_rnn_std_reshape_simple))\n",
        "true_val = val_out.iloc[:,1:-1]\n",
        "\n",
        "print('Metrics for LSTM model:')\n",
        "evaluate_mod(true_val, pred_rnn_std_simple)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/5\n",
            "7870/7870 [==============================] - 40s 5ms/step - loss: 0.0013 - val_loss: 2.6488e-05\n",
            "Epoch 2/5\n",
            "7870/7870 [==============================] - 38s 5ms/step - loss: 2.1144e-05 - val_loss: 9.6862e-06\n",
            "Epoch 3/5\n",
            "7870/7870 [==============================] - 39s 5ms/step - loss: 1.5822e-05 - val_loss: 1.4888e-05\n",
            "Epoch 4/5\n",
            "7870/7870 [==============================] - 38s 5ms/step - loss: 1.4162e-05 - val_loss: 1.2673e-05\n",
            "Epoch 5/5\n",
            "7870/7870 [==============================] - 39s 5ms/step - loss: 1.3159e-05 - val_loss: 1.3192e-05\n",
            "Model: \"sequential_3\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "simple_rnn_3 (SimpleRNN)     (None, 1, 256)            68096     \n",
            "_________________________________________________________________\n",
            "dense_6 (Dense)              (None, 1, 3)              771       \n",
            "=================================================================\n",
            "Total params: 68,867\n",
            "Trainable params: 68,867\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "Metrics for LSTM model:\n",
            "RMSE: Precursor: 0.00004, Gas: 0.00009, Aerosols: 0.00008\n",
            "R2: Precursor: 0.99999, Gas: 0.99999, Aerosols: 0.99999\n",
            "Hellenger Distance: Precursor: 0.00000, Gas: 0.00000, Aerosols: 0.00003\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5rM1P8ws6n7Y"
      },
      "source": [
        "output_columns = ['Precursor [ug/m3]', 'Gas [ug/m3]', 'Aerosol [ug_m3]']\n",
        "pred_rnn_std_simple.columns = output_columns\n",
        "\n",
        "pred_train_rnn_std_reshape_simple = mod_rnn_std_simple.predict(rnn_std_train_in).reshape(scaled_std_train_out.shape)\n",
        "pred_train_rnn_std_simple = pd.DataFrame(output_std_scaler.inverse_transform(pred_train_rnn_std_reshape_simple))\n",
        "\n",
        "pred_train_rnn_std_simple.columns = output_columns"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mya68FOv7IsT"
      },
      "source": [
        "LSTM"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CSznp-8t7MJH",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 459
        },
        "outputId": "f1ab86e4-3bb0-4d20-b352-0bef0a75d031"
      },
      "source": [
        "from tensorflow.keras.layers import LSTM, SimpleRNN\n",
        "from tensorflow.keras.models import Sequential\n",
        "\n",
        "tf.random.set_seed(seed)\n",
        "\n",
        "# Transform the model for use in RNN\n",
        "# previously applied transformer: output_std_scaler\n",
        "rnn_std_train_in = scaled_std_train_in.reshape(-1,1,9)\n",
        "rnn_std_train_out = scaled_std_train_out.reshape(-1,1,3)\n",
        "rnn_std_val_in = scaled_std_val_in.reshape(-1,1,9)\n",
        "rnn_std_val_out = scaled_std_val_out.reshape(-1,1,3)\n",
        "\n",
        "# Set up the model\n",
        "mod_rnn_std_lstm = Sequential()\n",
        "mod_rnn_std_lstm.add(SimpleRNN(256, input_shape=(1, 9), return_sequences=True))\n",
        "mod_rnn_std_lstm.add(Dense(3))\n",
        "mod_rnn_std_lstm.compile(Adam(learning_rate=0.001), \"mse\")\n",
        "\n",
        "# fit the model\n",
        "history = mod_rnn_std_lstm.fit(rnn_std_train_in, rnn_std_train_out, \n",
        "                      validation_data=(rnn_std_val_in, rnn_std_val_out), \n",
        "                      batch_size=256, epochs=5, verbose=1, workers=-1)\n",
        "mod_rnn_std_lstm.summary()\n",
        "\n",
        "# get predictions and truth\n",
        "pred_rnn_std_reshape_lstm = mod_rnn_std_lstm.predict(rnn_std_val_in).reshape(scaled_std_val_out.shape)\n",
        "pred_rnn_std_lstm = pd.DataFrame(output_std_scaler.inverse_transform(pred_rnn_std_reshape_lstm))\n",
        "true_val = val_out.iloc[:,1:-1]\n",
        "\n",
        "print('Metrics for LSTM model:')\n",
        "evaluate_mod(true_val, pred_rnn_std_lstm)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/5\n",
            "7870/7870 [==============================] - 38s 5ms/step - loss: 0.0013 - val_loss: 2.6489e-05\n",
            "Epoch 2/5\n",
            "7870/7870 [==============================] - 39s 5ms/step - loss: 2.1144e-05 - val_loss: 9.6868e-06\n",
            "Epoch 3/5\n",
            "7870/7870 [==============================] - 38s 5ms/step - loss: 1.5821e-05 - val_loss: 1.4885e-05\n",
            "Epoch 4/5\n",
            "7870/7870 [==============================] - 38s 5ms/step - loss: 1.4162e-05 - val_loss: 1.2684e-05\n",
            "Epoch 5/5\n",
            "7870/7870 [==============================] - 37s 5ms/step - loss: 1.3170e-05 - val_loss: 1.2955e-05\n",
            "Model: \"sequential_4\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "simple_rnn_4 (SimpleRNN)     (None, 1, 256)            68096     \n",
            "_________________________________________________________________\n",
            "dense_7 (Dense)              (None, 1, 3)              771       \n",
            "=================================================================\n",
            "Total params: 68,867\n",
            "Trainable params: 68,867\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "Metrics for LSTM model:\n",
            "RMSE: Precursor: 0.00004, Gas: 0.00009, Aerosols: 0.00008\n",
            "R2: Precursor: 0.99999, Gas: 0.99999, Aerosols: 0.99999\n",
            "Hellenger Distance: Precursor: 0.00000, Gas: 0.00000, Aerosols: 0.00003\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DhEH6WM57MJP"
      },
      "source": [
        "output_columns = ['Precursor [ug/m3]', 'Gas [ug/m3]', 'Aerosol [ug_m3]']\n",
        "pred_rnn_std_lstm.columns = output_columns\n",
        "\n",
        "pred_train_rnn_std_reshape_lstm = mod_rnn_std_lstm.predict(rnn_std_train_in).reshape(scaled_std_train_out.shape)\n",
        "pred_train_rnn_std_lstm = pd.DataFrame(output_std_scaler.inverse_transform(pred_train_rnn_std_reshape_lstm))\n",
        "\n",
        "pred_train_rnn_std_lstm.columns = output_columns"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yOpq4pHdZO1p"
      },
      "source": [
        "### Taylor Diagram"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iPAzm2UEZRAa",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 517
        },
        "outputId": "aba22307-bf3d-48b4-9ed6-c0d195f01a3e"
      },
      "source": [
        "train_model_outputs = {\n",
        "    'Linear Train': linear_mod_train,\n",
        "    'DNN Train': pred_train_nn,\n",
        "    'RNN Train': pred_train_rnn_std_2,\n",
        "    'Simple RNN Train': pred_train_rnn_std_simple,\n",
        "    'LSTM Train': pred_train_rnn_std_lstm,\n",
        "    'Random Forest Train': pred_train_rf,\n",
        "    'Linear Val':linear_mod_val,\n",
        "    'DNN Val': pred_nn,\n",
        "    'RNN Val': pred_rnn_std_2, \n",
        "    'Simple RNN Val': pred_rnn_std_simple,\n",
        "    'LSTM Val': pred_rnn_std_lstm,\n",
        "    'Random Forest Val': pred_rf\n",
        "}\n",
        "truth = {\n",
        "    'Train': true_val,\n",
        "    'Val': true_val\n",
        "}\n",
        "\n",
        "def get_stats(truth, models):\n",
        "    stats = pd.DataFrame(index=['MAE', 'RMSE', 'R', 'Hellinger Distance', 'STDVAR', 'Truth STDVAR', 'Col', 'Model Type'])\n",
        "    for col in truth.columns.values:\n",
        "        for model_name, model in models.items():\n",
        "            mae = (truth[col] - model[col]).abs().mean()\n",
        "            rmse = np.sqrt(((truth[col] - model[col])**2).mean())\n",
        "            pearson_corr = ((truth[col]*model[col]).mean() - truth[col].mean()*model[col].mean())/(truth[col].std()*model[col].std())\n",
        "            hd = hellinger_distance(truth[col].values, model[col].values)\n",
        "            model_std = model[col].std()\n",
        "            truth_std = truth[col].std()\n",
        "            stats[f'{model_name} {col}'] = np.array([mae, rmse, pearson_corr, hd, model_std, truth_std, col, model_name])\n",
        "    return stats.transpose().reset_index().drop(columns='index') #.reset_index().set_index(['Col', 'Model Type'])\n",
        "\n",
        "df_all_val = get_stats(truth['Val'], {f'{t} Val': train_model_outputs[f'{t} Val'] for t in ['Linear', 'DNN', 'RNN', 'Simple RNN', 'LSTM', 'Random Forest']})\n",
        "scores_val = {outname: df_out.set_index('Model Type').drop(columns='Col') for outname, df_out in df_all_val.groupby('Col')}\n",
        "df_all_train = get_stats(truth['Val'], {f'{t} Train': train_model_outputs[f'{t} Train'] for t in ['Linear', 'DNN', 'RNN', 'Simple RNN', 'LSTM', 'Random Forest']})\n",
        "scores_train = {outname: df_out.set_index('Model Type').drop(columns='Col') for outname, df_out in df_all_train.groupby('Col')}\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "fig, ax = plt.subplots(figsize=(8,8), subplot_kw=dict(projection='polar'))\n",
        "corrs = [0.1, 0.25, 0.5, 0.7, 0.9,0.95, 0.99]\n",
        "theta_corrs = np.arccos(corrs)*180./np.pi\n",
        "ax.set_thetamax(90)\n",
        "ax.set_thetagrids(theta_corrs, corrs)\n",
        "var_space, theta_space = np.meshgrid(np.linspace(0,2,30), np.linspace(0,np.pi/2, 30))\n",
        "bias_contours = np.sqrt(var_space**2+1-2*var_space*np.cos(theta_space))\n",
        "contours = ax.contour(theta_space, var_space, bias_contours, colors='k')\n",
        "ax.clabel(contours, contours.levels, inline=True, fmt='Bias: %2.1f', fontsize=10)\n",
        "colors = ['r', 'b', 'g', 'c', 'm', 'k']\n",
        "for modeltype, color in zip(['Linear', 'DNN', 'RNN', 'Simple RNN', 'LSTM', 'Random Forest'], colors):\n",
        "    r = []\n",
        "    stdvar_norms = []\n",
        "    labels = []\n",
        "    for datatype, dataset in zip(['Train', 'Val'], [scores_train, scores_val]):\n",
        "        for component in output_columns:\n",
        "            element_r = float(dataset[component].loc[f'{modeltype} {datatype}', 'R'])\n",
        "            r.append(element_r)\n",
        "            std = float(dataset[component].loc[f'{modeltype} {datatype}', 'STDVAR'])\n",
        "            truth_std = float(dataset[component].loc[f'{modeltype} {datatype}', 'Truth STDVAR'])\n",
        "            stdvar_norms.append(std/truth_std)\n",
        "#             ax.text(std/truth_std, np.arccos(element_r),f'{datatype}')\n",
        "    ax.scatter( stdvar_norms,np.arccos(r), color=color, label=modeltype, s = 25)\n",
        "ax.legend(loc=2, bbox_to_anchor=(1.05,1))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f4db83aa2e8>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 149
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "42D0GozvXPN_"
      },
      "source": [
        "### Validation Curves"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "c2XXlQa-DEzt"
      },
      "source": [
        "Validation curve here was plotted with a different model: this RNN model has its input formatted as 3-d array to show time series."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xQfcHxRYDU1b"
      },
      "source": [
        "# Data reprocessing\n",
        "def time_series(df, colnum):\n",
        "    '''\n",
        "    turn dataframe into multi-index dataframe, then to xarray dataset,\n",
        "    then to xarray dataarray. After that, transpose to the required \n",
        "    input shape ([batch, timesteps, feature]).\n",
        "\n",
        "    returns the 3-d np array above in 2-d (-1,colnum) format so it can be\n",
        "    standardized or scaled.\n",
        "    '''\n",
        "    data_array = df.set_index(['id','Time [s]']).to_xarray().to_array()\n",
        "    array = data_array.transpose('id','Time [s]', 'variable').values\n",
        "\n",
        "    return array.reshape(-1, colnum) # reshaped to scaled with xx_std_scaler\n",
        "\n",
        "# np.shape(time_series(train_in, 9))\n",
        "# # (2014600, 9)\n",
        "\n",
        "# actual shape is (1400, 1439, 9)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WLwUPiOHDU1j"
      },
      "source": [
        "# Data preprocessing\n",
        "# Standardizing\n",
        "input_std_scaler = StandardScaler()\n",
        "output_std_scaler = StandardScaler()\n",
        "\n",
        "scaled_std_train_in_ts = input_std_scaler.fit_transform(time_series(train_in, 9))\n",
        "scaled_std_train_out_ts = output_std_scaler.fit_transform(time_series(train_out, 3))\n",
        "scaled_std_val_in_ts = input_std_scaler.transform(time_series(val_in, 9))\n",
        "scaled_std_val_out_ts = output_std_scaler.transform(time_series(val_out, 3))\n",
        "\n",
        "# reshape. note this time data is in [batch, timesteps, feature] size.\n",
        "rnn_std_train_in_ts = scaled_std_train_in_ts.reshape(-1,1439,9)\n",
        "rnn_std_train_out_ts = scaled_std_train_out_ts.reshape(-1,1439,3)\n",
        "rnn_std_val_in_ts = scaled_std_val_in_ts.reshape(-1,1439,9)\n",
        "rnn_std_val_out_ts = scaled_std_val_out_ts.reshape(-1,1439,3)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Kp2uiAYKDU1m",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "d1b2f86b-a6b6-4222-a231-c3e877e84a22"
      },
      "source": [
        "from tensorflow.keras.layers import LSTM, SimpleRNN\n",
        "from tensorflow.keras.models import Sequential\n",
        "\n",
        "tf.random.set_seed(seed)\n",
        "\n",
        "# Set up the model\n",
        "mod_rnn_std_8 = Sequential()\n",
        "mod_rnn_std_8.add(SimpleRNN(256, input_shape=(1439, 9), return_sequences=True))\n",
        "# mod_rnn_std_8.add(LSTM(256, return_sequences=True))\n",
        "mod_rnn_std_8.add(Dense(3))\n",
        "mod_rnn_std_8.compile(Adam(learning_rate=0.001), \"mse\")\n",
        "\n",
        "# fit the model\n",
        "# history = mod_rnn_std_8.fit(rnn_std_train_in_ts, rnn_std_train_out_ts, \n",
        "#                       validation_data=(rnn_std_val_in_ts, rnn_std_val_out_ts), \n",
        "#                       batch_size=10, epochs=5, verbose=1, workers=-1)\n",
        "\n",
        "history = mod_rnn_std_8.fit(rnn_std_train_in_ts, rnn_std_train_out_ts, \n",
        "                      validation_data=(rnn_std_val_in_ts, rnn_std_val_out_ts), \n",
        "                      batch_size=256, epochs=1000, verbose=1, workers=-1)\n",
        "\n",
        "mod_rnn_std_8.summary()\n",
        "\n",
        "# get predictions and truth\n",
        "pred_rnn_std_8 = mod_rnn_std_8.predict(rnn_std_val_in_ts).reshape(-1,3)\n",
        "pred_rnn_std_reshape_8 = mod_rnn_std_8.predict(rnn_std_val_in_ts).reshape(scaled_std_val_out.shape)\n",
        "pred_rnn_std_reverse_8 = pd.DataFrame(output_std_scaler.inverse_transform(pred_rnn_std_reshape_8)).iloc[1:,:]\n",
        "true_val = val_out.iloc[:-1,1:4]\n",
        "\n",
        "print('Metrics for LSTM model:')\n",
        "evaluate_mod(true_val, pred_rnn_std_reverse_8)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.4068 - val_loss: 0.0827\n",
            "Epoch 2/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 0.0606 - val_loss: 0.0394\n",
            "Epoch 3/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 0.0305 - val_loss: 0.0201\n",
            "Epoch 4/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 0.0194 - val_loss: 0.0149\n",
            "Epoch 5/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0134 - val_loss: 0.0119\n",
            "Epoch 6/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0110 - val_loss: 0.0095\n",
            "Epoch 7/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0089 - val_loss: 0.0082\n",
            "Epoch 8/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 0.0077 - val_loss: 0.0073\n",
            "Epoch 9/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 0.0069 - val_loss: 0.0064\n",
            "Epoch 10/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0062 - val_loss: 0.0059\n",
            "Epoch 11/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0057 - val_loss: 0.0054\n",
            "Epoch 12/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0052 - val_loss: 0.0051\n",
            "Epoch 13/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0048 - val_loss: 0.0047\n",
            "Epoch 14/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0045 - val_loss: 0.0043\n",
            "Epoch 15/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0042 - val_loss: 0.0040\n",
            "Epoch 16/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 0.0039 - val_loss: 0.0038\n",
            "Epoch 17/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0036 - val_loss: 0.0035\n",
            "Epoch 18/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 0.0034 - val_loss: 0.0032\n",
            "Epoch 19/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 0.0031 - val_loss: 0.0030\n",
            "Epoch 20/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0029 - val_loss: 0.0028\n",
            "Epoch 21/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0027 - val_loss: 0.0026\n",
            "Epoch 22/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0025 - val_loss: 0.0024\n",
            "Epoch 23/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0024 - val_loss: 0.0023\n",
            "Epoch 24/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0022 - val_loss: 0.0021\n",
            "Epoch 25/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0020 - val_loss: 0.0020\n",
            "Epoch 26/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0019 - val_loss: 0.0018\n",
            "Epoch 27/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 0.0018 - val_loss: 0.0017\n",
            "Epoch 28/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0016 - val_loss: 0.0016\n",
            "Epoch 29/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0015 - val_loss: 0.0015\n",
            "Epoch 30/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0014 - val_loss: 0.0014\n",
            "Epoch 31/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0013 - val_loss: 0.0013\n",
            "Epoch 32/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0012 - val_loss: 0.0012\n",
            "Epoch 33/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0011 - val_loss: 0.0011\n",
            "Epoch 34/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0011 - val_loss: 0.0010\n",
            "Epoch 35/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 9.9060e-04 - val_loss: 9.6463e-04\n",
            "Epoch 36/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 9.2378e-04 - val_loss: 9.0002e-04\n",
            "Epoch 37/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 8.5886e-04 - val_loss: 8.3833e-04\n",
            "Epoch 38/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 8.0094e-04 - val_loss: 7.8364e-04\n",
            "Epoch 39/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 7.4417e-04 - val_loss: 7.2668e-04\n",
            "Epoch 40/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.9785e-04 - val_loss: 6.8525e-04\n",
            "Epoch 41/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 6.5360e-04 - val_loss: 6.3938e-04\n",
            "Epoch 42/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 6.1031e-04 - val_loss: 6.0981e-04\n",
            "Epoch 43/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.7767e-04 - val_loss: 5.6060e-04\n",
            "Epoch 44/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.4152e-04 - val_loss: 5.2719e-04\n",
            "Epoch 45/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 5.0608e-04 - val_loss: 4.9952e-04\n",
            "Epoch 46/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 4.7414e-04 - val_loss: 4.6420e-04\n",
            "Epoch 47/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 4.4547e-04 - val_loss: 4.4366e-04\n",
            "Epoch 48/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 4.1998e-04 - val_loss: 4.1892e-04\n",
            "Epoch 49/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.9697e-04 - val_loss: 4.0136e-04\n",
            "Epoch 50/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.7425e-04 - val_loss: 3.7326e-04\n",
            "Epoch 51/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.5580e-04 - val_loss: 3.5344e-04\n",
            "Epoch 52/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.3837e-04 - val_loss: 3.3698e-04\n",
            "Epoch 53/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.2140e-04 - val_loss: 3.1972e-04\n",
            "Epoch 54/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.0704e-04 - val_loss: 3.1513e-04\n",
            "Epoch 55/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.9379e-04 - val_loss: 2.9835e-04\n",
            "Epoch 56/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.8156e-04 - val_loss: 2.7995e-04\n",
            "Epoch 57/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.6542e-04 - val_loss: 2.6688e-04\n",
            "Epoch 58/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.5393e-04 - val_loss: 2.5661e-04\n",
            "Epoch 59/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.4616e-04 - val_loss: 2.4559e-04\n",
            "Epoch 60/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.3608e-04 - val_loss: 2.4032e-04\n",
            "Epoch 61/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.2813e-04 - val_loss: 2.2785e-04\n",
            "Epoch 62/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.1937e-04 - val_loss: 2.2250e-04\n",
            "Epoch 63/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.1312e-04 - val_loss: 2.2541e-04\n",
            "Epoch 64/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.0926e-04 - val_loss: 2.1210e-04\n",
            "Epoch 65/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0187e-04 - val_loss: 2.0739e-04\n",
            "Epoch 66/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9802e-04 - val_loss: 2.0261e-04\n",
            "Epoch 67/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9012e-04 - val_loss: 1.9320e-04\n",
            "Epoch 68/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8343e-04 - val_loss: 1.8818e-04\n",
            "Epoch 69/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7741e-04 - val_loss: 1.8554e-04\n",
            "Epoch 70/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7454e-04 - val_loss: 1.8007e-04\n",
            "Epoch 71/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7123e-04 - val_loss: 1.7247e-04\n",
            "Epoch 72/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6665e-04 - val_loss: 1.6977e-04\n",
            "Epoch 73/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6280e-04 - val_loss: 1.6698e-04\n",
            "Epoch 74/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5943e-04 - val_loss: 1.6385e-04\n",
            "Epoch 75/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.5657e-04 - val_loss: 1.5964e-04\n",
            "Epoch 76/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5149e-04 - val_loss: 1.6036e-04\n",
            "Epoch 77/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4979e-04 - val_loss: 1.5430e-04\n",
            "Epoch 78/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4651e-04 - val_loss: 1.5028e-04\n",
            "Epoch 79/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.4388e-04 - val_loss: 1.5168e-04\n",
            "Epoch 80/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4161e-04 - val_loss: 1.4867e-04\n",
            "Epoch 81/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4071e-04 - val_loss: 1.4408e-04\n",
            "Epoch 82/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3567e-04 - val_loss: 1.4100e-04\n",
            "Epoch 83/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3337e-04 - val_loss: 1.3926e-04\n",
            "Epoch 84/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3226e-04 - val_loss: 1.3846e-04\n",
            "Epoch 85/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.2893e-04 - val_loss: 1.3573e-04\n",
            "Epoch 86/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.2763e-04 - val_loss: 1.3716e-04\n",
            "Epoch 87/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.2581e-04 - val_loss: 1.3392e-04\n",
            "Epoch 88/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.2341e-04 - val_loss: 1.3074e-04\n",
            "Epoch 89/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.2159e-04 - val_loss: 1.2851e-04\n",
            "Epoch 90/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.1923e-04 - val_loss: 1.3078e-04\n",
            "Epoch 91/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.1926e-04 - val_loss: 1.2716e-04\n",
            "Epoch 92/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.1777e-04 - val_loss: 1.2550e-04\n",
            "Epoch 93/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.1608e-04 - val_loss: 1.2617e-04\n",
            "Epoch 94/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.1341e-04 - val_loss: 1.2512e-04\n",
            "Epoch 95/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.1311e-04 - val_loss: 1.2319e-04\n",
            "Epoch 96/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.1171e-04 - val_loss: 1.2858e-04\n",
            "Epoch 97/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.1096e-04 - val_loss: 1.2412e-04\n",
            "Epoch 98/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.1080e-04 - val_loss: 1.2188e-04\n",
            "Epoch 99/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.0959e-04 - val_loss: 1.2309e-04\n",
            "Epoch 100/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.0736e-04 - val_loss: 1.2474e-04\n",
            "Epoch 101/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.0620e-04 - val_loss: 1.2306e-04\n",
            "Epoch 102/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.0388e-04 - val_loss: 1.2626e-04\n",
            "Epoch 103/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.0199e-04 - val_loss: 1.2161e-04\n",
            "Epoch 104/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.0008e-04 - val_loss: 1.1996e-04\n",
            "Epoch 105/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 9.8118e-05 - val_loss: 1.2223e-04\n",
            "Epoch 106/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 9.8214e-05 - val_loss: 1.2573e-04\n",
            "Epoch 107/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 9.8909e-05 - val_loss: 1.2409e-04\n",
            "Epoch 108/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 9.6592e-05 - val_loss: 1.2313e-04\n",
            "Epoch 109/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 9.5119e-05 - val_loss: 1.2419e-04\n",
            "Epoch 110/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 9.4184e-05 - val_loss: 1.2675e-04\n",
            "Epoch 111/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 9.3962e-05 - val_loss: 1.2870e-04\n",
            "Epoch 112/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 9.2144e-05 - val_loss: 1.3190e-04\n",
            "Epoch 113/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 9.1286e-05 - val_loss: 1.3288e-04\n",
            "Epoch 114/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 9.0335e-05 - val_loss: 1.3275e-04\n",
            "Epoch 115/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 9.0522e-05 - val_loss: 1.3965e-04\n",
            "Epoch 116/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 9.2250e-05 - val_loss: 1.3903e-04\n",
            "Epoch 117/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 8.9735e-05 - val_loss: 1.4032e-04\n",
            "Epoch 118/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 8.6880e-05 - val_loss: 1.4422e-04\n",
            "Epoch 119/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 8.6321e-05 - val_loss: 1.4298e-04\n",
            "Epoch 120/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 8.4431e-05 - val_loss: 1.4364e-04\n",
            "Epoch 121/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 8.3424e-05 - val_loss: 1.4012e-04\n",
            "Epoch 122/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 8.2226e-05 - val_loss: 1.4145e-04\n",
            "Epoch 123/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 8.1506e-05 - val_loss: 1.3790e-04\n",
            "Epoch 124/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 8.0000e-05 - val_loss: 1.3671e-04\n",
            "Epoch 125/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 7.9151e-05 - val_loss: 1.3827e-04\n",
            "Epoch 126/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 7.8993e-05 - val_loss: 1.4136e-04\n",
            "Epoch 127/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 7.9680e-05 - val_loss: 1.4339e-04\n",
            "Epoch 128/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 8.0031e-05 - val_loss: 1.4025e-04\n",
            "Epoch 129/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 7.6922e-05 - val_loss: 1.4412e-04\n",
            "Epoch 130/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 7.7452e-05 - val_loss: 1.4313e-04\n",
            "Epoch 131/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 7.5986e-05 - val_loss: 1.3490e-04\n",
            "Epoch 132/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 7.4137e-05 - val_loss: 1.3581e-04\n",
            "Epoch 133/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 7.3146e-05 - val_loss: 1.3808e-04\n",
            "Epoch 134/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 7.2987e-05 - val_loss: 1.3921e-04\n",
            "Epoch 135/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 7.2143e-05 - val_loss: 1.3682e-04\n",
            "Epoch 136/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 7.0991e-05 - val_loss: 1.3680e-04\n",
            "Epoch 137/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 7.1740e-05 - val_loss: 1.3493e-04\n",
            "Epoch 138/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 7.0834e-05 - val_loss: 1.3200e-04\n",
            "Epoch 139/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 7.0557e-05 - val_loss: 1.3234e-04\n",
            "Epoch 140/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 7.0341e-05 - val_loss: 1.3168e-04\n",
            "Epoch 141/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.9365e-05 - val_loss: 1.2729e-04\n",
            "Epoch 142/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.8056e-05 - val_loss: 1.2922e-04\n",
            "Epoch 143/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.8832e-05 - val_loss: 1.3099e-04\n",
            "Epoch 144/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.7555e-05 - val_loss: 1.3311e-04\n",
            "Epoch 145/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.7946e-05 - val_loss: 1.3034e-04\n",
            "Epoch 146/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.7264e-05 - val_loss: 1.3502e-04\n",
            "Epoch 147/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.8988e-05 - val_loss: 1.3188e-04\n",
            "Epoch 148/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.6535e-05 - val_loss: 1.3127e-04\n",
            "Epoch 149/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.5029e-05 - val_loss: 1.3079e-04\n",
            "Epoch 150/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 6.4134e-05 - val_loss: 1.2886e-04\n",
            "Epoch 151/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.2791e-05 - val_loss: 1.2781e-04\n",
            "Epoch 152/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.2759e-05 - val_loss: 1.2359e-04\n",
            "Epoch 153/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.2593e-05 - val_loss: 1.2293e-04\n",
            "Epoch 154/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 6.0917e-05 - val_loss: 1.1566e-04\n",
            "Epoch 155/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 6.0048e-05 - val_loss: 1.1371e-04\n",
            "Epoch 156/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 6.0742e-05 - val_loss: 1.1376e-04\n",
            "Epoch 157/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 6.0179e-05 - val_loss: 1.1616e-04\n",
            "Epoch 158/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 5.9187e-05 - val_loss: 1.1177e-04\n",
            "Epoch 159/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 5.8335e-05 - val_loss: 1.1124e-04\n",
            "Epoch 160/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 5.8127e-05 - val_loss: 1.1038e-04\n",
            "Epoch 161/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.7579e-05 - val_loss: 1.0837e-04\n",
            "Epoch 162/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.7216e-05 - val_loss: 1.0158e-04\n",
            "Epoch 163/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.6545e-05 - val_loss: 1.0084e-04\n",
            "Epoch 164/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.6689e-05 - val_loss: 1.0524e-04\n",
            "Epoch 165/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.6312e-05 - val_loss: 1.0662e-04\n",
            "Epoch 166/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 5.7089e-05 - val_loss: 9.9912e-05\n",
            "Epoch 167/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.5853e-05 - val_loss: 1.0187e-04\n",
            "Epoch 168/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 5.6435e-05 - val_loss: 1.0199e-04\n",
            "Epoch 169/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.7063e-05 - val_loss: 1.0016e-04\n",
            "Epoch 170/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.4898e-05 - val_loss: 9.8314e-05\n",
            "Epoch 171/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 5.5455e-05 - val_loss: 9.5757e-05\n",
            "Epoch 172/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.4100e-05 - val_loss: 9.1137e-05\n",
            "Epoch 173/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.3055e-05 - val_loss: 8.8541e-05\n",
            "Epoch 174/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.2636e-05 - val_loss: 8.6045e-05\n",
            "Epoch 175/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 5.4721e-05 - val_loss: 9.0731e-05\n",
            "Epoch 176/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 5.3904e-05 - val_loss: 8.9909e-05\n",
            "Epoch 177/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.1829e-05 - val_loss: 8.5093e-05\n",
            "Epoch 178/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.1112e-05 - val_loss: 8.5924e-05\n",
            "Epoch 179/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.0248e-05 - val_loss: 7.9842e-05\n",
            "Epoch 180/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.9346e-05 - val_loss: 7.8643e-05\n",
            "Epoch 181/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.9777e-05 - val_loss: 7.7771e-05\n",
            "Epoch 182/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.9026e-05 - val_loss: 7.4882e-05\n",
            "Epoch 183/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.9037e-05 - val_loss: 7.3024e-05\n",
            "Epoch 184/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.9350e-05 - val_loss: 7.5169e-05\n",
            "Epoch 185/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.0294e-05 - val_loss: 7.2289e-05\n",
            "Epoch 186/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.9123e-05 - val_loss: 6.9022e-05\n",
            "Epoch 187/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.7226e-05 - val_loss: 6.8537e-05\n",
            "Epoch 188/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 4.7216e-05 - val_loss: 6.4351e-05\n",
            "Epoch 189/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.7078e-05 - val_loss: 6.3491e-05\n",
            "Epoch 190/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.6173e-05 - val_loss: 6.1922e-05\n",
            "Epoch 191/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.6233e-05 - val_loss: 6.4300e-05\n",
            "Epoch 192/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.6693e-05 - val_loss: 6.2346e-05\n",
            "Epoch 193/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.6578e-05 - val_loss: 6.0894e-05\n",
            "Epoch 194/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.5918e-05 - val_loss: 6.2757e-05\n",
            "Epoch 195/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.6552e-05 - val_loss: 5.9894e-05\n",
            "Epoch 196/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.6137e-05 - val_loss: 5.7084e-05\n",
            "Epoch 197/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.4618e-05 - val_loss: 5.6442e-05\n",
            "Epoch 198/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.3932e-05 - val_loss: 5.4252e-05\n",
            "Epoch 199/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.3683e-05 - val_loss: 5.5605e-05\n",
            "Epoch 200/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.5275e-05 - val_loss: 5.2784e-05\n",
            "Epoch 201/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.4268e-05 - val_loss: 5.0955e-05\n",
            "Epoch 202/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 4.3497e-05 - val_loss: 4.9136e-05\n",
            "Epoch 203/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.2391e-05 - val_loss: 5.1852e-05\n",
            "Epoch 204/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.3066e-05 - val_loss: 5.1345e-05\n",
            "Epoch 205/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.2922e-05 - val_loss: 5.0503e-05\n",
            "Epoch 206/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.3754e-05 - val_loss: 5.0554e-05\n",
            "Epoch 207/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.2506e-05 - val_loss: 4.7194e-05\n",
            "Epoch 208/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.1802e-05 - val_loss: 4.7980e-05\n",
            "Epoch 209/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.1752e-05 - val_loss: 4.6453e-05\n",
            "Epoch 210/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.0785e-05 - val_loss: 4.5080e-05\n",
            "Epoch 211/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.0166e-05 - val_loss: 4.2880e-05\n",
            "Epoch 212/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.9068e-05 - val_loss: 4.5405e-05\n",
            "Epoch 213/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.1059e-05 - val_loss: 4.7151e-05\n",
            "Epoch 214/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.1583e-05 - val_loss: 4.6145e-05\n",
            "Epoch 215/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.0398e-05 - val_loss: 4.5021e-05\n",
            "Epoch 216/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.3023e-05 - val_loss: 4.8434e-05\n",
            "Epoch 217/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.2321e-05 - val_loss: 4.3754e-05\n",
            "Epoch 218/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.0162e-05 - val_loss: 4.3959e-05\n",
            "Epoch 219/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.9670e-05 - val_loss: 4.1842e-05\n",
            "Epoch 220/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.8964e-05 - val_loss: 3.9630e-05\n",
            "Epoch 221/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.8103e-05 - val_loss: 4.8535e-05\n",
            "Epoch 222/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.0963e-05 - val_loss: 4.1536e-05\n",
            "Epoch 223/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 4.0504e-05 - val_loss: 4.6132e-05\n",
            "Epoch 224/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.2300e-05 - val_loss: 4.7492e-05\n",
            "Epoch 225/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.1644e-05 - val_loss: 4.6069e-05\n",
            "Epoch 226/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.5803e-05 - val_loss: 5.5818e-05\n",
            "Epoch 227/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.5764e-05 - val_loss: 4.4818e-05\n",
            "Epoch 228/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.3634e-05 - val_loss: 5.8263e-05\n",
            "Epoch 229/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.5069e-05 - val_loss: 4.1383e-05\n",
            "Epoch 230/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.2451e-05 - val_loss: 4.9162e-05\n",
            "Epoch 231/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.1312e-05 - val_loss: 3.8332e-05\n",
            "Epoch 232/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.8390e-05 - val_loss: 4.3748e-05\n",
            "Epoch 233/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.7137e-05 - val_loss: 3.8390e-05\n",
            "Epoch 234/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.6209e-05 - val_loss: 3.9939e-05\n",
            "Epoch 235/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.6072e-05 - val_loss: 3.8612e-05\n",
            "Epoch 236/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.6038e-05 - val_loss: 4.1416e-05\n",
            "Epoch 237/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.5042e-05 - val_loss: 3.6727e-05\n",
            "Epoch 238/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.4325e-05 - val_loss: 4.1718e-05\n",
            "Epoch 239/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.5992e-05 - val_loss: 4.1775e-05\n",
            "Epoch 240/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.4990e-05 - val_loss: 3.8096e-05\n",
            "Epoch 241/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.5368e-05 - val_loss: 3.7719e-05\n",
            "Epoch 242/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.4613e-05 - val_loss: 3.8407e-05\n",
            "Epoch 243/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.3756e-05 - val_loss: 3.5924e-05\n",
            "Epoch 244/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.3272e-05 - val_loss: 3.5598e-05\n",
            "Epoch 245/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.2897e-05 - val_loss: 3.4629e-05\n",
            "Epoch 246/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.2020e-05 - val_loss: 3.5636e-05\n",
            "Epoch 247/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.2390e-05 - val_loss: 3.4742e-05\n",
            "Epoch 248/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.1929e-05 - val_loss: 3.5497e-05\n",
            "Epoch 249/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.3292e-05 - val_loss: 3.5016e-05\n",
            "Epoch 250/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.2507e-05 - val_loss: 3.4279e-05\n",
            "Epoch 251/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.1633e-05 - val_loss: 3.4447e-05\n",
            "Epoch 252/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.2131e-05 - val_loss: 3.3799e-05\n",
            "Epoch 253/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.1350e-05 - val_loss: 3.3575e-05\n",
            "Epoch 254/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.1421e-05 - val_loss: 3.4024e-05\n",
            "Epoch 255/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.1542e-05 - val_loss: 3.4003e-05\n",
            "Epoch 256/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.1301e-05 - val_loss: 3.3292e-05\n",
            "Epoch 257/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.1121e-05 - val_loss: 3.2488e-05\n",
            "Epoch 258/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.0303e-05 - val_loss: 3.2441e-05\n",
            "Epoch 259/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.0577e-05 - val_loss: 3.2205e-05\n",
            "Epoch 260/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.0225e-05 - val_loss: 3.2392e-05\n",
            "Epoch 261/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.0345e-05 - val_loss: 3.2285e-05\n",
            "Epoch 262/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.9537e-05 - val_loss: 3.4133e-05\n",
            "Epoch 263/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.0110e-05 - val_loss: 3.0818e-05\n",
            "Epoch 264/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.8647e-05 - val_loss: 3.0498e-05\n",
            "Epoch 265/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.8375e-05 - val_loss: 3.1610e-05\n",
            "Epoch 266/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.9474e-05 - val_loss: 3.4415e-05\n",
            "Epoch 267/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.2161e-05 - val_loss: 3.3797e-05\n",
            "Epoch 268/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.2371e-05 - val_loss: 3.3692e-05\n",
            "Epoch 269/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.1059e-05 - val_loss: 3.1219e-05\n",
            "Epoch 270/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.9134e-05 - val_loss: 3.1036e-05\n",
            "Epoch 271/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.0614e-05 - val_loss: 3.1574e-05\n",
            "Epoch 272/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.2383e-05 - val_loss: 3.6840e-05\n",
            "Epoch 273/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.2398e-05 - val_loss: 3.5995e-05\n",
            "Epoch 274/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.0861e-05 - val_loss: 3.8347e-05\n",
            "Epoch 275/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.1906e-05 - val_loss: 3.5188e-05\n",
            "Epoch 276/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.1577e-05 - val_loss: 3.5027e-05\n",
            "Epoch 277/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.2085e-05 - val_loss: 3.5383e-05\n",
            "Epoch 278/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.0695e-05 - val_loss: 3.7762e-05\n",
            "Epoch 279/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.1685e-05 - val_loss: 3.5928e-05\n",
            "Epoch 280/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.0382e-05 - val_loss: 3.1155e-05\n",
            "Epoch 281/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.8518e-05 - val_loss: 2.8758e-05\n",
            "Epoch 282/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.7269e-05 - val_loss: 2.7967e-05\n",
            "Epoch 283/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.5809e-05 - val_loss: 2.8019e-05\n",
            "Epoch 284/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.6110e-05 - val_loss: 2.8798e-05\n",
            "Epoch 285/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.6331e-05 - val_loss: 2.7466e-05\n",
            "Epoch 286/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.5703e-05 - val_loss: 2.7183e-05\n",
            "Epoch 287/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.5420e-05 - val_loss: 2.6844e-05\n",
            "Epoch 288/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.5537e-05 - val_loss: 2.7826e-05\n",
            "Epoch 289/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.5814e-05 - val_loss: 2.7437e-05\n",
            "Epoch 290/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.5122e-05 - val_loss: 2.7235e-05\n",
            "Epoch 291/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.4777e-05 - val_loss: 2.7284e-05\n",
            "Epoch 292/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.4823e-05 - val_loss: 2.7734e-05\n",
            "Epoch 293/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.4900e-05 - val_loss: 2.6888e-05\n",
            "Epoch 294/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.4889e-05 - val_loss: 2.7176e-05\n",
            "Epoch 295/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.5035e-05 - val_loss: 2.8855e-05\n",
            "Epoch 296/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.5881e-05 - val_loss: 2.7397e-05\n",
            "Epoch 297/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.6043e-05 - val_loss: 2.8319e-05\n",
            "Epoch 298/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.6008e-05 - val_loss: 2.7983e-05\n",
            "Epoch 299/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.5222e-05 - val_loss: 2.7600e-05\n",
            "Epoch 300/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.5475e-05 - val_loss: 2.6797e-05\n",
            "Epoch 301/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.6564e-05 - val_loss: 3.7680e-05\n",
            "Epoch 302/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.0490e-05 - val_loss: 2.9646e-05\n",
            "Epoch 303/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.3266e-05 - val_loss: 3.0315e-05\n",
            "Epoch 304/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.5924e-05 - val_loss: 5.4069e-05\n",
            "Epoch 305/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.8648e-05 - val_loss: 3.5978e-05\n",
            "Epoch 306/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.7131e-05 - val_loss: 4.2645e-05\n",
            "Epoch 307/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.4874e-05 - val_loss: 5.4636e-05\n",
            "Epoch 308/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.6997e-05 - val_loss: 7.2138e-05\n",
            "Epoch 309/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 7.0337e-05 - val_loss: 3.6383e-05\n",
            "Epoch 310/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.2637e-05 - val_loss: 4.4382e-05\n",
            "Epoch 311/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.8385e-05 - val_loss: 4.1998e-05\n",
            "Epoch 312/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.2532e-05 - val_loss: 3.2455e-05\n",
            "Epoch 313/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.7778e-05 - val_loss: 3.2107e-05\n",
            "Epoch 314/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.7095e-05 - val_loss: 2.9503e-05\n",
            "Epoch 315/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.6607e-05 - val_loss: 2.7701e-05\n",
            "Epoch 316/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.7340e-05 - val_loss: 2.8241e-05\n",
            "Epoch 317/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.1034e-05 - val_loss: 3.0094e-05\n",
            "Epoch 318/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.9129e-05 - val_loss: 2.4796e-05\n",
            "Epoch 319/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.5810e-05 - val_loss: 2.9809e-05\n",
            "Epoch 320/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.6496e-05 - val_loss: 2.8494e-05\n",
            "Epoch 321/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.5133e-05 - val_loss: 2.6830e-05\n",
            "Epoch 322/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.3654e-05 - val_loss: 2.7871e-05\n",
            "Epoch 323/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.3498e-05 - val_loss: 2.7998e-05\n",
            "Epoch 324/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.2910e-05 - val_loss: 2.4693e-05\n",
            "Epoch 325/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.5745e-05 - val_loss: 2.4605e-05\n",
            "Epoch 326/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.3252e-05 - val_loss: 2.6321e-05\n",
            "Epoch 327/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.3408e-05 - val_loss: 2.5595e-05\n",
            "Epoch 328/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.2921e-05 - val_loss: 2.4086e-05\n",
            "Epoch 329/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.1742e-05 - val_loss: 2.4662e-05\n",
            "Epoch 330/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.4914e-05 - val_loss: 2.4470e-05\n",
            "Epoch 331/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.3522e-05 - val_loss: 2.4658e-05\n",
            "Epoch 332/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.1344e-05 - val_loss: 2.3157e-05\n",
            "Epoch 333/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.1447e-05 - val_loss: 2.2749e-05\n",
            "Epoch 334/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2873e-05 - val_loss: 2.4978e-05\n",
            "Epoch 335/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.2152e-05 - val_loss: 2.6494e-05\n",
            "Epoch 336/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.3933e-05 - val_loss: 2.1966e-05\n",
            "Epoch 337/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.1328e-05 - val_loss: 2.5105e-05\n",
            "Epoch 338/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.2151e-05 - val_loss: 2.3400e-05\n",
            "Epoch 339/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.2930e-05 - val_loss: 2.2139e-05\n",
            "Epoch 340/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.1312e-05 - val_loss: 2.2355e-05\n",
            "Epoch 341/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0241e-05 - val_loss: 2.2076e-05\n",
            "Epoch 342/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.0320e-05 - val_loss: 2.3634e-05\n",
            "Epoch 343/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.1391e-05 - val_loss: 2.3991e-05\n",
            "Epoch 344/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.0180e-05 - val_loss: 2.1469e-05\n",
            "Epoch 345/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9684e-05 - val_loss: 2.2694e-05\n",
            "Epoch 346/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0497e-05 - val_loss: 2.2058e-05\n",
            "Epoch 347/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0013e-05 - val_loss: 2.0238e-05\n",
            "Epoch 348/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.8820e-05 - val_loss: 2.0086e-05\n",
            "Epoch 349/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.8434e-05 - val_loss: 2.0961e-05\n",
            "Epoch 350/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0227e-05 - val_loss: 2.1782e-05\n",
            "Epoch 351/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.0094e-05 - val_loss: 2.2046e-05\n",
            "Epoch 352/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9033e-05 - val_loss: 2.1058e-05\n",
            "Epoch 353/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.8942e-05 - val_loss: 2.2834e-05\n",
            "Epoch 354/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.0909e-05 - val_loss: 2.3599e-05\n",
            "Epoch 355/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.0631e-05 - val_loss: 2.3167e-05\n",
            "Epoch 356/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.1531e-05 - val_loss: 2.4605e-05\n",
            "Epoch 357/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.6256e-05 - val_loss: 3.5458e-05\n",
            "Epoch 358/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.1954e-05 - val_loss: 3.3952e-05\n",
            "Epoch 359/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.5240e-05 - val_loss: 4.6292e-05\n",
            "Epoch 360/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.2847e-05 - val_loss: 2.7992e-05\n",
            "Epoch 361/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 6.2033e-05 - val_loss: 1.1437e-04\n",
            "Epoch 362/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 9.5280e-05 - val_loss: 4.1761e-05\n",
            "Epoch 363/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 8.8607e-05 - val_loss: 8.5432e-05\n",
            "Epoch 364/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.2963e-05 - val_loss: 6.8163e-05\n",
            "Epoch 365/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.5039e-05 - val_loss: 4.9961e-05\n",
            "Epoch 366/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.5276e-05 - val_loss: 4.3815e-05\n",
            "Epoch 367/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.0431e-05 - val_loss: 3.2177e-05\n",
            "Epoch 368/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.7842e-05 - val_loss: 2.8971e-05\n",
            "Epoch 369/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.5520e-05 - val_loss: 2.5480e-05\n",
            "Epoch 370/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.6318e-05 - val_loss: 2.8172e-05\n",
            "Epoch 371/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.4864e-05 - val_loss: 2.6914e-05\n",
            "Epoch 372/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.4362e-05 - val_loss: 3.1505e-05\n",
            "Epoch 373/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.5245e-05 - val_loss: 2.8590e-05\n",
            "Epoch 374/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2978e-05 - val_loss: 3.1375e-05\n",
            "Epoch 375/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2654e-05 - val_loss: 2.2752e-05\n",
            "Epoch 376/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0109e-05 - val_loss: 2.2974e-05\n",
            "Epoch 377/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9349e-05 - val_loss: 2.0806e-05\n",
            "Epoch 378/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7728e-05 - val_loss: 2.1396e-05\n",
            "Epoch 379/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7847e-05 - val_loss: 2.0064e-05\n",
            "Epoch 380/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8791e-05 - val_loss: 1.9133e-05\n",
            "Epoch 381/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7614e-05 - val_loss: 2.2775e-05\n",
            "Epoch 382/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8528e-05 - val_loss: 2.0616e-05\n",
            "Epoch 383/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.7786e-05 - val_loss: 1.9889e-05\n",
            "Epoch 384/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.7307e-05 - val_loss: 2.0678e-05\n",
            "Epoch 385/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7536e-05 - val_loss: 2.9578e-05\n",
            "Epoch 386/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2865e-05 - val_loss: 2.3249e-05\n",
            "Epoch 387/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.1182e-05 - val_loss: 2.2693e-05\n",
            "Epoch 388/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0531e-05 - val_loss: 2.2495e-05\n",
            "Epoch 389/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0117e-05 - val_loss: 2.2958e-05\n",
            "Epoch 390/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.4340e-05 - val_loss: 3.3847e-05\n",
            "Epoch 391/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.8395e-05 - val_loss: 3.2012e-05\n",
            "Epoch 392/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.6696e-05 - val_loss: 2.5156e-05\n",
            "Epoch 393/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.4496e-05 - val_loss: 2.5731e-05\n",
            "Epoch 394/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.5777e-05 - val_loss: 2.7920e-05\n",
            "Epoch 395/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.7124e-05 - val_loss: 3.0658e-05\n",
            "Epoch 396/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.3844e-05 - val_loss: 3.0828e-05\n",
            "Epoch 397/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.2273e-05 - val_loss: 3.0770e-05\n",
            "Epoch 398/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2150e-05 - val_loss: 2.2716e-05\n",
            "Epoch 399/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.8981e-05 - val_loss: 2.4347e-05\n",
            "Epoch 400/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0157e-05 - val_loss: 2.2977e-05\n",
            "Epoch 401/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8836e-05 - val_loss: 2.3333e-05\n",
            "Epoch 402/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8981e-05 - val_loss: 1.9608e-05\n",
            "Epoch 403/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6163e-05 - val_loss: 2.3510e-05\n",
            "Epoch 404/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8869e-05 - val_loss: 2.2042e-05\n",
            "Epoch 405/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7804e-05 - val_loss: 2.3198e-05\n",
            "Epoch 406/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9145e-05 - val_loss: 2.1889e-05\n",
            "Epoch 407/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.7421e-05 - val_loss: 2.0653e-05\n",
            "Epoch 408/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.8216e-05 - val_loss: 2.0549e-05\n",
            "Epoch 409/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7465e-05 - val_loss: 2.1077e-05\n",
            "Epoch 410/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9441e-05 - val_loss: 2.7293e-05\n",
            "Epoch 411/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0211e-05 - val_loss: 2.2984e-05\n",
            "Epoch 412/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7838e-05 - val_loss: 2.0949e-05\n",
            "Epoch 413/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7243e-05 - val_loss: 1.7695e-05\n",
            "Epoch 414/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7017e-05 - val_loss: 2.0536e-05\n",
            "Epoch 415/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9509e-05 - val_loss: 2.0541e-05\n",
            "Epoch 416/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9006e-05 - val_loss: 2.1693e-05\n",
            "Epoch 417/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0385e-05 - val_loss: 2.0837e-05\n",
            "Epoch 418/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8219e-05 - val_loss: 2.8029e-05\n",
            "Epoch 419/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9104e-05 - val_loss: 2.0557e-05\n",
            "Epoch 420/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6681e-05 - val_loss: 2.0025e-05\n",
            "Epoch 421/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6307e-05 - val_loss: 2.0313e-05\n",
            "Epoch 422/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7547e-05 - val_loss: 2.4962e-05\n",
            "Epoch 423/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.1104e-05 - val_loss: 2.7310e-05\n",
            "Epoch 424/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.1399e-05 - val_loss: 1.9363e-05\n",
            "Epoch 425/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0928e-05 - val_loss: 2.2188e-05\n",
            "Epoch 426/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0047e-05 - val_loss: 2.5769e-05\n",
            "Epoch 427/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9450e-05 - val_loss: 2.3327e-05\n",
            "Epoch 428/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9686e-05 - val_loss: 2.1163e-05\n",
            "Epoch 429/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8942e-05 - val_loss: 2.2296e-05\n",
            "Epoch 430/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.0237e-05 - val_loss: 1.8809e-05\n",
            "Epoch 431/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7972e-05 - val_loss: 2.8029e-05\n",
            "Epoch 432/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9446e-05 - val_loss: 2.0473e-05\n",
            "Epoch 433/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7263e-05 - val_loss: 1.8680e-05\n",
            "Epoch 434/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.8998e-05 - val_loss: 2.0878e-05\n",
            "Epoch 435/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7378e-05 - val_loss: 2.0167e-05\n",
            "Epoch 436/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6231e-05 - val_loss: 2.1113e-05\n",
            "Epoch 437/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6490e-05 - val_loss: 1.8003e-05\n",
            "Epoch 438/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5175e-05 - val_loss: 1.9107e-05\n",
            "Epoch 439/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6910e-05 - val_loss: 2.2922e-05\n",
            "Epoch 440/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.4999e-05 - val_loss: 4.4976e-05\n",
            "Epoch 441/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.8425e-05 - val_loss: 2.7823e-05\n",
            "Epoch 442/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2279e-05 - val_loss: 2.8470e-05\n",
            "Epoch 443/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.1772e-05 - val_loss: 2.0607e-05\n",
            "Epoch 444/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9240e-05 - val_loss: 2.1611e-05\n",
            "Epoch 445/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9017e-05 - val_loss: 2.1441e-05\n",
            "Epoch 446/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7599e-05 - val_loss: 1.9618e-05\n",
            "Epoch 447/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.1300e-05 - val_loss: 3.1150e-05\n",
            "Epoch 448/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.0502e-05 - val_loss: 2.5188e-05\n",
            "Epoch 449/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.7208e-05 - val_loss: 3.0108e-05\n",
            "Epoch 450/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.6341e-05 - val_loss: 2.9164e-05\n",
            "Epoch 451/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.8155e-05 - val_loss: 8.3326e-05\n",
            "Epoch 452/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.2077e-05 - val_loss: 5.1741e-05\n",
            "Epoch 453/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.4721e-05 - val_loss: 2.7812e-05\n",
            "Epoch 454/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.5329e-05 - val_loss: 3.4716e-05\n",
            "Epoch 455/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.9952e-05 - val_loss: 3.9855e-05\n",
            "Epoch 456/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.9934e-05 - val_loss: 4.1916e-05\n",
            "Epoch 457/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 4.2140e-05 - val_loss: 3.6592e-05\n",
            "Epoch 458/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.1092e-05 - val_loss: 3.4745e-05\n",
            "Epoch 459/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.4399e-05 - val_loss: 2.0978e-05\n",
            "Epoch 460/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.9475e-05 - val_loss: 8.8654e-05\n",
            "Epoch 461/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.8460e-05 - val_loss: 3.8182e-05\n",
            "Epoch 462/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 4.7039e-05 - val_loss: 6.2683e-05\n",
            "Epoch 463/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 4.1187e-05 - val_loss: 3.0659e-05\n",
            "Epoch 464/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.5078e-05 - val_loss: 5.2935e-05\n",
            "Epoch 465/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.1148e-05 - val_loss: 2.1103e-05\n",
            "Epoch 466/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.4070e-05 - val_loss: 1.9209e-05\n",
            "Epoch 467/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.9315e-05 - val_loss: 1.7319e-05\n",
            "Epoch 468/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.5551e-05 - val_loss: 1.8799e-05\n",
            "Epoch 469/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6774e-05 - val_loss: 1.4349e-05\n",
            "Epoch 470/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5920e-05 - val_loss: 1.7957e-05\n",
            "Epoch 471/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5032e-05 - val_loss: 1.5668e-05\n",
            "Epoch 472/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4301e-05 - val_loss: 1.6427e-05\n",
            "Epoch 473/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4593e-05 - val_loss: 1.5201e-05\n",
            "Epoch 474/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3314e-05 - val_loss: 1.3636e-05\n",
            "Epoch 475/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.2343e-05 - val_loss: 1.5603e-05\n",
            "Epoch 476/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3210e-05 - val_loss: 1.4766e-05\n",
            "Epoch 477/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3764e-05 - val_loss: 1.4671e-05\n",
            "Epoch 478/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3254e-05 - val_loss: 2.3395e-05\n",
            "Epoch 479/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.4500e-05 - val_loss: 6.2235e-05\n",
            "Epoch 480/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 6.1337e-05 - val_loss: 1.4843e-04\n",
            "Epoch 481/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.0234e-04 - val_loss: 5.0240e-05\n",
            "Epoch 482/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.9774e-05 - val_loss: 3.6418e-05\n",
            "Epoch 483/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.0792e-05 - val_loss: 4.8659e-05\n",
            "Epoch 484/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.1862e-05 - val_loss: 2.9953e-05\n",
            "Epoch 485/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.3666e-05 - val_loss: 1.8555e-05\n",
            "Epoch 486/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7257e-05 - val_loss: 1.5756e-05\n",
            "Epoch 487/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5430e-05 - val_loss: 1.5536e-05\n",
            "Epoch 488/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4270e-05 - val_loss: 1.5337e-05\n",
            "Epoch 489/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6418e-05 - val_loss: 1.9833e-05\n",
            "Epoch 490/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6258e-05 - val_loss: 1.5942e-05\n",
            "Epoch 491/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4640e-05 - val_loss: 1.6015e-05\n",
            "Epoch 492/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7244e-05 - val_loss: 1.5984e-05\n",
            "Epoch 493/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.7262e-05 - val_loss: 1.9182e-05\n",
            "Epoch 494/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6638e-05 - val_loss: 2.3780e-05\n",
            "Epoch 495/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0532e-05 - val_loss: 2.0009e-05\n",
            "Epoch 496/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9159e-05 - val_loss: 2.2837e-05\n",
            "Epoch 497/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9830e-05 - val_loss: 3.7972e-05\n",
            "Epoch 498/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.7645e-05 - val_loss: 2.1802e-05\n",
            "Epoch 499/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.5103e-05 - val_loss: 4.0905e-05\n",
            "Epoch 500/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.1283e-05 - val_loss: 3.7925e-05\n",
            "Epoch 501/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.7064e-05 - val_loss: 2.0751e-05\n",
            "Epoch 502/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0489e-05 - val_loss: 2.3326e-05\n",
            "Epoch 503/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.9438e-05 - val_loss: 1.7471e-05\n",
            "Epoch 504/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.9675e-05 - val_loss: 1.8428e-05\n",
            "Epoch 505/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.1494e-05 - val_loss: 1.5872e-05\n",
            "Epoch 506/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2630e-05 - val_loss: 2.7827e-05\n",
            "Epoch 507/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.3112e-05 - val_loss: 1.6366e-05\n",
            "Epoch 508/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8180e-05 - val_loss: 2.3897e-05\n",
            "Epoch 509/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.8841e-05 - val_loss: 4.4746e-05\n",
            "Epoch 510/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.5440e-05 - val_loss: 3.5011e-05\n",
            "Epoch 511/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.4224e-05 - val_loss: 2.3667e-05\n",
            "Epoch 512/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.3109e-05 - val_loss: 2.2321e-05\n",
            "Epoch 513/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9666e-05 - val_loss: 1.7089e-05\n",
            "Epoch 514/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7485e-05 - val_loss: 1.7359e-05\n",
            "Epoch 515/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8799e-05 - val_loss: 2.4921e-05\n",
            "Epoch 516/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.1417e-05 - val_loss: 1.7181e-05\n",
            "Epoch 517/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9219e-05 - val_loss: 1.5794e-05\n",
            "Epoch 518/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6401e-05 - val_loss: 1.6861e-05\n",
            "Epoch 519/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5363e-05 - val_loss: 1.8127e-05\n",
            "Epoch 520/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.5904e-05 - val_loss: 2.1437e-05\n",
            "Epoch 521/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8582e-05 - val_loss: 2.0430e-05\n",
            "Epoch 522/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7821e-05 - val_loss: 1.6687e-05\n",
            "Epoch 523/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9025e-05 - val_loss: 1.8020e-05\n",
            "Epoch 524/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.7586e-05 - val_loss: 2.2959e-05\n",
            "Epoch 525/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9892e-05 - val_loss: 3.0770e-05\n",
            "Epoch 526/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2902e-05 - val_loss: 2.7800e-05\n",
            "Epoch 527/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.8674e-05 - val_loss: 1.5836e-05\n",
            "Epoch 528/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.9579e-05 - val_loss: 2.0437e-05\n",
            "Epoch 529/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.1566e-05 - val_loss: 2.5596e-05\n",
            "Epoch 530/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.1617e-05 - val_loss: 1.6958e-05\n",
            "Epoch 531/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8821e-05 - val_loss: 1.7254e-05\n",
            "Epoch 532/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.7095e-05 - val_loss: 4.1834e-05\n",
            "Epoch 533/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.6747e-05 - val_loss: 3.0454e-05\n",
            "Epoch 534/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.2981e-05 - val_loss: 3.5457e-05\n",
            "Epoch 535/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.7776e-05 - val_loss: 2.7454e-05\n",
            "Epoch 536/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.4529e-05 - val_loss: 1.4817e-05\n",
            "Epoch 537/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2776e-05 - val_loss: 2.7059e-05\n",
            "Epoch 538/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.1383e-05 - val_loss: 2.5206e-05\n",
            "Epoch 539/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.3784e-05 - val_loss: 2.5590e-05\n",
            "Epoch 540/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9749e-05 - val_loss: 1.4543e-05\n",
            "Epoch 541/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6268e-05 - val_loss: 1.6969e-05\n",
            "Epoch 542/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8084e-05 - val_loss: 1.9929e-05\n",
            "Epoch 543/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8658e-05 - val_loss: 2.8103e-05\n",
            "Epoch 544/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2460e-05 - val_loss: 3.7973e-05\n",
            "Epoch 545/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.6531e-05 - val_loss: 4.3405e-05\n",
            "Epoch 546/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.7829e-05 - val_loss: 5.1435e-05\n",
            "Epoch 547/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.7109e-05 - val_loss: 6.4481e-05\n",
            "Epoch 548/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.1215e-05 - val_loss: 4.4004e-05\n",
            "Epoch 549/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.6007e-05 - val_loss: 4.6274e-05\n",
            "Epoch 550/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.4141e-05 - val_loss: 2.2761e-05\n",
            "Epoch 551/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0306e-05 - val_loss: 1.6180e-05\n",
            "Epoch 552/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6692e-05 - val_loss: 1.2315e-05\n",
            "Epoch 553/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5356e-05 - val_loss: 1.8403e-05\n",
            "Epoch 554/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6419e-05 - val_loss: 2.5316e-05\n",
            "Epoch 555/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8667e-05 - val_loss: 1.4241e-05\n",
            "Epoch 556/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4991e-05 - val_loss: 1.7680e-05\n",
            "Epoch 557/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3636e-05 - val_loss: 2.1165e-05\n",
            "Epoch 558/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7593e-05 - val_loss: 2.2929e-05\n",
            "Epoch 559/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.5607e-05 - val_loss: 2.0082e-05\n",
            "Epoch 560/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8036e-05 - val_loss: 1.7605e-05\n",
            "Epoch 561/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4923e-05 - val_loss: 1.6105e-05\n",
            "Epoch 562/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5647e-05 - val_loss: 1.4430e-05\n",
            "Epoch 563/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7728e-05 - val_loss: 1.4043e-05\n",
            "Epoch 564/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0743e-05 - val_loss: 1.5774e-05\n",
            "Epoch 565/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5312e-05 - val_loss: 1.6843e-05\n",
            "Epoch 566/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4481e-05 - val_loss: 1.4149e-05\n",
            "Epoch 567/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.4395e-05 - val_loss: 1.5422e-05\n",
            "Epoch 568/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5595e-05 - val_loss: 1.6328e-05\n",
            "Epoch 569/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7376e-05 - val_loss: 1.5246e-05\n",
            "Epoch 570/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7310e-05 - val_loss: 6.0397e-05\n",
            "Epoch 571/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.3087e-05 - val_loss: 7.7847e-05\n",
            "Epoch 572/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 6.1552e-05 - val_loss: 1.1622e-04\n",
            "Epoch 573/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.0914e-04 - val_loss: 7.0988e-05\n",
            "Epoch 574/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 9.0952e-05 - val_loss: 9.2048e-05\n",
            "Epoch 575/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 8.0992e-05 - val_loss: 4.8796e-05\n",
            "Epoch 576/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.7985e-05 - val_loss: 7.0335e-05\n",
            "Epoch 577/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.3923e-05 - val_loss: 2.9517e-05\n",
            "Epoch 578/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.8456e-05 - val_loss: 2.9534e-05\n",
            "Epoch 579/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.3117e-05 - val_loss: 2.1215e-05\n",
            "Epoch 580/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0204e-05 - val_loss: 2.4652e-05\n",
            "Epoch 581/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8310e-05 - val_loss: 1.5377e-05\n",
            "Epoch 582/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5399e-05 - val_loss: 2.1781e-05\n",
            "Epoch 583/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7415e-05 - val_loss: 1.5619e-05\n",
            "Epoch 584/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7380e-05 - val_loss: 2.4235e-05\n",
            "Epoch 585/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.5443e-05 - val_loss: 1.8062e-05\n",
            "Epoch 586/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.6161e-05 - val_loss: 2.9918e-05\n",
            "Epoch 587/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.0815e-05 - val_loss: 2.0335e-05\n",
            "Epoch 588/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0947e-05 - val_loss: 2.1430e-05\n",
            "Epoch 589/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9564e-05 - val_loss: 2.6480e-05\n",
            "Epoch 590/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8134e-05 - val_loss: 2.0081e-05\n",
            "Epoch 591/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.6571e-05 - val_loss: 1.5434e-05\n",
            "Epoch 592/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5624e-05 - val_loss: 1.7511e-05\n",
            "Epoch 593/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8611e-05 - val_loss: 1.3842e-05\n",
            "Epoch 594/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3336e-05 - val_loss: 1.5080e-05\n",
            "Epoch 595/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.2816e-05 - val_loss: 1.1937e-05\n",
            "Epoch 596/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.3016e-05 - val_loss: 2.2056e-05\n",
            "Epoch 597/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6947e-05 - val_loss: 1.9706e-05\n",
            "Epoch 598/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0354e-05 - val_loss: 2.8791e-05\n",
            "Epoch 599/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.0576e-05 - val_loss: 1.5253e-05\n",
            "Epoch 600/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.5525e-05 - val_loss: 2.8662e-05\n",
            "Epoch 601/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.6168e-05 - val_loss: 1.4689e-05\n",
            "Epoch 602/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9162e-05 - val_loss: 3.0442e-05\n",
            "Epoch 603/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.8661e-05 - val_loss: 1.5995e-05\n",
            "Epoch 604/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2046e-05 - val_loss: 1.9066e-05\n",
            "Epoch 605/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8186e-05 - val_loss: 1.7604e-05\n",
            "Epoch 606/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.7014e-05 - val_loss: 1.7179e-05\n",
            "Epoch 607/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.7852e-05 - val_loss: 1.9181e-05\n",
            "Epoch 608/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.2347e-05 - val_loss: 2.2190e-05\n",
            "Epoch 609/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.2423e-05 - val_loss: 4.5761e-05\n",
            "Epoch 610/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.0605e-05 - val_loss: 6.4791e-05\n",
            "Epoch 611/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.8399e-05 - val_loss: 6.5178e-05\n",
            "Epoch 612/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.4558e-05 - val_loss: 2.8833e-05\n",
            "Epoch 613/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.8776e-05 - val_loss: 2.7591e-05\n",
            "Epoch 614/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.3400e-05 - val_loss: 1.7701e-05\n",
            "Epoch 615/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.3363e-05 - val_loss: 3.5102e-05\n",
            "Epoch 616/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.7597e-05 - val_loss: 1.0430e-04\n",
            "Epoch 617/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.6110e-05 - val_loss: 3.9487e-05\n",
            "Epoch 618/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.2708e-05 - val_loss: 4.1856e-05\n",
            "Epoch 619/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.3535e-05 - val_loss: 4.8726e-05\n",
            "Epoch 620/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.8056e-05 - val_loss: 3.7278e-05\n",
            "Epoch 621/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.8854e-05 - val_loss: 3.4232e-05\n",
            "Epoch 622/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.7377e-05 - val_loss: 3.9458e-05\n",
            "Epoch 623/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.5821e-05 - val_loss: 2.8458e-05\n",
            "Epoch 624/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.5320e-05 - val_loss: 3.5937e-05\n",
            "Epoch 625/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.4009e-05 - val_loss: 2.0852e-05\n",
            "Epoch 626/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.0793e-05 - val_loss: 2.0868e-05\n",
            "Epoch 627/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.8388e-05 - val_loss: 2.5068e-05\n",
            "Epoch 628/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.1672e-05 - val_loss: 2.3851e-05\n",
            "Epoch 629/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0795e-05 - val_loss: 2.2896e-05\n",
            "Epoch 630/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.9298e-05 - val_loss: 2.1930e-05\n",
            "Epoch 631/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.8696e-05 - val_loss: 1.7566e-05\n",
            "Epoch 632/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.6098e-05 - val_loss: 1.6476e-05\n",
            "Epoch 633/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4523e-05 - val_loss: 1.4194e-05\n",
            "Epoch 634/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3862e-05 - val_loss: 1.4928e-05\n",
            "Epoch 635/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6584e-05 - val_loss: 2.1490e-05\n",
            "Epoch 636/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.6016e-05 - val_loss: 1.1836e-05\n",
            "Epoch 637/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.3659e-05 - val_loss: 1.4641e-05\n",
            "Epoch 638/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.2652e-05 - val_loss: 1.1735e-05\n",
            "Epoch 639/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.1303e-05 - val_loss: 1.2442e-05\n",
            "Epoch 640/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.1718e-05 - val_loss: 1.1788e-05\n",
            "Epoch 641/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.2400e-05 - val_loss: 9.8068e-06\n",
            "Epoch 642/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.0566e-05 - val_loss: 1.3337e-05\n",
            "Epoch 643/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.0952e-05 - val_loss: 1.1017e-05\n",
            "Epoch 644/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.0084e-05 - val_loss: 9.3425e-06\n",
            "Epoch 645/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 9.2654e-06 - val_loss: 9.7930e-06\n",
            "Epoch 646/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 8.9940e-06 - val_loss: 9.8012e-06\n",
            "Epoch 647/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 8.7942e-06 - val_loss: 1.0041e-05\n",
            "Epoch 648/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.1499e-05 - val_loss: 1.2164e-05\n",
            "Epoch 649/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4040e-05 - val_loss: 1.8987e-05\n",
            "Epoch 650/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6349e-05 - val_loss: 2.5226e-05\n",
            "Epoch 651/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.8503e-05 - val_loss: 5.2342e-05\n",
            "Epoch 652/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.3839e-05 - val_loss: 9.5336e-05\n",
            "Epoch 653/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 5.8478e-05 - val_loss: 4.6093e-05\n",
            "Epoch 654/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.4989e-05 - val_loss: 3.5484e-05\n",
            "Epoch 655/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.6272e-05 - val_loss: 3.2729e-05\n",
            "Epoch 656/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2782e-05 - val_loss: 1.7472e-05\n",
            "Epoch 657/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6115e-05 - val_loss: 1.6773e-05\n",
            "Epoch 658/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4693e-05 - val_loss: 2.8949e-05\n",
            "Epoch 659/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.4905e-05 - val_loss: 1.2642e-05\n",
            "Epoch 660/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5154e-05 - val_loss: 1.5422e-05\n",
            "Epoch 661/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.2281e-05 - val_loss: 1.5362e-05\n",
            "Epoch 662/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4946e-05 - val_loss: 1.7679e-05\n",
            "Epoch 663/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.5287e-05 - val_loss: 2.1820e-05\n",
            "Epoch 664/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9080e-05 - val_loss: 1.6852e-05\n",
            "Epoch 665/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.3022e-05 - val_loss: 1.4694e-05\n",
            "Epoch 666/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9359e-05 - val_loss: 2.9262e-05\n",
            "Epoch 667/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.3063e-05 - val_loss: 3.3903e-05\n",
            "Epoch 668/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.0912e-05 - val_loss: 2.4017e-05\n",
            "Epoch 669/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.3034e-05 - val_loss: 1.9984e-05\n",
            "Epoch 670/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2561e-05 - val_loss: 2.8965e-05\n",
            "Epoch 671/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2477e-05 - val_loss: 2.0967e-05\n",
            "Epoch 672/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.5887e-05 - val_loss: 1.5258e-05\n",
            "Epoch 673/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3634e-05 - val_loss: 1.2112e-05\n",
            "Epoch 674/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.1977e-05 - val_loss: 1.0924e-05\n",
            "Epoch 675/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.1700e-05 - val_loss: 1.2039e-05\n",
            "Epoch 676/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.1982e-05 - val_loss: 1.2738e-05\n",
            "Epoch 677/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.3341e-05 - val_loss: 1.6063e-05\n",
            "Epoch 678/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.3072e-05 - val_loss: 1.8478e-05\n",
            "Epoch 679/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.6385e-05 - val_loss: 1.7618e-05\n",
            "Epoch 680/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.9627e-05 - val_loss: 2.5085e-05\n",
            "Epoch 681/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.5743e-05 - val_loss: 2.6083e-05\n",
            "Epoch 682/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.6429e-05 - val_loss: 3.2621e-05\n",
            "Epoch 683/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.8221e-05 - val_loss: 1.8304e-05\n",
            "Epoch 684/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.2075e-05 - val_loss: 4.0978e-05\n",
            "Epoch 685/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.7892e-05 - val_loss: 2.5028e-05\n",
            "Epoch 686/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.1297e-05 - val_loss: 1.8053e-05\n",
            "Epoch 687/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.3319e-05 - val_loss: 1.7248e-05\n",
            "Epoch 688/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.3071e-05 - val_loss: 6.1211e-05\n",
            "Epoch 689/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.6935e-05 - val_loss: 9.4661e-05\n",
            "Epoch 690/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 8.1899e-05 - val_loss: 4.9959e-05\n",
            "Epoch 691/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.7983e-05 - val_loss: 3.8584e-05\n",
            "Epoch 692/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.0158e-05 - val_loss: 4.3881e-05\n",
            "Epoch 693/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.2268e-05 - val_loss: 2.6910e-05\n",
            "Epoch 694/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.9764e-05 - val_loss: 4.2741e-05\n",
            "Epoch 695/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.6542e-05 - val_loss: 4.4366e-05\n",
            "Epoch 696/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.9740e-05 - val_loss: 3.0238e-05\n",
            "Epoch 697/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.5880e-05 - val_loss: 3.1148e-05\n",
            "Epoch 698/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2903e-05 - val_loss: 1.6979e-05\n",
            "Epoch 699/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.7312e-05 - val_loss: 2.6342e-05\n",
            "Epoch 700/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.1177e-05 - val_loss: 1.6364e-05\n",
            "Epoch 701/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7317e-05 - val_loss: 1.5437e-05\n",
            "Epoch 702/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4036e-05 - val_loss: 1.7909e-05\n",
            "Epoch 703/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4740e-05 - val_loss: 1.7651e-05\n",
            "Epoch 704/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.9623e-05 - val_loss: 3.8449e-05\n",
            "Epoch 705/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.7022e-05 - val_loss: 3.3015e-05\n",
            "Epoch 706/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.4977e-05 - val_loss: 2.7493e-05\n",
            "Epoch 707/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0616e-05 - val_loss: 3.2323e-05\n",
            "Epoch 708/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.0163e-05 - val_loss: 2.1246e-05\n",
            "Epoch 709/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.9508e-05 - val_loss: 3.7306e-05\n",
            "Epoch 710/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.6809e-05 - val_loss: 3.9523e-05\n",
            "Epoch 711/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.5179e-05 - val_loss: 3.7462e-05\n",
            "Epoch 712/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.4761e-05 - val_loss: 2.4792e-05\n",
            "Epoch 713/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.7552e-05 - val_loss: 2.5397e-05\n",
            "Epoch 714/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.7705e-05 - val_loss: 1.9750e-05\n",
            "Epoch 715/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.8212e-05 - val_loss: 1.9567e-05\n",
            "Epoch 716/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.3797e-05 - val_loss: 9.2806e-06\n",
            "Epoch 717/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.2906e-05 - val_loss: 1.4833e-05\n",
            "Epoch 718/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6471e-05 - val_loss: 1.7699e-05\n",
            "Epoch 719/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.5674e-05 - val_loss: 1.8783e-05\n",
            "Epoch 720/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6287e-05 - val_loss: 2.4095e-05\n",
            "Epoch 721/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.0530e-05 - val_loss: 1.4605e-05\n",
            "Epoch 722/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.4434e-05 - val_loss: 1.8251e-05\n",
            "Epoch 723/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2703e-05 - val_loss: 2.7209e-05\n",
            "Epoch 724/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.3475e-05 - val_loss: 2.9747e-05\n",
            "Epoch 725/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.1261e-05 - val_loss: 1.8384e-05\n",
            "Epoch 726/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0467e-05 - val_loss: 1.6501e-05\n",
            "Epoch 727/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8525e-05 - val_loss: 1.8598e-05\n",
            "Epoch 728/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0201e-05 - val_loss: 3.0583e-05\n",
            "Epoch 729/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.1352e-05 - val_loss: 2.5876e-05\n",
            "Epoch 730/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.5091e-05 - val_loss: 2.5549e-05\n",
            "Epoch 731/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.3528e-05 - val_loss: 1.6796e-05\n",
            "Epoch 732/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5740e-05 - val_loss: 1.3873e-05\n",
            "Epoch 733/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.8451e-05 - val_loss: 2.3834e-05\n",
            "Epoch 734/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.1266e-05 - val_loss: 1.8637e-05\n",
            "Epoch 735/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3401e-05 - val_loss: 1.3163e-05\n",
            "Epoch 736/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.2085e-05 - val_loss: 1.8977e-05\n",
            "Epoch 737/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.6008e-05 - val_loss: 3.0481e-05\n",
            "Epoch 738/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.0786e-05 - val_loss: 2.8154e-05\n",
            "Epoch 739/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.2278e-05 - val_loss: 3.9754e-05\n",
            "Epoch 740/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 4.0477e-05 - val_loss: 4.0831e-05\n",
            "Epoch 741/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.6270e-05 - val_loss: 7.8728e-05\n",
            "Epoch 742/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.6239e-05 - val_loss: 5.9587e-05\n",
            "Epoch 743/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.3207e-05 - val_loss: 6.3196e-05\n",
            "Epoch 744/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 7.2590e-05 - val_loss: 7.0255e-05\n",
            "Epoch 745/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 8.4282e-05 - val_loss: 8.0957e-05\n",
            "Epoch 746/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 9.9388e-05 - val_loss: 9.1479e-05\n",
            "Epoch 747/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 9.2249e-05 - val_loss: 7.8450e-05\n",
            "Epoch 748/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.5867e-05 - val_loss: 7.8177e-05\n",
            "Epoch 749/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 8.1695e-05 - val_loss: 9.2618e-05\n",
            "Epoch 750/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 7.8944e-05 - val_loss: 9.3618e-05\n",
            "Epoch 751/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.8602e-05 - val_loss: 5.6582e-05\n",
            "Epoch 752/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 4.5029e-05 - val_loss: 4.3005e-05\n",
            "Epoch 753/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.6917e-05 - val_loss: 2.7614e-05\n",
            "Epoch 754/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.6399e-05 - val_loss: 2.7724e-05\n",
            "Epoch 755/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9305e-05 - val_loss: 1.4590e-05\n",
            "Epoch 756/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5267e-05 - val_loss: 1.8970e-05\n",
            "Epoch 757/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5289e-05 - val_loss: 1.3293e-05\n",
            "Epoch 758/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.5341e-05 - val_loss: 1.9214e-05\n",
            "Epoch 759/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.9583e-05 - val_loss: 1.9851e-05\n",
            "Epoch 760/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0659e-05 - val_loss: 2.4365e-05\n",
            "Epoch 761/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9664e-05 - val_loss: 2.0932e-05\n",
            "Epoch 762/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.6751e-05 - val_loss: 1.4536e-05\n",
            "Epoch 763/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.1435e-05 - val_loss: 9.7937e-06\n",
            "Epoch 764/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 8.4035e-06 - val_loss: 7.4272e-06\n",
            "Epoch 765/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 8.6030e-06 - val_loss: 1.0612e-05\n",
            "Epoch 766/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 8.9620e-06 - val_loss: 7.9634e-06\n",
            "Epoch 767/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 8.0748e-06 - val_loss: 7.9018e-06\n",
            "Epoch 768/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 7.7454e-06 - val_loss: 9.2700e-06\n",
            "Epoch 769/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 8.3661e-06 - val_loss: 7.1345e-06\n",
            "Epoch 770/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 7.1696e-06 - val_loss: 8.8706e-06\n",
            "Epoch 771/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 8.0429e-06 - val_loss: 8.5701e-06\n",
            "Epoch 772/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 8.9295e-06 - val_loss: 1.1460e-05\n",
            "Epoch 773/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.0461e-05 - val_loss: 1.2438e-05\n",
            "Epoch 774/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.0173e-05 - val_loss: 8.7328e-06\n",
            "Epoch 775/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 8.7033e-06 - val_loss: 1.4253e-05\n",
            "Epoch 776/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.2540e-05 - val_loss: 1.1412e-05\n",
            "Epoch 777/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.2804e-05 - val_loss: 1.0514e-05\n",
            "Epoch 778/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.8726e-05 - val_loss: 5.1794e-05\n",
            "Epoch 779/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.5424e-05 - val_loss: 5.6196e-05\n",
            "Epoch 780/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.8060e-05 - val_loss: 4.0676e-05\n",
            "Epoch 781/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.8662e-05 - val_loss: 2.7812e-05\n",
            "Epoch 782/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.9439e-05 - val_loss: 2.4986e-05\n",
            "Epoch 783/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.8771e-05 - val_loss: 1.4735e-05\n",
            "Epoch 784/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5398e-05 - val_loss: 1.6493e-05\n",
            "Epoch 785/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5213e-05 - val_loss: 1.4686e-05\n",
            "Epoch 786/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3489e-05 - val_loss: 1.2869e-05\n",
            "Epoch 787/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.1127e-05 - val_loss: 1.1287e-05\n",
            "Epoch 788/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.0583e-05 - val_loss: 9.0595e-06\n",
            "Epoch 789/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 9.9486e-06 - val_loss: 1.1514e-05\n",
            "Epoch 790/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.2331e-05 - val_loss: 1.0678e-05\n",
            "Epoch 791/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.2167e-05 - val_loss: 8.8428e-06\n",
            "Epoch 792/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 9.1866e-06 - val_loss: 1.3566e-05\n",
            "Epoch 793/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.0092e-05 - val_loss: 1.0685e-05\n",
            "Epoch 794/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.0200e-05 - val_loss: 1.2957e-05\n",
            "Epoch 795/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 9.6208e-06 - val_loss: 1.1380e-05\n",
            "Epoch 796/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.0949e-05 - val_loss: 1.0413e-05\n",
            "Epoch 797/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.0120e-05 - val_loss: 9.9334e-06\n",
            "Epoch 798/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.0237e-05 - val_loss: 1.3457e-05\n",
            "Epoch 799/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.1180e-05 - val_loss: 1.2023e-05\n",
            "Epoch 800/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4944e-05 - val_loss: 1.0280e-05\n",
            "Epoch 801/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.3758e-05 - val_loss: 1.5240e-05\n",
            "Epoch 802/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3111e-05 - val_loss: 1.6677e-05\n",
            "Epoch 803/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6432e-05 - val_loss: 1.6496e-05\n",
            "Epoch 804/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4587e-05 - val_loss: 1.5775e-05\n",
            "Epoch 805/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.4966e-05 - val_loss: 1.3271e-05\n",
            "Epoch 806/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7269e-05 - val_loss: 1.5885e-05\n",
            "Epoch 807/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0384e-05 - val_loss: 2.2067e-05\n",
            "Epoch 808/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8794e-05 - val_loss: 1.2113e-05\n",
            "Epoch 809/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.1576e-05 - val_loss: 1.2214e-05\n",
            "Epoch 810/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.0907e-05 - val_loss: 1.2182e-05\n",
            "Epoch 811/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.2054e-05 - val_loss: 1.2032e-05\n",
            "Epoch 812/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4815e-05 - val_loss: 1.7142e-05\n",
            "Epoch 813/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3647e-05 - val_loss: 1.3326e-05\n",
            "Epoch 814/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3292e-05 - val_loss: 1.8757e-05\n",
            "Epoch 815/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5577e-05 - val_loss: 1.2098e-05\n",
            "Epoch 816/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.2821e-05 - val_loss: 1.8774e-05\n",
            "Epoch 817/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5743e-05 - val_loss: 1.7838e-05\n",
            "Epoch 818/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.7792e-05 - val_loss: 1.4315e-05\n",
            "Epoch 819/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.2947e-05 - val_loss: 2.3341e-05\n",
            "Epoch 820/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.1912e-05 - val_loss: 2.5681e-05\n",
            "Epoch 821/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.9500e-05 - val_loss: 1.5486e-05\n",
            "Epoch 822/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.5305e-05 - val_loss: 1.0706e-05\n",
            "Epoch 823/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.3767e-05 - val_loss: 1.3632e-05\n",
            "Epoch 824/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.3564e-05 - val_loss: 1.9229e-05\n",
            "Epoch 825/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.0629e-05 - val_loss: 1.8506e-05\n",
            "Epoch 826/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.0046e-05 - val_loss: 2.6994e-05\n",
            "Epoch 827/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.6113e-05 - val_loss: 2.7115e-05\n",
            "Epoch 828/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.2073e-05 - val_loss: 2.9883e-05\n",
            "Epoch 829/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.0074e-05 - val_loss: 4.4910e-05\n",
            "Epoch 830/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 6.4969e-05 - val_loss: 8.7275e-05\n",
            "Epoch 831/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.0308e-04 - val_loss: 6.9963e-05\n",
            "Epoch 832/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.1086e-04 - val_loss: 2.9030e-04\n",
            "Epoch 833/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.8169e-04 - val_loss: 0.0364\n",
            "Epoch 834/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 0.1640 - val_loss: 0.0465\n",
            "Epoch 835/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0431 - val_loss: 0.0289\n",
            "Epoch 836/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 0.0267 - val_loss: 0.0220\n",
            "Epoch 837/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 0.0200 - val_loss: 0.0176\n",
            "Epoch 838/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0120 - val_loss: 0.0086\n",
            "Epoch 839/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 0.0076 - val_loss: 0.0063\n",
            "Epoch 840/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0049 - val_loss: 0.0037\n",
            "Epoch 841/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 0.0037 - val_loss: 0.0027\n",
            "Epoch 842/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0028 - val_loss: 0.0026\n",
            "Epoch 843/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 0.0021 - val_loss: 0.0019\n",
            "Epoch 844/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 0.0016 - val_loss: 0.0015\n",
            "Epoch 845/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 0.0014 - val_loss: 0.0015\n",
            "Epoch 846/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 0.0013 - val_loss: 0.0012\n",
            "Epoch 847/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 0.0011 - val_loss: 0.0011\n",
            "Epoch 848/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 0.0010 - val_loss: 0.0010\n",
            "Epoch 849/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 9.2056e-04 - val_loss: 9.9258e-04\n",
            "Epoch 850/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 9.0869e-04 - val_loss: 9.7529e-04\n",
            "Epoch 851/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 8.3036e-04 - val_loss: 8.0696e-04\n",
            "Epoch 852/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 7.4397e-04 - val_loss: 7.8883e-04\n",
            "Epoch 853/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 7.1409e-04 - val_loss: 7.4126e-04\n",
            "Epoch 854/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 6.7259e-04 - val_loss: 7.0154e-04\n",
            "Epoch 855/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 6.4559e-04 - val_loss: 6.8982e-04\n",
            "Epoch 856/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 6.3717e-04 - val_loss: 6.5670e-04\n",
            "Epoch 857/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 6.0941e-04 - val_loss: 6.3864e-04\n",
            "Epoch 858/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.8799e-04 - val_loss: 6.0698e-04\n",
            "Epoch 859/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.4980e-04 - val_loss: 5.7478e-04\n",
            "Epoch 860/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.3982e-04 - val_loss: 5.7743e-04\n",
            "Epoch 861/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.1663e-04 - val_loss: 5.7922e-04\n",
            "Epoch 862/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 5.1872e-04 - val_loss: 5.4464e-04\n",
            "Epoch 863/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.9860e-04 - val_loss: 5.0959e-04\n",
            "Epoch 864/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.5970e-04 - val_loss: 4.8579e-04\n",
            "Epoch 865/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.4651e-04 - val_loss: 4.8961e-04\n",
            "Epoch 866/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.4584e-04 - val_loss: 4.6246e-04\n",
            "Epoch 867/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.2452e-04 - val_loss: 4.4204e-04\n",
            "Epoch 868/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 4.1276e-04 - val_loss: 4.4731e-04\n",
            "Epoch 869/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.2508e-04 - val_loss: 4.7419e-04\n",
            "Epoch 870/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.2226e-04 - val_loss: 4.3578e-04\n",
            "Epoch 871/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.9675e-04 - val_loss: 4.1193e-04\n",
            "Epoch 872/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.8023e-04 - val_loss: 3.9560e-04\n",
            "Epoch 873/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.6759e-04 - val_loss: 4.1727e-04\n",
            "Epoch 874/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.7244e-04 - val_loss: 4.5305e-04\n",
            "Epoch 875/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.1855e-04 - val_loss: 3.7862e-04\n",
            "Epoch 876/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 4.2487e-04 - val_loss: 4.6319e-04\n",
            "Epoch 877/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.8303e-04 - val_loss: 3.7394e-04\n",
            "Epoch 878/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.5137e-04 - val_loss: 3.6537e-04\n",
            "Epoch 879/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.3456e-04 - val_loss: 3.4693e-04\n",
            "Epoch 880/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.1748e-04 - val_loss: 3.6176e-04\n",
            "Epoch 881/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.2722e-04 - val_loss: 3.5150e-04\n",
            "Epoch 882/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.2491e-04 - val_loss: 3.6249e-04\n",
            "Epoch 883/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.1478e-04 - val_loss: 3.3176e-04\n",
            "Epoch 884/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.0272e-04 - val_loss: 3.3650e-04\n",
            "Epoch 885/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.0331e-04 - val_loss: 3.1482e-04\n",
            "Epoch 886/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.8449e-04 - val_loss: 3.0437e-04\n",
            "Epoch 887/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.9044e-04 - val_loss: 3.2573e-04\n",
            "Epoch 888/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.0192e-04 - val_loss: 3.2186e-04\n",
            "Epoch 889/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.9247e-04 - val_loss: 3.0141e-04\n",
            "Epoch 890/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.7162e-04 - val_loss: 2.9911e-04\n",
            "Epoch 891/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.6317e-04 - val_loss: 2.7493e-04\n",
            "Epoch 892/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.6694e-04 - val_loss: 2.8571e-04\n",
            "Epoch 893/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.7115e-04 - val_loss: 2.8036e-04\n",
            "Epoch 894/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.7067e-04 - val_loss: 2.6566e-04\n",
            "Epoch 895/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.6705e-04 - val_loss: 2.8805e-04\n",
            "Epoch 896/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.5996e-04 - val_loss: 2.6671e-04\n",
            "Epoch 897/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.4582e-04 - val_loss: 2.7467e-04\n",
            "Epoch 898/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.5133e-04 - val_loss: 2.6058e-04\n",
            "Epoch 899/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.4860e-04 - val_loss: 2.9758e-04\n",
            "Epoch 900/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.5216e-04 - val_loss: 2.4617e-04\n",
            "Epoch 901/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.3359e-04 - val_loss: 2.5455e-04\n",
            "Epoch 902/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.4732e-04 - val_loss: 2.4794e-04\n",
            "Epoch 903/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.4099e-04 - val_loss: 2.3571e-04\n",
            "Epoch 904/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.2614e-04 - val_loss: 2.4395e-04\n",
            "Epoch 905/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.3284e-04 - val_loss: 2.7429e-04\n",
            "Epoch 906/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.4172e-04 - val_loss: 2.5622e-04\n",
            "Epoch 907/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2312e-04 - val_loss: 2.4546e-04\n",
            "Epoch 908/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.2545e-04 - val_loss: 2.3796e-04\n",
            "Epoch 909/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.2280e-04 - val_loss: 2.2674e-04\n",
            "Epoch 910/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.2482e-04 - val_loss: 2.7165e-04\n",
            "Epoch 911/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2542e-04 - val_loss: 2.6275e-04\n",
            "Epoch 912/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2690e-04 - val_loss: 2.2507e-04\n",
            "Epoch 913/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.5436e-04 - val_loss: 2.6588e-04\n",
            "Epoch 914/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.6635e-04 - val_loss: 2.2980e-04\n",
            "Epoch 915/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.3468e-04 - val_loss: 2.1653e-04\n",
            "Epoch 916/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.1111e-04 - val_loss: 2.3496e-04\n",
            "Epoch 917/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.1396e-04 - val_loss: 2.1539e-04\n",
            "Epoch 918/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0616e-04 - val_loss: 2.1159e-04\n",
            "Epoch 919/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.1831e-04 - val_loss: 2.1918e-04\n",
            "Epoch 920/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.3281e-04 - val_loss: 2.0545e-04\n",
            "Epoch 921/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9388e-04 - val_loss: 1.8991e-04\n",
            "Epoch 922/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.9772e-04 - val_loss: 1.9795e-04\n",
            "Epoch 923/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8518e-04 - val_loss: 1.8974e-04\n",
            "Epoch 924/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9393e-04 - val_loss: 2.0280e-04\n",
            "Epoch 925/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9083e-04 - val_loss: 2.6414e-04\n",
            "Epoch 926/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2402e-04 - val_loss: 1.8486e-04\n",
            "Epoch 927/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.4266e-04 - val_loss: 2.0029e-04\n",
            "Epoch 928/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7793e-04 - val_loss: 1.9896e-04\n",
            "Epoch 929/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9794e-04 - val_loss: 2.3746e-04\n",
            "Epoch 930/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9720e-04 - val_loss: 2.4137e-04\n",
            "Epoch 931/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.1746e-04 - val_loss: 1.7040e-04\n",
            "Epoch 932/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7041e-04 - val_loss: 1.7628e-04\n",
            "Epoch 933/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6046e-04 - val_loss: 1.9057e-04\n",
            "Epoch 934/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9073e-04 - val_loss: 1.6392e-04\n",
            "Epoch 935/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0311e-04 - val_loss: 2.7362e-04\n",
            "Epoch 936/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.0552e-04 - val_loss: 2.6985e-04\n",
            "Epoch 937/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9134e-04 - val_loss: 2.4423e-04\n",
            "Epoch 938/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.9907e-04 - val_loss: 2.1519e-04\n",
            "Epoch 939/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.3441e-04 - val_loss: 1.7124e-04\n",
            "Epoch 940/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9366e-04 - val_loss: 1.6947e-04\n",
            "Epoch 941/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7205e-04 - val_loss: 1.7963e-04\n",
            "Epoch 942/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7182e-04 - val_loss: 1.8883e-04\n",
            "Epoch 943/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6738e-04 - val_loss: 2.2384e-04\n",
            "Epoch 944/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6616e-04 - val_loss: 1.6225e-04\n",
            "Epoch 945/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4942e-04 - val_loss: 1.6940e-04\n",
            "Epoch 946/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4889e-04 - val_loss: 1.5213e-04\n",
            "Epoch 947/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4655e-04 - val_loss: 1.6536e-04\n",
            "Epoch 948/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6235e-04 - val_loss: 2.1218e-04\n",
            "Epoch 949/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.8521e-04 - val_loss: 2.0241e-04\n",
            "Epoch 950/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.7205e-04 - val_loss: 1.8003e-04\n",
            "Epoch 951/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.7166e-04 - val_loss: 1.5341e-04\n",
            "Epoch 952/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4645e-04 - val_loss: 1.5933e-04\n",
            "Epoch 953/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.4192e-04 - val_loss: 1.5307e-04\n",
            "Epoch 954/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4683e-04 - val_loss: 1.7627e-04\n",
            "Epoch 955/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5593e-04 - val_loss: 1.6099e-04\n",
            "Epoch 956/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6694e-04 - val_loss: 1.3939e-04\n",
            "Epoch 957/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.5314e-04 - val_loss: 1.4842e-04\n",
            "Epoch 958/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6340e-04 - val_loss: 2.8203e-04\n",
            "Epoch 959/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 3.1145e-04 - val_loss: 2.8157e-04\n",
            "Epoch 960/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.3177e-04 - val_loss: 1.5084e-04\n",
            "Epoch 961/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4510e-04 - val_loss: 1.6903e-04\n",
            "Epoch 962/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.3532e-04 - val_loss: 2.3731e-04\n",
            "Epoch 963/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.8208e-04 - val_loss: 1.6914e-04\n",
            "Epoch 964/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.5689e-04 - val_loss: 1.7715e-04\n",
            "Epoch 965/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.5429e-04 - val_loss: 1.6847e-04\n",
            "Epoch 966/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.6785e-04 - val_loss: 2.4386e-04\n",
            "Epoch 967/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.0935e-04 - val_loss: 2.7991e-04\n",
            "Epoch 968/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9836e-04 - val_loss: 1.3879e-04\n",
            "Epoch 969/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.4430e-04 - val_loss: 1.8349e-04\n",
            "Epoch 970/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.9009e-04 - val_loss: 1.9565e-04\n",
            "Epoch 971/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.7012e-04 - val_loss: 1.4340e-04\n",
            "Epoch 972/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.2485e-04 - val_loss: 1.6960e-04\n",
            "Epoch 973/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.5745e-04 - val_loss: 1.5657e-04\n",
            "Epoch 974/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.8447e-04 - val_loss: 2.3586e-04\n",
            "Epoch 975/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.2008e-04 - val_loss: 2.8407e-04\n",
            "Epoch 976/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.8141e-04 - val_loss: 2.2568e-04\n",
            "Epoch 977/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.3807e-04 - val_loss: 1.4064e-04\n",
            "Epoch 978/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.6693e-04 - val_loss: 2.4214e-04\n",
            "Epoch 979/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.9567e-04 - val_loss: 1.4240e-04\n",
            "Epoch 980/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.5727e-04 - val_loss: 1.4038e-04\n",
            "Epoch 981/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.7525e-04 - val_loss: 3.2884e-04\n",
            "Epoch 982/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 2.5756e-04 - val_loss: 1.7287e-04\n",
            "Epoch 983/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.4810e-04 - val_loss: 2.4010e-04\n",
            "Epoch 984/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2401e-04 - val_loss: 1.7925e-04\n",
            "Epoch 985/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.4047e-04 - val_loss: 1.4344e-04\n",
            "Epoch 986/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3457e-04 - val_loss: 1.1771e-04\n",
            "Epoch 987/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3993e-04 - val_loss: 1.2107e-04\n",
            "Epoch 988/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.2815e-04 - val_loss: 1.3389e-04\n",
            "Epoch 989/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.3401e-04 - val_loss: 1.2261e-04\n",
            "Epoch 990/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.0968e-04 - val_loss: 1.3668e-04\n",
            "Epoch 991/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.4757e-04 - val_loss: 1.1540e-04\n",
            "Epoch 992/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.1380e-04 - val_loss: 1.4303e-04\n",
            "Epoch 993/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3728e-04 - val_loss: 2.0447e-04\n",
            "Epoch 994/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 3.4035e-04 - val_loss: 7.7827e-04\n",
            "Epoch 995/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 7.7253e-04 - val_loss: 2.2787e-04\n",
            "Epoch 996/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.4853e-04 - val_loss: 2.9437e-04\n",
            "Epoch 997/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 2.2219e-04 - val_loss: 1.6115e-04\n",
            "Epoch 998/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.3647e-04 - val_loss: 1.3691e-04\n",
            "Epoch 999/1000\n",
            "6/6 [==============================] - 8s 1s/step - loss: 1.1998e-04 - val_loss: 1.3340e-04\n",
            "Epoch 1000/1000\n",
            "6/6 [==============================] - 7s 1s/step - loss: 1.1121e-04 - val_loss: 1.2346e-04\n",
            "Model: \"sequential_45\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "simple_rnn_38 (SimpleRNN)    (None, 1439, 256)         68096     \n",
            "_________________________________________________________________\n",
            "dense_51 (Dense)             (None, 1439, 3)           771       \n",
            "=================================================================\n",
            "Total params: 68,867\n",
            "Trainable params: 68,867\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "Metrics for LSTM model:\n",
            "RMSE: Precursor: 0.00181, Gas: 0.00050, Aerosols: 0.00104\n",
            "R2: Precursor: 0.96509, Gas: 0.99959, Aerosols: 0.99779\n",
            "Hellenger Distance: Precursor: 0.00005, Gas: 0.00004, Aerosols: 0.00121\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qLu_mYLb3JXG",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "outputId": "df7e4ec1-20ad-47c6-c352-1487d37445e5"
      },
      "source": [
        "def plot_history():\n",
        "    plt.figure()\n",
        "    plt.xlabel('Epoch')\n",
        "    plt.ylabel('Mean Square Error')\n",
        "    plt.plot(hist['epoch'], hist['loss'], label='Train Error')\n",
        "    plt.plot(hist['epoch'], hist['val_loss'], label = 'Val Error')\n",
        "    plt.legend()\n",
        "\n",
        "hist = pd.DataFrame(history.history)\n",
        "hist['epoch'] = history.epoch\n",
        "\n",
        "plot_history()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VU99eQFM3bua",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 283
        },
        "outputId": "d9f42f90-515a-46d6-85c8-978006a6e8fe"
      },
      "source": [
        "def plot_history_short():\n",
        "    plt.figure()\n",
        "    plt.xlabel('Epoch')\n",
        "    plt.ylabel('Mean Square Error')\n",
        "    plt.plot(hist['epoch'], hist['loss'], label='Train Error')\n",
        "    plt.plot(hist['epoch'], hist['val_loss'], label = 'Val Error')\n",
        "    plt.legend()\n",
        "    plt.xlim([0,100])\n",
        "    plt.ylim([0,0.10])\n",
        "\n",
        "plot_history_short()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AT76H8TMUN2t"
      },
      "source": [
        "\n",
        "## Compare interpretation of baseline with your approach"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lMn1R1MwyVY1",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 238
        },
        "outputId": "15460896-3076-43a7-ef26-571e886f4b4d"
      },
      "source": [
        "j=0\n",
        "print('====================================')\n",
        "print(output_columns[j])\n",
        "print('====================================')\n",
        "foo = scores_val[output_columns[j]]\n",
        "for col in foo.columns:\n",
        "    foo[col] = pd.to_numeric(foo[col])\n",
        "foo.style.background_gradient(cmap='Reds', subset=['MAE', 'RMSE', 'Hellinger Distance']).background_gradient(cmap='Greens', subset=['R'])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "====================================\n",
            "Precursor [ug/m3]\n",
            "====================================\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<style  type=\"text/css\" >\n",
              "    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row0_col0 {\n",
              "            background-color:  #67000d;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row0_col1 {\n",
              "            background-color:  #67000d;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row0_col2 {\n",
              "            background-color:  #157f3b;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row0_col3 {\n",
              "            background-color:  #fff5f0;\n",
              "            color:  #000000;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row1_col0 {\n",
              "            background-color:  #cb181d;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row1_col1 {\n",
              "            background-color:  #bc141a;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row1_col2 {\n",
              "            background-color:  #006529;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row1_col3 {\n",
              "            background-color:  #fff5f0;\n",
              "            color:  #000000;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row2_col0 {\n",
              "            background-color:  #c7171c;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row2_col1 {\n",
              "            background-color:  #b61319;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row2_col2 {\n",
              "            background-color:  #006529;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row2_col3 {\n",
              "            background-color:  #fff5f0;\n",
              "            color:  #000000;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row3_col0 {\n",
              "            background-color:  #77040f;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row3_col1 {\n",
              "            background-color:  #7e0610;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row3_col2 {\n",
              "            background-color:  #00441b;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row3_col3 {\n",
              "            background-color:  #fff5f0;\n",
              "            color:  #000000;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row4_col0 {\n",
              "            background-color:  #75030f;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row4_col1 {\n",
              "            background-color:  #7c0510;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row4_col2 {\n",
              "            background-color:  #00471c;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row4_col3 {\n",
              "            background-color:  #fff5f0;\n",
              "            color:  #000000;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row5_col0 {\n",
              "            background-color:  #fff5f0;\n",
              "            color:  #000000;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row5_col1 {\n",
              "            background-color:  #fff5f0;\n",
              "            color:  #000000;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row5_col2 {\n",
              "            background-color:  #f7fcf5;\n",
              "            color:  #000000;\n",
              "        }    #T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row5_col3 {\n",
              "            background-color:  #67000d;\n",
              "            color:  #f1f1f1;\n",
              "        }</style><table id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002\" ><thead>    <tr>        <th class=\"blank level0\" ></th>        <th class=\"col_heading level0 col0\" >MAE</th>        <th class=\"col_heading level0 col1\" >RMSE</th>        <th class=\"col_heading level0 col2\" >R</th>        <th class=\"col_heading level0 col3\" >Hellinger Distance</th>        <th class=\"col_heading level0 col4\" >STDVAR</th>        <th class=\"col_heading level0 col5\" >Truth STDVAR</th>    </tr>    <tr>        <th class=\"index_name level0\" >Model Type</th>        <th class=\"blank\" ></th>        <th class=\"blank\" ></th>        <th class=\"blank\" ></th>        <th class=\"blank\" ></th>        <th class=\"blank\" ></th>        <th class=\"blank\" ></th>    </tr></thead><tbody>\n",
              "                <tr>\n",
              "                        <th id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002level0_row0\" class=\"row_heading level0 row0\" >Linear Val</th>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row0_col0\" class=\"data row0 col0\" >0.005001</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row0_col1\" class=\"data row0 col1\" >0.013297</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row0_col2\" class=\"data row0 col2\" >0.001411</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row0_col3\" class=\"data row0 col3\" >0.000012</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row0_col4\" class=\"data row0 col4\" >0.009640</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row0_col5\" class=\"data row0 col5\" >0.009633</td>\n",
              "            </tr>\n",
              "            <tr>\n",
              "                        <th id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002level0_row1\" class=\"row_heading level0 row1\" >DNN Val</th>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row1_col0\" class=\"data row1 col0\" >0.004986</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row1_col1\" class=\"data row1 col1\" >0.013287</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row1_col2\" class=\"data row1 col2\" >0.001939</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row1_col3\" class=\"data row1 col3\" >0.000001</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row1_col4\" class=\"data row1 col4\" >0.009637</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row1_col5\" class=\"data row1 col5\" >0.009633</td>\n",
              "            </tr>\n",
              "            <tr>\n",
              "                        <th id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002level0_row2\" class=\"row_heading level0 row2\" >RNN Val</th>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row2_col0\" class=\"data row2 col0\" >0.004987</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row2_col1\" class=\"data row2 col1\" >0.013288</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row2_col2\" class=\"data row2 col2\" >0.001937</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row2_col3\" class=\"data row2 col3\" >0.000002</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row2_col4\" class=\"data row2 col4\" >0.009643</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row2_col5\" class=\"data row2 col5\" >0.009633</td>\n",
              "            </tr>\n",
              "            <tr>\n",
              "                        <th id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002level0_row3\" class=\"row_heading level0 row3\" >Simple RNN Val</th>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row3_col0\" class=\"data row3 col0\" >0.004999</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row3_col1\" class=\"data row3 col1\" >0.013295</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row3_col2\" class=\"data row3 col2\" >0.002481</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row3_col3\" class=\"data row3 col3\" >0.000002</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row3_col4\" class=\"data row3 col4\" >0.009655</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row3_col5\" class=\"data row3 col5\" >0.009633</td>\n",
              "            </tr>\n",
              "            <tr>\n",
              "                        <th id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002level0_row4\" class=\"row_heading level0 row4\" >LSTM Val</th>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row4_col0\" class=\"data row4 col0\" >0.004999</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row4_col1\" class=\"data row4 col1\" >0.013295</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row4_col2\" class=\"data row4 col2\" >0.002426</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row4_col3\" class=\"data row4 col3\" >0.000001</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row4_col4\" class=\"data row4 col4\" >0.009654</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row4_col5\" class=\"data row4 col5\" >0.009633</td>\n",
              "            </tr>\n",
              "            <tr>\n",
              "                        <th id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002level0_row5\" class=\"row_heading level0 row5\" >Random Forest Val</th>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row5_col0\" class=\"data row5 col0\" >0.004940</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row5_col1\" class=\"data row5 col1\" >0.013250</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row5_col2\" class=\"data row5 col2\" >-0.002852</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row5_col3\" class=\"data row5 col3\" >0.043186</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row5_col4\" class=\"data row5 col4\" >0.009509</td>\n",
              "                        <td id=\"T_9e7a2ef4_b743_11ea_a58a_0242ac1c0002row5_col5\" class=\"data row5 col5\" >0.009633</td>\n",
              "            </tr>\n",
              "    </tbody></table>"
            ],
            "text/plain": [
              "<pandas.io.formats.style.Styler at 0x7f4daa43f550>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 145
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "b6aecGyqyccH",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 238
        },
        "outputId": "565ae4b0-f386-47e7-ca5b-da7f9aa364b0"
      },
      "source": [
        "j=0\n",
        "print('====================================')\n",
        "print(output_columns[j])\n",
        "print('====================================')\n",
        "foo = scores_train[output_columns[j]]\n",
        "for col in foo.columns:\n",
        "    foo[col] = pd.to_numeric(foo[col])\n",
        "foo.style.background_gradient(cmap='Reds', subset=['MAE', 'RMSE', 'Hellinger Distance']).background_gradient(cmap='Greens', subset=['R'])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "====================================\n",
            "Precursor [ug/m3]\n",
            "====================================\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<style  type=\"text/css\" >\n",
              "    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row0_col0 {\n",
              "            background-color:  #9c0d14;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row0_col1 {\n",
              "            background-color:  #79040f;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row0_col2 {\n",
              "            background-color:  #f7fcf5;\n",
              "            color:  #000000;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row0_col3 {\n",
              "            background-color:  #fff5f0;\n",
              "            color:  #000000;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row1_col0 {\n",
              "            background-color:  #cb181d;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row1_col1 {\n",
              "            background-color:  #8a0812;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row1_col2 {\n",
              "            background-color:  #87cd86;\n",
              "            color:  #000000;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row1_col3 {\n",
              "            background-color:  #fff5f0;\n",
              "            color:  #000000;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row2_col0 {\n",
              "            background-color:  #e63328;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row2_col1 {\n",
              "            background-color:  #840711;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row2_col2 {\n",
              "            background-color:  #edf8e9;\n",
              "            color:  #000000;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row2_col3 {\n",
              "            background-color:  #fff5f0;\n",
              "            color:  #000000;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row3_col0 {\n",
              "            background-color:  #67000d;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row3_col1 {\n",
              "            background-color:  #67000d;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row3_col2 {\n",
              "            background-color:  #d4eece;\n",
              "            color:  #000000;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row3_col3 {\n",
              "            background-color:  #fff5f0;\n",
              "            color:  #000000;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row4_col0 {\n",
              "            background-color:  #6b010e;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row4_col1 {\n",
              "            background-color:  #67000d;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row4_col2 {\n",
              "            background-color:  #d6efd0;\n",
              "            color:  #000000;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row4_col3 {\n",
              "            background-color:  #fff5f0;\n",
              "            color:  #000000;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row5_col0 {\n",
              "            background-color:  #fff5f0;\n",
              "            color:  #000000;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row5_col1 {\n",
              "            background-color:  #fff5f0;\n",
              "            color:  #000000;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row5_col2 {\n",
              "            background-color:  #00441b;\n",
              "            color:  #f1f1f1;\n",
              "        }    #T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row5_col3 {\n",
              "            background-color:  #67000d;\n",
              "            color:  #f1f1f1;\n",
              "        }</style><table id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002\" ><thead>    <tr>        <th class=\"blank level0\" ></th>        <th class=\"col_heading level0 col0\" >MAE</th>        <th class=\"col_heading level0 col1\" >RMSE</th>        <th class=\"col_heading level0 col2\" >R</th>        <th class=\"col_heading level0 col3\" >Hellinger Distance</th>        <th class=\"col_heading level0 col4\" >STDVAR</th>        <th class=\"col_heading level0 col5\" >Truth STDVAR</th>    </tr>    <tr>        <th class=\"index_name level0\" >Model Type</th>        <th class=\"blank\" ></th>        <th class=\"blank\" ></th>        <th class=\"blank\" ></th>        <th class=\"blank\" ></th>        <th class=\"blank\" ></th>        <th class=\"blank\" ></th>    </tr></thead><tbody>\n",
              "                <tr>\n",
              "                        <th id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002level0_row0\" class=\"row_heading level0 row0\" >Linear Train</th>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row0_col0\" class=\"data row0 col0\" >0.005334</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row0_col1\" class=\"data row0 col1\" >0.013560</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row0_col2\" class=\"data row0 col2\" >-0.019108</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row0_col3\" class=\"data row0 col3\" >0.000066</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row0_col4\" class=\"data row0 col4\" >0.009846</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row0_col5\" class=\"data row0 col5\" >0.009633</td>\n",
              "            </tr>\n",
              "            <tr>\n",
              "                        <th id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002level0_row1\" class=\"row_heading level0 row1\" >DNN Train</th>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row1_col0\" class=\"data row1 col0\" >0.005329</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row1_col1\" class=\"data row1 col1\" >0.013556</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row1_col2\" class=\"data row1 col2\" >-0.018503</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row1_col3\" class=\"data row1 col3\" >0.000063</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row1_col4\" class=\"data row1 col4\" >0.009892</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row1_col5\" class=\"data row1 col5\" >0.009633</td>\n",
              "            </tr>\n",
              "            <tr>\n",
              "                        <th id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002level0_row2\" class=\"row_heading level0 row2\" >RNN Train</th>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row2_col0\" class=\"data row2 col0\" >0.005325</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row2_col1\" class=\"data row2 col1\" >0.013558</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row2_col2\" class=\"data row2 col2\" >-0.019011</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row2_col3\" class=\"data row2 col3\" >0.000046</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row2_col4\" class=\"data row2 col4\" >0.009895</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row2_col5\" class=\"data row2 col5\" >0.009633</td>\n",
              "            </tr>\n",
              "            <tr>\n",
              "                        <th id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002level0_row3\" class=\"row_heading level0 row3\" >Simple RNN Train</th>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row3_col0\" class=\"data row3 col0\" >0.005337</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row3_col1\" class=\"data row3 col1\" >0.013564</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row3_col2\" class=\"data row3 col2\" >-0.018839</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row3_col3\" class=\"data row3 col3\" >0.000066</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row3_col4\" class=\"data row3 col4\" >0.009914</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row3_col5\" class=\"data row3 col5\" >0.009633</td>\n",
              "            </tr>\n",
              "            <tr>\n",
              "                        <th id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002level0_row4\" class=\"row_heading level0 row4\" >LSTM Train</th>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row4_col0\" class=\"data row4 col0\" >0.005337</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row4_col1\" class=\"data row4 col1\" >0.013564</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row4_col2\" class=\"data row4 col2\" >-0.018851</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row4_col3\" class=\"data row4 col3\" >0.000065</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row4_col4\" class=\"data row4 col4\" >0.009913</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row4_col5\" class=\"data row4 col5\" >0.009633</td>\n",
              "            </tr>\n",
              "            <tr>\n",
              "                        <th id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002level0_row5\" class=\"row_heading level0 row5\" >Random Forest Train</th>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row5_col0\" class=\"data row5 col0\" >0.005303</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row5_col1\" class=\"data row5 col1\" >0.013452</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row5_col2\" class=\"data row5 col2\" >-0.017753</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row5_col3\" class=\"data row5 col3\" >0.042422</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row5_col4\" class=\"data row5 col4\" >0.009722</td>\n",
              "                        <td id=\"T_a8ad97a8_b743_11ea_a58a_0242ac1c0002row5_col5\" class=\"data row5 col5\" >0.009633</td>\n",
              "            </tr>\n",
              "    </tbody></table>"
            ],
            "text/plain": [
              "<pandas.io.formats.style.Styler at 0x7f4cc10b70b8>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 146
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "tXWN-a4gUO1V"
      },
      "source": [
        "\n",
        "## Submit best results on project to leaderboard"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ps4Z69v5fW2P"
      },
      "source": [
        "# New Section"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Z8wvXC8sUPoU"
      },
      "source": [
        "\n",
        "## Prepare 2 Google Slides on team's approach and submit them \n",
        "\n",
        "[Slides](https://docs.google.com/presentation/d/1DFd5SPJTGjpRcuIGF9-72q-eDUBZBjES7WHLHv9-eb8/edit?usp=sharing)"
      ]
    }
  ]
}