notebooks/7_neural_networks.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-05-20T12:28:32.643440Z",
     "start_time": "2018-05-20T12:28:32.606527Z"
    }
   },
   "source": [
    "### Neural Networks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-28T22:46:09.054032Z",
     "start_time": "2019-06-28T22:46:09.048101Z"
    }
   },
   "outputs": [],
   "source": [
    "import os\n",
    "import glob\n",
    "import warnings\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from tqdm import tqdm_notebook as tqdm\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "from tensorflow import set_random_seed\n",
    "\n",
    "from keras import backend as K\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout\n",
    "from keras.wrappers.scikit_learn import KerasRegressor\n",
    "from keras.callbacks import ModelCheckpoint\n",
    "\n",
    "import utils\n",
    "from constants import DATA_DIR"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-28T22:46:09.707601Z",
     "start_time": "2019-06-28T22:46:09.695879Z"
    }
   },
   "outputs": [],
   "source": [
    "np.random.seed(23)\n",
    "set_random_seed(23)\n",
    "\n",
    "warnings.filterwarnings(\"ignore\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-28T22:46:10.330037Z",
     "start_time": "2019-06-28T22:46:10.305936Z"
    }
   },
   "outputs": [],
   "source": [
    "def root_mean_squared_error(y_true, y_pred):\n",
    "    # Custom loss function for keras\n",
    "    return K.sqrt(K.mean(K.square(y_pred - y_true)))\n",
    "\n",
    "\n",
    "def get_modelcheckpoint_path(model_num):\n",
    "    # Create a file path for a model and save models in hdf5 files with datetime, validation losses and epochs\n",
    "    parent = DATA_DIR + \"/Models/NN/Model_{}/\".format(model_num)\n",
    "    child = (\n",
    "        pd.Timestamp.now().strftime(\"%Y%m%d-%Hh%Mm\")\n",
    "        + \"-model-epoch_{epoch:02d}-rmse_{val_loss:.5f}.hdf5\"\n",
    "    )\n",
    "    return parent + child\n",
    "\n",
    "\n",
    "def get_weights_path_and_epoch(model_num):\n",
    "    filepaths = glob.glob(\n",
    "        DATA_DIR + \"/Models/NN/Model_{}/*.hdf5\".format(str(model_num))\n",
    "    )\n",
    "    losses = [float(filepath[-12:-5]) for filepath in filepaths]\n",
    "    epochs = losses.index(min(losses))\n",
    "    print(\n",
    "        \"Model {} | Lowest Valid Error: {} at Epoch {}\".format(\n",
    "            model_num, min(losses), epochs\n",
    "        )\n",
    "    )\n",
    "    return (filepaths[losses.index(min(losses))], epochs)\n",
    "\n",
    "\n",
    "def plot_learning_process(hist_list):\n",
    "    for i, hist in enumerate(hist_list):\n",
    "        plt.subplot(1, 1, 1)\n",
    "        plt.plot(hist.history[\"loss\"])\n",
    "        plt.plot(hist.history[\"val_loss\"])\n",
    "        plt.title(\"Model Loss\")\n",
    "        plt.ylabel(\"Loss\")\n",
    "        plt.xlabel(\"Epoch\")\n",
    "        plt.legend([\"Train\", \"Validation\"], loc=\"upper right\")\n",
    "        plt.show()\n",
    "    return None\n",
    "\n",
    "\n",
    "def run_models():\n",
    "    for i in range(1, 4):\n",
    "        print(create_model(i)().summary())\n",
    "        model = KerasRegressor(\n",
    "            build_fn=create_model(i),\n",
    "            epochs=100,\n",
    "            batch_size=32,\n",
    "            validation_split=0.2,\n",
    "            shuffle=True,\n",
    "            verbose=1,\n",
    "        )\n",
    "\n",
    "        filepath = get_modelcheckpoint_path(i)\n",
    "        checkpoint = ModelCheckpoint(\n",
    "            filepath, monitor=\"val_loss\", save_best_only=True, verbose=0, mode=\"min\"\n",
    "        )\n",
    "        callbacks_list = [checkpoint]\n",
    "\n",
    "        if i == 1:\n",
    "            h1 = model.fit(X, y, callbacks=callbacks_list)\n",
    "        elif i == 2:\n",
    "            h2 = model.fit(X, y, callbacks=callbacks_list)\n",
    "        elif i == 3:\n",
    "            h3 = model.fit(X, y, callbacks=callbacks_list)\n",
    "\n",
    "    return [h1, h2, h3]\n",
    "\n",
    "\n",
    "def create_model(model_num):\n",
    "    def model_1():\n",
    "        model = Sequential()\n",
    "        model.add(Dense(X.shape[1], input_dim=X.shape[1], activation=\"relu\"))\n",
    "        model.add(Dense(64, activation=\"relu\"))\n",
    "        model.add(Dense(32, activation=\"relu\"))\n",
    "        model.add(Dense(1))\n",
    "        model.compile(loss=root_mean_squared_error, optimizer=\"adam\")\n",
    "        return model\n",
    "\n",
    "    def model_2():\n",
    "        model = Sequential()\n",
    "        model.add(Dense(X.shape[1], input_dim=X.shape[1], activation=\"relu\"))\n",
    "        model.add(Dense(64, activation=\"relu\"))\n",
    "        model.add(Dense(128, activation=\"relu\"))\n",
    "        model.add(Dense(32, activation=\"relu\"))\n",
    "        model.add(Dropout(0.2))\n",
    "        model.add(Dense(1))\n",
    "        model.compile(loss=root_mean_squared_error, optimizer=\"adam\")\n",
    "        return model\n",
    "\n",
    "    def model_3():\n",
    "        model = Sequential()\n",
    "        model.add(Dense(X.shape[1], input_dim=X.shape[1], activation=\"relu\"))\n",
    "        model.add(Dense(128, activation=\"relu\"))\n",
    "        model.add(Dense(256, activation=\"relu\"))\n",
    "        model.add(Dense(64, activation=\"relu\"))\n",
    "        model.add(Dropout(0.2))\n",
    "        model.add(Dense(1))\n",
    "        model.compile(loss=root_mean_squared_error, optimizer=\"adam\")\n",
    "        return model\n",
    "\n",
    "    if model_num == 1:\n",
    "        return model_1\n",
    "    elif model_num == 2:\n",
    "        return model_2\n",
    "    elif model_num == 3:\n",
    "        return model_3\n",
    "    else:\n",
    "        return \"invalid model_num\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-28T22:46:13.500083Z",
     "start_time": "2019-06-28T22:46:11.225860Z"
    }
   },
   "outputs": [],
   "source": [
    "X, y = utils.load_full_dataset(\"quad\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-28T22:46:28.205365Z",
     "start_time": "2019-06-28T22:46:28.201772Z"
    },
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Takes about ~ 30 mins for training 100 epochs for all models\n",
    "hist_list = run_models()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-29T16:55:52.344495Z",
     "start_time": "2019-05-29T16:55:51.714626Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# The models seem to overfit at an early stage for all models\n",
    "plot_learning_process(hist_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-29T11:39:45.668288Z",
     "start_time": "2019-05-29T11:32:19.437421Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model 1 | Lowest Valid Error: 9.06407 at Epoch 4\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "85719207a527436f95a975bd5770235d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=5), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>MAE</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>train</th>\n",
       "      <td>7.104619</td>\n",
       "      <td>7.093637</td>\n",
       "      <td>7.081982</td>\n",
       "      <td>6.987312</td>\n",
       "      <td>7.041672</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>valid</th>\n",
       "      <td>7.157601</td>\n",
       "      <td>7.155964</td>\n",
       "      <td>7.048330</td>\n",
       "      <td>6.970658</td>\n",
       "      <td>7.026474</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                       \n",
       "MAE                                                    \n",
       "train  7.104619  7.093637  7.081982  6.987312  7.041672\n",
       "valid  7.157601  7.155964  7.048330  6.970658  7.026474"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>RMSE</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>train</th>\n",
       "      <td>9.200789</td>\n",
       "      <td>9.180906</td>\n",
       "      <td>9.312114</td>\n",
       "      <td>9.164618</td>\n",
       "      <td>9.156726</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>valid</th>\n",
       "      <td>9.259978</td>\n",
       "      <td>9.276065</td>\n",
       "      <td>9.303126</td>\n",
       "      <td>9.119281</td>\n",
       "      <td>9.132811</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                       \n",
       "RMSE                                                   \n",
       "train  9.200789  9.180906  9.312114  9.164618  9.156726\n",
       "valid  9.259978  9.276065  9.303126  9.119281  9.132811"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "   <--- Validation Errors --->\n",
      "MAE  | Mean: 7.07181, SD: 0.07387\n",
      "RMSE | Mean: 9.21825, SD: 0.07666\n",
      "\n",
      "Model 2 | Lowest Valid Error: 9.08894 at Epoch 6\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e21b160520594ce893c734387328ded7",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=5), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>MAE</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>train</th>\n",
       "      <td>6.974793</td>\n",
       "      <td>7.075141</td>\n",
       "      <td>7.240280</td>\n",
       "      <td>7.003266</td>\n",
       "      <td>7.018079</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>valid</th>\n",
       "      <td>7.048455</td>\n",
       "      <td>7.134515</td>\n",
       "      <td>7.225957</td>\n",
       "      <td>6.992152</td>\n",
       "      <td>6.990194</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                       \n",
       "MAE                                                    \n",
       "train  6.974793  7.075141  7.240280  7.003266  7.018079\n",
       "valid  7.048455  7.134515  7.225957  6.992152  6.990194"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>RMSE</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>train</th>\n",
       "      <td>9.154584</td>\n",
       "      <td>9.191060</td>\n",
       "      <td>9.324433</td>\n",
       "      <td>9.267034</td>\n",
       "      <td>9.259054</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>valid</th>\n",
       "      <td>9.237600</td>\n",
       "      <td>9.287955</td>\n",
       "      <td>9.316102</td>\n",
       "      <td>9.222004</td>\n",
       "      <td>9.234972</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                       \n",
       "RMSE                                                   \n",
       "train  9.154584  9.191060  9.324433  9.267034  9.259054\n",
       "valid  9.237600  9.287955  9.316102  9.222004  9.234972"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "   <--- Validation Errors --->\n",
      "MAE  | Mean: 7.07825, SD: 0.0906\n",
      "RMSE | Mean: 9.25973, SD: 0.03606\n",
      "\n",
      "Model 3 | Lowest Valid Error: 9.06403 at Epoch 7\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "fcc4ae7eae15461f9c4d58d823f831f1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=5), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>MAE</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>train</th>\n",
       "      <td>6.937657</td>\n",
       "      <td>6.953649</td>\n",
       "      <td>7.044424</td>\n",
       "      <td>7.104365</td>\n",
       "      <td>7.236153</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>valid</th>\n",
       "      <td>6.999789</td>\n",
       "      <td>7.014042</td>\n",
       "      <td>7.038410</td>\n",
       "      <td>7.108518</td>\n",
       "      <td>7.247965</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                       \n",
       "MAE                                                    \n",
       "train  6.937657  6.953649  7.044424  7.104365  7.236153\n",
       "valid  6.999789  7.014042  7.038410  7.108518  7.247965"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>RMSE</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>train</th>\n",
       "      <td>9.115793</td>\n",
       "      <td>9.087052</td>\n",
       "      <td>9.153248</td>\n",
       "      <td>9.234147</td>\n",
       "      <td>9.343228</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>valid</th>\n",
       "      <td>9.192974</td>\n",
       "      <td>9.185803</td>\n",
       "      <td>9.168075</td>\n",
       "      <td>9.221036</td>\n",
       "      <td>9.335152</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                       \n",
       "RMSE                                                   \n",
       "train  9.115793  9.087052  9.153248  9.234147  9.343228\n",
       "valid  9.192974  9.185803  9.168075  9.221036  9.335152"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "   <--- Validation Errors --->\n",
      "MAE  | Mean: 7.08174, SD: 0.09113\n",
      "RMSE | Mean: 9.22061, SD: 0.05976\n",
      "\n"
     ]
    }
   ],
   "source": [
    "for i in range(1, 4):\n",
    "    weights_path, epochs = get_weights_path_and_epoch(i)\n",
    "    model = KerasRegressor(\n",
    "        build_fn=create_model(i),\n",
    "        epochs=epochs,\n",
    "        batch_size=32,\n",
    "        validation_split=0.2,\n",
    "        shuffle=True,\n",
    "        verbose=0,\n",
    "    )\n",
    "\n",
    "    errors = utils.cross_val(model, X, y, n_folds=5)\n",
    "    utils.summarize_errors(errors, verbose=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-29T16:55:22.726002Z",
     "start_time": "2019-05-29T16:55:22.721544Z"
    }
   },
   "source": [
    "### Neural Networks (Model 1) trained to Epoch 28\n",
    " - MAE  | Mean: 7.07181, SD: 0.07387\n",
    " - RMSE | Mean: 9.21825, SD: 0.07666"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.1"
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}