(Keras) Cornell Movie Review Dataset.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# (Keras) Cornell Movie Review Dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Based on https://github.com/yaringal/BayesianRNN/blob/master/Example/sentiment_lstm_regression.py"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from tensorflow.contrib.keras.python.keras.optimizers import SGD, RMSprop, Adagrad\n",
    "from tensorflow.contrib.keras.python.keras.models import Sequential\n",
    "from tensorflow.contrib.keras.python.keras.layers.core import Dense, Dropout\n",
    "from tensorflow.contrib.keras.python.keras.layers.embeddings import Embedding\n",
    "from tensorflow.contrib.keras.python.keras.layers.recurrent import LSTM, GRU, SimpleRNN\n",
    "from tensorflow.contrib.keras.python.keras.regularizers import l2\n",
    "from tensorflow.contrib.keras.python.keras.optimizers import Adam\n",
    "from sklearn.metrics import mean_squared_error\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from yaringal_callbacks import ModelTest\n",
    "from yaringal_dataset import loader\n",
    "\n",
    "%matplotlib inline\n",
    "plt.style.use('fivethirtyeight')\n",
    "plt.rcParams[\"figure.figsize\"] = (8, 5)\n",
    "\n",
    "# Global params:\n",
    "NB_WORDS = 20000\n",
    "SKIP_TOP = 0\n",
    "TEST_SPLIT = 0.2\n",
    "INIT_SEED = 2017\n",
    "GLOBAL_SEED = 2018\n",
    "MAXLEN = 200\n",
    "BATCH_SIZE = 128\n",
    "TEST_BATCH_SIZE = 512"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10620 train sequences\n",
      "2655 test sequences\n",
      "Pad sequences (samples x time)\n",
      "X_train shape: (10620, 200)\n",
      "X_test shape: (2655, 200)\n"
     ]
    }
   ],
   "source": [
    "dataset = loader(INIT_SEED, MAXLEN, NB_WORDS, SKIP_TOP, TEST_SPLIT)\n",
    "\n",
    "X_train, X_test, Y_train, Y_test = dataset.X_train, dataset.X_test, dataset.Y_train, dataset.Y_test\n",
    "mean_y_train, std_y_train = dataset.mean_y_train, dataset.std_y_train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def get_model(idrop=0.2, edrop=0.1, odrop=0.25, rdrop=0.2, weight_decay=1e-4, lr=1e-3):\n",
    "    model = Sequential()\n",
    "    model.add(Embedding(NB_WORDS, 128, embeddings_regularizer=l2(weight_decay),\n",
    "                        input_length=MAXLEN)) \n",
    "    if edrop:\n",
    "        model.add(Dropout(edrop))\n",
    "    model.add(LSTM(128, kernel_regularizer=l2(weight_decay), recurrent_regularizer=l2(weight_decay),\n",
    "                   bias_regularizer=l2(weight_decay), dropout=idrop, recurrent_dropout=rdrop))\n",
    "    if odrop:\n",
    "        model.add(Dropout(odrop))\n",
    "    model.add(Dense(1, kernel_regularizer=l2(weight_decay),\n",
    "                    bias_regularizer=l2(weight_decay)))\n",
    "    optimizer = Adam(lr)\n",
    "    model.compile(loss='mse', metrics=[\"mse\"], optimizer=optimizer)\n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Normal Variational LSTM (w/o Embedding Dropout)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Build model...\n"
     ]
    }
   ],
   "source": [
    "print('Build model...')\n",
    "model = get_model(rdrop=0.25, odrop=0.25, edrop=0, idrop=0.25, weight_decay=1e-4, lr=1e-3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "modeltest_1 = ModelTest(X_test, Yt=Y_test,\n",
    "                        test_every_X_epochs=2, verbose=0, T=10,\n",
    "                        mean_y_train=mean_y_train, std_y_train=std_y_train,\n",
    "                        loss='euclidean', batch_size=TEST_BATCH_SIZE)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/200\n",
      "Standard error at epoch 000: 1.0040/0.1857\n",
      "MC error at epoch 000: 1.0064/0.1859\n",
      "15s - loss: 76.7824 - mean_squared_error: 1.0282\n",
      "Epoch 2/200\n",
      "11s - loss: 59.7187 - mean_squared_error: 1.0073\n",
      "Epoch 3/200\n",
      "Standard error at epoch 002: 1.0038/0.1856\n",
      "MC error at epoch 002: 1.0048/0.1857\n",
      "14s - loss: 46.2964 - mean_squared_error: 1.0051\n",
      "Epoch 4/200\n",
      "11s - loss: 35.7194 - mean_squared_error: 1.0037\n",
      "Epoch 5/200\n",
      "Standard error at epoch 004: 1.0056/0.1858\n",
      "MC error at epoch 004: 1.0067/0.1859\n",
      "14s - loss: 27.4081 - mean_squared_error: 0.9986\n",
      "Epoch 6/200\n",
      "11s - loss: 20.9169 - mean_squared_error: 0.9945\n",
      "Epoch 7/200\n",
      "Standard error at epoch 006: 1.0006/0.1853\n",
      "MC error at epoch 006: 1.0023/0.1855\n",
      "14s - loss: 15.8839 - mean_squared_error: 0.9916\n",
      "Epoch 8/200\n",
      "11s - loss: 12.0078 - mean_squared_error: 0.9824\n",
      "Epoch 9/200\n",
      "Standard error at epoch 008: 0.9954/0.1849\n",
      "MC error at epoch 008: 0.9967/0.1850\n",
      "14s - loss: 9.0588 - mean_squared_error: 0.9768\n",
      "Epoch 10/200\n",
      "11s - loss: 6.8289 - mean_squared_error: 0.9618\n",
      "Epoch 11/200\n",
      "Standard error at epoch 010: 0.9966/0.1850\n",
      "MC error at epoch 010: 0.9980/0.1851\n",
      "14s - loss: 5.1613 - mean_squared_error: 0.9398\n",
      "Epoch 12/200\n",
      "11s - loss: 3.9136 - mean_squared_error: 0.8944\n",
      "Epoch 13/200\n",
      "Standard error at epoch 012: 0.9373/0.1794\n",
      "MC error at epoch 012: 0.9371/0.1794\n",
      "14s - loss: 2.9606 - mean_squared_error: 0.7966\n",
      "Epoch 14/200\n",
      "11s - loss: 2.2306 - mean_squared_error: 0.6571\n",
      "Epoch 15/200\n",
      "Standard error at epoch 014: 0.8171/0.1675\n",
      "MC error at epoch 014: 0.8175/0.1675\n",
      "14s - loss: 1.6629 - mean_squared_error: 0.4910\n",
      "Epoch 16/200\n",
      "11s - loss: 1.2618 - mean_squared_error: 0.3653\n",
      "Epoch 17/200\n",
      "Standard error at epoch 016: 0.7679/0.1624\n",
      "MC error at epoch 016: 0.7718/0.1628\n",
      "14s - loss: 0.9780 - mean_squared_error: 0.2780\n",
      "Epoch 18/200\n",
      "11s - loss: 0.7959 - mean_squared_error: 0.2360\n",
      "Epoch 19/200\n",
      "Standard error at epoch 018: 0.7650/0.1621\n",
      "MC error at epoch 018: 0.7681/0.1624\n",
      "14s - loss: 0.6602 - mean_squared_error: 0.1992\n",
      "Epoch 20/200\n",
      "11s - loss: 0.5636 - mean_squared_error: 0.1762\n",
      "Epoch 21/200\n",
      "Standard error at epoch 020: 0.7752/0.1631\n",
      "MC error at epoch 020: 0.7806/0.1637\n",
      "14s - loss: 0.5129 - mean_squared_error: 0.1772\n",
      "Epoch 22/200\n",
      "11s - loss: 0.4572 - mean_squared_error: 0.1590\n",
      "Epoch 23/200\n",
      "Standard error at epoch 022: 0.7650/0.1621\n",
      "MC error at epoch 022: 0.7672/0.1623\n",
      "14s - loss: 0.4238 - mean_squared_error: 0.1550\n",
      "Epoch 24/200\n",
      "11s - loss: 0.4020 - mean_squared_error: 0.1553\n",
      "Epoch 25/200\n",
      "Standard error at epoch 024: 0.7614/0.1617\n",
      "MC error at epoch 024: 0.7636/0.1619\n",
      "14s - loss: 0.3724 - mean_squared_error: 0.1427\n",
      "Epoch 26/200\n",
      "11s - loss: 0.3566 - mean_squared_error: 0.1416\n",
      "Epoch 27/200\n",
      "Standard error at epoch 026: 0.7664/0.1622\n",
      "MC error at epoch 026: 0.7664/0.1622\n",
      "14s - loss: 0.3436 - mean_squared_error: 0.1403\n",
      "Epoch 28/200\n",
      "11s - loss: 0.3322 - mean_squared_error: 0.1383\n",
      "Epoch 29/200\n",
      "Standard error at epoch 028: 0.7633/0.1619\n",
      "MC error at epoch 028: 0.7632/0.1619\n",
      "14s - loss: 0.3170 - mean_squared_error: 0.1318\n",
      "Epoch 30/200\n",
      "11s - loss: 0.3115 - mean_squared_error: 0.1335\n",
      "Epoch 31/200\n",
      "Standard error at epoch 030: 0.7575/0.1613\n",
      "MC error at epoch 030: 0.7599/0.1615\n",
      "14s - loss: 0.2999 - mean_squared_error: 0.1279\n",
      "Epoch 32/200\n",
      "11s - loss: 0.2974 - mean_squared_error: 0.1306\n",
      "Epoch 33/200\n",
      "Standard error at epoch 032: 0.7608/0.1616\n",
      "MC error at epoch 032: 0.7587/0.1614\n",
      "14s - loss: 0.2897 - mean_squared_error: 0.1277\n",
      "Epoch 34/200\n",
      "11s - loss: 0.2796 - mean_squared_error: 0.1228\n",
      "Epoch 35/200\n",
      "Standard error at epoch 034: 0.7544/0.1609\n",
      "MC error at epoch 034: 0.7544/0.1609\n",
      "14s - loss: 0.2731 - mean_squared_error: 0.1209\n",
      "Epoch 36/200\n",
      "11s - loss: 0.2658 - mean_squared_error: 0.1179\n",
      "Epoch 37/200\n",
      "Standard error at epoch 036: 0.7467/0.1601\n",
      "MC error at epoch 036: 0.7498/0.1604\n",
      "14s - loss: 0.2649 - mean_squared_error: 0.1208\n",
      "Epoch 38/200\n",
      "11s - loss: 0.2621 - mean_squared_error: 0.1207\n",
      "Epoch 39/200\n",
      "Standard error at epoch 038: 0.7516/0.1606\n",
      "MC error at epoch 038: 0.7523/0.1607\n",
      "14s - loss: 0.2569 - mean_squared_error: 0.1184\n",
      "Epoch 40/200\n",
      "11s - loss: 0.2526 - mean_squared_error: 0.1169\n",
      "Epoch 41/200\n",
      "Standard error at epoch 040: 0.7452/0.1599\n",
      "MC error at epoch 040: 0.7473/0.1602\n",
      "14s - loss: 0.2458 - mean_squared_error: 0.1127\n",
      "Epoch 42/200\n",
      "11s - loss: 0.2444 - mean_squared_error: 0.1139\n",
      "Epoch 43/200\n",
      "Standard error at epoch 042: 0.7244/0.1577\n",
      "MC error at epoch 042: 0.7249/0.1578\n",
      "14s - loss: 0.2440 - mean_squared_error: 0.1153\n",
      "Epoch 44/200\n",
      "11s - loss: 0.2399 - mean_squared_error: 0.1129\n",
      "Epoch 45/200\n",
      "Standard error at epoch 044: 0.7318/0.1585\n",
      "MC error at epoch 044: 0.7308/0.1584\n",
      "14s - loss: 0.2336 - mean_squared_error: 0.1089\n",
      "Epoch 46/200\n",
      "11s - loss: 0.2316 - mean_squared_error: 0.1091\n",
      "Epoch 47/200\n",
      "Standard error at epoch 046: 0.7383/0.1592\n",
      "MC error at epoch 046: 0.7323/0.1586\n",
      "14s - loss: 0.2287 - mean_squared_error: 0.1080\n",
      "Epoch 48/200\n",
      "11s - loss: 0.2288 - mean_squared_error: 0.1094\n",
      "Epoch 49/200\n",
      "Standard error at epoch 048: 0.7270/0.1580\n",
      "MC error at epoch 048: 0.7240/0.1577\n",
      "14s - loss: 0.2266 - mean_squared_error: 0.1083\n",
      "Epoch 50/200\n",
      "11s - loss: 0.2269 - mean_squared_error: 0.1101\n",
      "Epoch 51/200\n",
      "Standard error at epoch 050: 0.7356/0.1589\n",
      "MC error at epoch 050: 0.7308/0.1584\n",
      "14s - loss: 0.2216 - mean_squared_error: 0.1062\n",
      "Epoch 52/200\n",
      "11s - loss: 0.2190 - mean_squared_error: 0.1052\n",
      "Epoch 53/200\n",
      "Standard error at epoch 052: 0.7219/0.1574\n",
      "MC error at epoch 052: 0.7198/0.1572\n",
      "14s - loss: 0.2192 - mean_squared_error: 0.1066\n",
      "Epoch 54/200\n",
      "11s - loss: 0.2135 - mean_squared_error: 0.1020\n",
      "Epoch 55/200\n",
      "Standard error at epoch 054: 0.7357/0.1589\n",
      "MC error at epoch 054: 0.7318/0.1585\n",
      "14s - loss: 0.2158 - mean_squared_error: 0.1053\n",
      "Epoch 56/200\n",
      "11s - loss: 0.2149 - mean_squared_error: 0.1051\n",
      "Epoch 57/200\n",
      "Standard error at epoch 056: 0.7158/0.1568\n",
      "MC error at epoch 056: 0.7152/0.1567\n",
      "14s - loss: 0.2094 - mean_squared_error: 0.1005\n",
      "Epoch 58/200\n",
      "11s - loss: 0.2132 - mean_squared_error: 0.1048\n",
      "Epoch 59/200\n",
      "Standard error at epoch 058: 0.7233/0.1576\n",
      "MC error at epoch 058: 0.7201/0.1572\n",
      "14s - loss: 0.2105 - mean_squared_error: 0.1028\n",
      "Epoch 60/200\n",
      "11s - loss: 0.2099 - mean_squared_error: 0.1032\n",
      "Epoch 61/200\n",
      "Standard error at epoch 060: 0.6936/0.1543\n",
      "MC error at epoch 060: 0.6927/0.1542\n",
      "14s - loss: 0.2075 - mean_squared_error: 0.1015\n",
      "Epoch 62/200\n",
      "11s - loss: 0.2064 - mean_squared_error: 0.1012\n",
      "Epoch 63/200\n",
      "Standard error at epoch 062: 0.7014/0.1552\n",
      "MC error at epoch 062: 0.6964/0.1546\n",
      "14s - loss: 0.2070 - mean_squared_error: 0.1026\n",
      "Epoch 64/200\n",
      "11s - loss: 0.2058 - mean_squared_error: 0.1018\n",
      "Epoch 65/200\n",
      "Standard error at epoch 064: 0.7164/0.1568\n",
      "MC error at epoch 064: 0.7124/0.1564\n",
      "14s - loss: 0.2016 - mean_squared_error: 0.0983\n",
      "Epoch 66/200\n",
      "11s - loss: 0.2002 - mean_squared_error: 0.0979\n",
      "Epoch 67/200\n",
      "Standard error at epoch 066: 0.6998/0.1550\n",
      "MC error at epoch 066: 0.7006/0.1551\n",
      "14s - loss: 0.2041 - mean_squared_error: 0.1023\n",
      "Epoch 68/200\n",
      "11s - loss: 0.2013 - mean_squared_error: 0.0997\n",
      "Epoch 69/200\n",
      "Standard error at epoch 068: 0.7060/0.1557\n",
      "MC error at epoch 068: 0.7053/0.1556\n",
      "14s - loss: 0.1980 - mean_squared_error: 0.0968\n",
      "Epoch 70/200\n",
      "11s - loss: 0.2013 - mean_squared_error: 0.1007\n",
      "Epoch 71/200\n",
      "Standard error at epoch 070: 0.7039/0.1554\n",
      "MC error at epoch 070: 0.7002/0.1550\n",
      "14s - loss: 0.1986 - mean_squared_error: 0.0983\n",
      "Epoch 72/200\n",
      "11s - loss: 0.1989 - mean_squared_error: 0.0992\n",
      "Epoch 73/200\n",
      "Standard error at epoch 072: 0.7183/0.1570\n",
      "MC error at epoch 072: 0.7159/0.1568\n",
      "14s - loss: 0.1944 - mean_squared_error: 0.0951\n",
      "Epoch 74/200\n",
      "11s - loss: 0.1966 - mean_squared_error: 0.0982\n",
      "Epoch 75/200\n",
      "Standard error at epoch 074: 0.7209/0.1573\n",
      "MC error at epoch 074: 0.7177/0.1570\n",
      "14s - loss: 0.1966 - mean_squared_error: 0.0986\n",
      "Epoch 76/200\n",
      "11s - loss: 0.1955 - mean_squared_error: 0.0973\n",
      "Epoch 77/200\n",
      "Standard error at epoch 076: 0.6904/0.1540\n",
      "MC error at epoch 076: 0.6893/0.1538\n",
      "14s - loss: 0.1938 - mean_squared_error: 0.0961\n",
      "Epoch 78/200\n",
      "11s - loss: 0.1935 - mean_squared_error: 0.0963\n",
      "Epoch 79/200\n",
      "Standard error at epoch 078: 0.7109/0.1562\n",
      "MC error at epoch 078: 0.7077/0.1559\n",
      "14s - loss: 0.1916 - mean_squared_error: 0.0946\n",
      "Epoch 80/200\n",
      "11s - loss: 0.1900 - mean_squared_error: 0.0938\n",
      "Epoch 81/200\n",
      "Standard error at epoch 080: 0.7019/0.1552\n",
      "MC error at epoch 080: 0.7020/0.1552\n",
      "14s - loss: 0.1947 - mean_squared_error: 0.0987\n",
      "Epoch 82/200\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "11s - loss: 0.1898 - mean_squared_error: 0.0938\n",
      "Epoch 83/200\n",
      "Standard error at epoch 082: 0.7009/0.1551\n",
      "MC error at epoch 082: 0.6929/0.1542\n",
      "14s - loss: 0.1904 - mean_squared_error: 0.0951\n",
      "Epoch 84/200\n",
      "11s - loss: 0.1919 - mean_squared_error: 0.0965\n",
      "Epoch 85/200\n",
      "Standard error at epoch 084: 0.6783/0.1526\n",
      "MC error at epoch 084: 0.6797/0.1528\n",
      "14s - loss: 0.1936 - mean_squared_error: 0.0979\n",
      "Epoch 86/200\n",
      "11s - loss: 0.1936 - mean_squared_error: 0.0976\n",
      "Epoch 87/200\n",
      "Standard error at epoch 086: 0.6898/0.1539\n",
      "MC error at epoch 086: 0.6877/0.1536\n",
      "14s - loss: 0.1869 - mean_squared_error: 0.0915\n",
      "Epoch 88/200\n",
      "11s - loss: 0.1889 - mean_squared_error: 0.0944\n",
      "Epoch 89/200\n",
      "Standard error at epoch 088: 0.6833/0.1532\n",
      "MC error at epoch 088: 0.6797/0.1528\n",
      "14s - loss: 0.1911 - mean_squared_error: 0.0964\n",
      "Epoch 90/200\n",
      "11s - loss: 0.1918 - mean_squared_error: 0.0971\n",
      "Epoch 91/200\n",
      "Standard error at epoch 090: 0.6814/0.1530\n",
      "MC error at epoch 090: 0.6768/0.1524\n",
      "14s - loss: 0.1886 - mean_squared_error: 0.0944\n",
      "Epoch 92/200\n",
      "11s - loss: 0.1841 - mean_squared_error: 0.0907\n",
      "Epoch 93/200\n",
      "Standard error at epoch 092: 0.6995/0.1550\n",
      "MC error at epoch 092: 0.6944/0.1544\n",
      "14s - loss: 0.1863 - mean_squared_error: 0.0933\n",
      "Epoch 94/200\n",
      "11s - loss: 0.1892 - mean_squared_error: 0.0954\n",
      "Epoch 95/200\n",
      "Standard error at epoch 094: 0.6814/0.1530\n",
      "MC error at epoch 094: 0.6775/0.1525\n",
      "14s - loss: 0.1856 - mean_squared_error: 0.0922\n",
      "Epoch 96/200\n",
      "11s - loss: 0.1816 - mean_squared_error: 0.0888\n",
      "Epoch 97/200\n",
      "Standard error at epoch 096: 0.6783/0.1526\n",
      "MC error at epoch 096: 0.6739/0.1521\n",
      "14s - loss: 0.1816 - mean_squared_error: 0.0897\n",
      "Epoch 98/200\n",
      "11s - loss: 0.1864 - mean_squared_error: 0.0947\n",
      "Epoch 99/200\n",
      "Standard error at epoch 098: 0.6780/0.1526\n",
      "MC error at epoch 098: 0.6731/0.1520\n",
      "14s - loss: 0.1844 - mean_squared_error: 0.0925\n",
      "Epoch 100/200\n",
      "11s - loss: 0.1835 - mean_squared_error: 0.0913\n",
      "Epoch 101/200\n",
      "Standard error at epoch 100: 0.6867/0.1535\n",
      "MC error at epoch 100: 0.6809/0.1529\n",
      "14s - loss: 0.1841 - mean_squared_error: 0.0920\n",
      "Epoch 102/200\n",
      "11s - loss: 0.1850 - mean_squared_error: 0.0931\n",
      "Epoch 103/200\n",
      "Standard error at epoch 102: 0.6644/0.1510\n",
      "MC error at epoch 102: 0.6586/0.1504\n",
      "14s - loss: 0.1815 - mean_squared_error: 0.0905\n",
      "Epoch 104/200\n",
      "11s - loss: 0.1810 - mean_squared_error: 0.0902\n",
      "Epoch 105/200\n",
      "Standard error at epoch 104: 0.7013/0.1552\n",
      "MC error at epoch 104: 0.6944/0.1544\n",
      "14s - loss: 0.1821 - mean_squared_error: 0.0914\n",
      "Epoch 106/200\n",
      "11s - loss: 0.1834 - mean_squared_error: 0.0924\n",
      "Epoch 107/200\n",
      "Standard error at epoch 106: 0.6902/0.1539\n",
      "MC error at epoch 106: 0.6861/0.1535\n",
      "14s - loss: 0.1834 - mean_squared_error: 0.0924\n",
      "Epoch 108/200\n",
      "11s - loss: 0.1813 - mean_squared_error: 0.0905\n",
      "Epoch 109/200\n",
      "Standard error at epoch 108: 0.6770/0.1525\n",
      "MC error at epoch 108: 0.6753/0.1523\n",
      "14s - loss: 0.1813 - mean_squared_error: 0.0904\n",
      "Epoch 110/200\n",
      "11s - loss: 0.1782 - mean_squared_error: 0.0879\n",
      "Epoch 111/200\n",
      "Standard error at epoch 110: 0.6706/0.1517\n",
      "MC error at epoch 110: 0.6672/0.1513\n",
      "14s - loss: 0.1817 - mean_squared_error: 0.0916\n",
      "Epoch 112/200\n",
      "12s - loss: 0.1775 - mean_squared_error: 0.0878\n",
      "Epoch 113/200\n",
      "Standard error at epoch 112: 0.6965/0.1546\n",
      "MC error at epoch 112: 0.6916/0.1541\n",
      "14s - loss: 0.1832 - mean_squared_error: 0.0933\n",
      "Epoch 114/200\n",
      "11s - loss: 0.1802 - mean_squared_error: 0.0901\n",
      "Epoch 115/200\n",
      "Standard error at epoch 114: 0.6724/0.1519\n",
      "MC error at epoch 114: 0.6691/0.1516\n",
      "14s - loss: 0.1786 - mean_squared_error: 0.0889\n",
      "Epoch 116/200\n",
      "11s - loss: 0.1785 - mean_squared_error: 0.0891\n",
      "Epoch 117/200\n",
      "Standard error at epoch 116: 0.6829/0.1531\n",
      "MC error at epoch 116: 0.6790/0.1527\n",
      "14s - loss: 0.1789 - mean_squared_error: 0.0894\n",
      "Epoch 118/200\n",
      "11s - loss: 0.1758 - mean_squared_error: 0.0867\n",
      "Epoch 119/200\n",
      "Standard error at epoch 118: 0.6807/0.1529\n",
      "MC error at epoch 118: 0.6810/0.1529\n",
      "14s - loss: 0.1752 - mean_squared_error: 0.0866\n",
      "Epoch 120/200\n",
      "11s - loss: 0.1760 - mean_squared_error: 0.0876\n",
      "Epoch 121/200\n",
      "Standard error at epoch 120: 0.6950/0.1545\n",
      "MC error at epoch 120: 0.6894/0.1538\n",
      "14s - loss: 0.1772 - mean_squared_error: 0.0887\n",
      "Epoch 122/200\n",
      "11s - loss: 0.1772 - mean_squared_error: 0.0885\n",
      "Epoch 123/200\n",
      "Standard error at epoch 122: 0.6830/0.1531\n",
      "MC error at epoch 122: 0.6763/0.1524\n",
      "14s - loss: 0.1756 - mean_squared_error: 0.0873\n",
      "Epoch 124/200\n",
      "11s - loss: 0.1781 - mean_squared_error: 0.0896\n",
      "Epoch 125/200\n",
      "Standard error at epoch 124: 0.6901/0.1539\n",
      "MC error at epoch 124: 0.6836/0.1532\n",
      "14s - loss: 0.1779 - mean_squared_error: 0.0892\n",
      "Epoch 126/200\n",
      "11s - loss: 0.1768 - mean_squared_error: 0.0886\n",
      "Epoch 127/200\n",
      "Standard error at epoch 126: 0.6852/0.1534\n",
      "MC error at epoch 126: 0.6785/0.1526\n",
      "14s - loss: 0.1775 - mean_squared_error: 0.0890\n",
      "Epoch 128/200\n",
      "11s - loss: 0.1753 - mean_squared_error: 0.0869\n",
      "Epoch 129/200\n",
      "Standard error at epoch 128: 0.6823/0.1531\n",
      "MC error at epoch 128: 0.6742/0.1521\n",
      "14s - loss: 0.1755 - mean_squared_error: 0.0872\n",
      "Epoch 130/200\n",
      "11s - loss: 0.1777 - mean_squared_error: 0.0889\n",
      "Epoch 131/200\n",
      "Standard error at epoch 130: 0.6950/0.1545\n",
      "MC error at epoch 130: 0.6822/0.1530\n",
      "14s - loss: 0.1742 - mean_squared_error: 0.0858\n",
      "Epoch 132/200\n",
      "11s - loss: 0.1728 - mean_squared_error: 0.0851\n",
      "Epoch 133/200\n",
      "Standard error at epoch 132: 0.6894/0.1538\n",
      "MC error at epoch 132: 0.6835/0.1532\n",
      "14s - loss: 0.1720 - mean_squared_error: 0.0846\n",
      "Epoch 134/200\n",
      "11s - loss: 0.1761 - mean_squared_error: 0.0886\n",
      "Epoch 135/200\n",
      "Standard error at epoch 134: 0.6789/0.1527\n",
      "MC error at epoch 134: 0.6767/0.1524\n",
      "14s - loss: 0.1745 - mean_squared_error: 0.0864\n",
      "Epoch 136/200\n",
      "11s - loss: 0.1745 - mean_squared_error: 0.0863\n",
      "Epoch 137/200\n",
      "Standard error at epoch 136: 0.6742/0.1521\n",
      "MC error at epoch 136: 0.6715/0.1518\n",
      "14s - loss: 0.1727 - mean_squared_error: 0.0850\n",
      "Epoch 138/200\n",
      "11s - loss: 0.1715 - mean_squared_error: 0.0845\n",
      "Epoch 139/200\n",
      "Standard error at epoch 138: 0.6711/0.1518\n",
      "MC error at epoch 138: 0.6651/0.1511\n",
      "14s - loss: 0.1704 - mean_squared_error: 0.0835\n",
      "Epoch 140/200\n",
      "11s - loss: 0.1716 - mean_squared_error: 0.0851\n",
      "Epoch 141/200\n",
      "Standard error at epoch 140: 0.7014/0.1552\n",
      "MC error at epoch 140: 0.6885/0.1537\n",
      "14s - loss: 0.1736 - mean_squared_error: 0.0870\n",
      "Epoch 142/200\n",
      "11s - loss: 0.1740 - mean_squared_error: 0.0872\n",
      "Epoch 143/200\n",
      "Standard error at epoch 142: 0.6694/0.1516\n",
      "MC error at epoch 142: 0.6650/0.1511\n",
      "14s - loss: 0.1717 - mean_squared_error: 0.0847\n",
      "Epoch 144/200\n",
      "11s - loss: 0.1750 - mean_squared_error: 0.0878\n",
      "Epoch 145/200\n",
      "Standard error at epoch 144: 0.6935/0.1543\n",
      "MC error at epoch 144: 0.6846/0.1533\n",
      "14s - loss: 0.1769 - mean_squared_error: 0.0893\n",
      "Epoch 146/200\n",
      "11s - loss: 0.1768 - mean_squared_error: 0.0883\n",
      "Epoch 147/200\n",
      "Standard error at epoch 146: 0.6734/0.1520\n",
      "MC error at epoch 146: 0.6674/0.1514\n",
      "14s - loss: 0.1749 - mean_squared_error: 0.0861\n",
      "Epoch 148/200\n",
      "11s - loss: 0.2010 - mean_squared_error: 0.1108\n",
      "Epoch 149/200\n",
      "Standard error at epoch 148: 0.6765/0.1524\n",
      "MC error at epoch 148: 0.6724/0.1519\n",
      "14s - loss: 0.1975 - mean_squared_error: 0.1006\n",
      "Epoch 150/200\n",
      "11s - loss: 0.1806 - mean_squared_error: 0.0861\n",
      "Epoch 151/200\n",
      "Standard error at epoch 150: 0.6717/0.1519\n",
      "MC error at epoch 150: 0.6625/0.1508\n",
      "14s - loss: 0.1718 - mean_squared_error: 0.0802\n",
      "Epoch 152/200\n",
      "11s - loss: 0.1721 - mean_squared_error: 0.0824\n",
      "Epoch 153/200\n",
      "Standard error at epoch 152: 0.6819/0.1530\n",
      "MC error at epoch 152: 0.6709/0.1518\n",
      "14s - loss: 0.1695 - mean_squared_error: 0.0808\n",
      "Epoch 154/200\n",
      "11s - loss: 0.1702 - mean_squared_error: 0.0822\n",
      "Epoch 155/200\n",
      "Standard error at epoch 154: 0.6916/0.1541\n",
      "MC error at epoch 154: 0.6833/0.1532\n",
      "14s - loss: 0.1699 - mean_squared_error: 0.0822\n",
      "Epoch 156/200\n",
      "11s - loss: 0.1687 - mean_squared_error: 0.0817\n",
      "Epoch 157/200\n",
      "Standard error at epoch 156: 0.6867/0.1535\n",
      "MC error at epoch 156: 0.6773/0.1525\n",
      "14s - loss: 0.1664 - mean_squared_error: 0.0804\n",
      "Epoch 158/200\n",
      "11s - loss: 0.1672 - mean_squared_error: 0.0816\n",
      "Epoch 159/200\n",
      "Standard error at epoch 158: 0.6687/0.1515\n",
      "MC error at epoch 158: 0.6631/0.1509\n",
      "14s - loss: 0.1684 - mean_squared_error: 0.0830\n",
      "Epoch 160/200\n",
      "11s - loss: 0.1694 - mean_squared_error: 0.0835\n",
      "Epoch 161/200\n",
      "Standard error at epoch 160: 0.6810/0.1529\n",
      "MC error at epoch 160: 0.6724/0.1519\n",
      "14s - loss: 0.1751 - mean_squared_error: 0.0882\n",
      "Epoch 162/200\n",
      "11s - loss: 0.1749 - mean_squared_error: 0.0862\n",
      "Epoch 163/200\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Standard error at epoch 162: 0.6687/0.1515\n",
      "MC error at epoch 162: 0.6637/0.1509\n",
      "14s - loss: 0.1706 - mean_squared_error: 0.0820\n",
      "Epoch 164/200\n",
      "11s - loss: 0.1682 - mean_squared_error: 0.0807\n",
      "Epoch 165/200\n",
      "Standard error at epoch 164: 0.6683/0.1515\n",
      "MC error at epoch 164: 0.6635/0.1509\n",
      "14s - loss: 0.1721 - mean_squared_error: 0.0849\n",
      "Epoch 166/200\n",
      "11s - loss: 0.1687 - mean_squared_error: 0.0809\n",
      "Epoch 167/200\n",
      "Standard error at epoch 166: 0.6711/0.1518\n",
      "MC error at epoch 166: 0.6676/0.1514\n",
      "14s - loss: 0.1679 - mean_squared_error: 0.0809\n",
      "Epoch 168/200\n",
      "11s - loss: 0.1668 - mean_squared_error: 0.0808\n",
      "Epoch 169/200\n",
      "Standard error at epoch 168: 0.6823/0.1530\n",
      "MC error at epoch 168: 0.6764/0.1524\n",
      "14s - loss: 0.1667 - mean_squared_error: 0.0807\n",
      "Epoch 170/200\n",
      "11s - loss: 0.1837 - mean_squared_error: 0.0961\n",
      "Epoch 171/200\n",
      "Standard error at epoch 170: 0.6656/0.1512\n",
      "MC error at epoch 170: 0.6634/0.1509\n",
      "14s - loss: 0.1822 - mean_squared_error: 0.0907\n",
      "Epoch 172/200\n",
      "11s - loss: 0.1752 - mean_squared_error: 0.0846\n",
      "Epoch 173/200\n",
      "Standard error at epoch 172: 0.6863/0.1535\n",
      "MC error at epoch 172: 0.6770/0.1525\n",
      "14s - loss: 0.1711 - mean_squared_error: 0.0815\n",
      "Epoch 174/200\n",
      "11s - loss: 0.1670 - mean_squared_error: 0.0790\n",
      "Epoch 175/200\n",
      "Standard error at epoch 174: 0.6753/0.1523\n",
      "MC error at epoch 174: 0.6685/0.1515\n",
      "14s - loss: 0.1660 - mean_squared_error: 0.0791\n",
      "Epoch 176/200\n",
      "11s - loss: 0.1664 - mean_squared_error: 0.0802\n",
      "Epoch 177/200\n",
      "Standard error at epoch 176: 0.6744/0.1522\n",
      "MC error at epoch 176: 0.6664/0.1513\n",
      "14s - loss: 0.1712 - mean_squared_error: 0.0848\n",
      "Epoch 178/200\n",
      "11s - loss: 0.1726 - mean_squared_error: 0.0849\n",
      "Epoch 179/200\n",
      "Standard error at epoch 178: 0.6639/0.1510\n",
      "MC error at epoch 178: 0.6584/0.1503\n",
      "14s - loss: 0.1711 - mean_squared_error: 0.0833\n",
      "Epoch 180/200\n",
      "11s - loss: 0.1671 - mean_squared_error: 0.0801\n",
      "Epoch 181/200\n",
      "Standard error at epoch 180: 0.6841/0.1532\n",
      "MC error at epoch 180: 0.6758/0.1523\n",
      "14s - loss: 0.1670 - mean_squared_error: 0.0807\n",
      "Epoch 182/200\n",
      "11s - loss: 0.1625 - mean_squared_error: 0.0767\n",
      "Epoch 183/200\n",
      "Standard error at epoch 182: 0.6690/0.1516\n",
      "MC error at epoch 182: 0.6644/0.1510\n",
      "14s - loss: 0.1633 - mean_squared_error: 0.0785\n",
      "Epoch 184/200\n",
      "11s - loss: 0.1669 - mean_squared_error: 0.0822\n",
      "Epoch 185/200\n",
      "Standard error at epoch 184: 0.6708/0.1517\n",
      "MC error at epoch 184: 0.6643/0.1510\n",
      "14s - loss: 0.1689 - mean_squared_error: 0.0836\n",
      "Epoch 186/200\n",
      "11s - loss: 0.1676 - mean_squared_error: 0.0816\n",
      "Epoch 187/200\n",
      "Standard error at epoch 186: 0.6670/0.1513\n",
      "MC error at epoch 186: 0.6636/0.1509\n",
      "14s - loss: 0.1705 - mean_squared_error: 0.0838\n",
      "Epoch 188/200\n",
      "11s - loss: 0.1717 - mean_squared_error: 0.0848\n",
      "Epoch 189/200\n",
      "Standard error at epoch 188: 0.6811/0.1529\n",
      "MC error at epoch 188: 0.6754/0.1523\n",
      "14s - loss: 0.1710 - mean_squared_error: 0.0834\n",
      "Epoch 190/200\n",
      "11s - loss: 0.1674 - mean_squared_error: 0.0807\n",
      "Epoch 191/200\n",
      "Standard error at epoch 190: 0.6833/0.1532\n",
      "MC error at epoch 190: 0.6755/0.1523\n",
      "14s - loss: 0.1721 - mean_squared_error: 0.0854\n",
      "Epoch 192/200\n",
      "11s - loss: 0.1667 - mean_squared_error: 0.0803\n",
      "Epoch 193/200\n",
      "Standard error at epoch 192: 0.6639/0.1510\n",
      "MC error at epoch 192: 0.6624/0.1508\n",
      "14s - loss: 0.1659 - mean_squared_error: 0.0804\n",
      "Epoch 194/200\n",
      "11s - loss: 0.1664 - mean_squared_error: 0.0814\n",
      "Epoch 195/200\n",
      "Standard error at epoch 194: 0.6868/0.1536\n",
      "MC error at epoch 194: 0.6757/0.1523\n",
      "14s - loss: 0.1670 - mean_squared_error: 0.0821\n",
      "Epoch 196/200\n",
      "11s - loss: 0.1666 - mean_squared_error: 0.0815\n",
      "Epoch 197/200\n",
      "Standard error at epoch 196: 0.6719/0.1519\n",
      "MC error at epoch 196: 0.6675/0.1514\n",
      "14s - loss: 0.1658 - mean_squared_error: 0.0807\n",
      "Epoch 198/200\n",
      "11s - loss: 0.1621 - mean_squared_error: 0.0774\n",
      "Epoch 199/200\n",
      "Standard error at epoch 198: 0.6776/0.1525\n",
      "MC error at epoch 198: 0.6682/0.1515\n",
      "14s - loss: 0.1633 - mean_squared_error: 0.0794\n",
      "Epoch 200/200\n",
      "11s - loss: 0.1596 - mean_squared_error: 0.0761\n"
     ]
    }
   ],
   "source": [
    "history_1 = model.fit(\n",
    "    X_train, Y_train,\n",
    "    verbose=2,\n",
    "    shuffle=True,\n",
    "    # validation_data=[X_test, Y_test],\n",
    "    batch_size=BATCH_SIZE, epochs=200, callbacks=[modeltest_1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best RMSE: 0.1503 Best Epoch: 180\n"
     ]
    }
   ],
   "source": [
    "print(\"Best RMSE: {:.4f} Best Epoch: {}\".format(\n",
    "    np.min([x[1] ** 0.5 for x in modeltest_1.history]), \n",
    "    (np.argmin([x[1] ** 0.5 for x in modeltest_1.history]) + 1)*2\n",
    "))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Standard LSTM w/o Dropout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Build model...\n"
     ]
    }
   ],
   "source": [
    "print('Build model...')\n",
    "model = get_model(edrop=0, rdrop=0, odrop=0, idrop=0, weight_decay=1e-10, lr=1e-3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "modeltest_2 = ModelTest(X_test, Yt=Y_test,\n",
    "                        test_every_X_epochs=2, verbose=0, T=1,\n",
    "                        mean_y_train=mean_y_train, std_y_train=std_y_train,\n",
    "                        loss='euclidean', batch_size=TEST_BATCH_SIZE)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/200\n",
      "Standard error at epoch 000: 1.0037/0.1856\n",
      "MC error at epoch 000: 1.0037/0.1856\n",
      "12s - loss: 1.0221 - mean_squared_error: 1.0220\n",
      "Epoch 2/200\n",
      "11s - loss: 0.9974 - mean_squared_error: 0.9973\n",
      "Epoch 3/200\n",
      "Standard error at epoch 002: 0.9893/0.1843\n",
      "MC error at epoch 002: 0.9893/0.1843\n",
      "12s - loss: 0.9676 - mean_squared_error: 0.9675\n",
      "Epoch 4/200\n",
      "11s - loss: 0.8916 - mean_squared_error: 0.8915\n",
      "Epoch 5/200\n",
      "Standard error at epoch 004: 0.7380/0.1592\n",
      "MC error at epoch 004: 0.7380/0.1592\n",
      "12s - loss: 0.6700 - mean_squared_error: 0.6699\n",
      "Epoch 6/200\n",
      "11s - loss: 0.4326 - mean_squared_error: 0.4325\n",
      "Epoch 7/200\n",
      "Standard error at epoch 006: 0.7081/0.1559\n",
      "MC error at epoch 006: 0.7081/0.1559\n",
      "12s - loss: 0.3097 - mean_squared_error: 0.3096\n",
      "Epoch 8/200\n",
      "11s - loss: 0.2355 - mean_squared_error: 0.2354\n",
      "Epoch 9/200\n",
      "Standard error at epoch 008: 0.7181/0.1570\n",
      "MC error at epoch 008: 0.7181/0.1570\n",
      "12s - loss: 0.1798 - mean_squared_error: 0.1797\n",
      "Epoch 10/200\n",
      "11s - loss: 0.1473 - mean_squared_error: 0.1472\n",
      "Epoch 11/200\n",
      "Standard error at epoch 010: 0.7334/0.1587\n",
      "MC error at epoch 010: 0.7334/0.1587\n",
      "12s - loss: 0.1229 - mean_squared_error: 0.1228\n",
      "Epoch 12/200\n",
      "11s - loss: 0.1038 - mean_squared_error: 0.1037\n",
      "Epoch 13/200\n",
      "Standard error at epoch 012: 0.8068/0.1664\n",
      "MC error at epoch 012: 0.8068/0.1664\n",
      "12s - loss: 0.0876 - mean_squared_error: 0.0875\n",
      "Epoch 14/200\n",
      "11s - loss: 0.0817 - mean_squared_error: 0.0816\n",
      "Epoch 15/200\n",
      "Standard error at epoch 014: 0.8103/0.1668\n",
      "MC error at epoch 014: 0.8103/0.1668\n",
      "12s - loss: 0.0716 - mean_squared_error: 0.0715\n",
      "Epoch 16/200\n",
      "11s - loss: 0.0635 - mean_squared_error: 0.0634\n",
      "Epoch 17/200\n",
      "Standard error at epoch 016: 0.8308/0.1689\n",
      "MC error at epoch 016: 0.8308/0.1689\n",
      "12s - loss: 0.0578 - mean_squared_error: 0.0577\n",
      "Epoch 18/200\n",
      "11s - loss: 0.0549 - mean_squared_error: 0.0548\n",
      "Epoch 19/200\n",
      "Standard error at epoch 018: 0.8556/0.1714\n",
      "MC error at epoch 018: 0.8556/0.1714\n",
      "12s - loss: 0.0563 - mean_squared_error: 0.0562\n",
      "Epoch 20/200\n",
      "11s - loss: 0.0521 - mean_squared_error: 0.0520\n",
      "Epoch 21/200\n",
      "Standard error at epoch 020: 0.8500/0.1708\n",
      "MC error at epoch 020: 0.8500/0.1708\n",
      "12s - loss: 0.0538 - mean_squared_error: 0.0537\n",
      "Epoch 22/200\n",
      "11s - loss: 0.0541 - mean_squared_error: 0.0540\n",
      "Epoch 23/200\n",
      "Standard error at epoch 022: 0.8588/0.1717\n",
      "MC error at epoch 022: 0.8588/0.1717\n",
      "12s - loss: 0.0496 - mean_squared_error: 0.0495\n",
      "Epoch 24/200\n",
      "11s - loss: 0.0440 - mean_squared_error: 0.0440\n",
      "Epoch 25/200\n",
      "Standard error at epoch 024: 0.8280/0.1686\n",
      "MC error at epoch 024: 0.8280/0.1686\n",
      "12s - loss: 0.0437 - mean_squared_error: 0.0437\n",
      "Epoch 26/200\n",
      "11s - loss: 0.0426 - mean_squared_error: 0.0425\n",
      "Epoch 27/200\n",
      "Standard error at epoch 026: 0.8503/0.1709\n",
      "MC error at epoch 026: 0.8503/0.1709\n",
      "12s - loss: 0.0499 - mean_squared_error: 0.0498\n",
      "Epoch 28/200\n",
      "11s - loss: 0.0440 - mean_squared_error: 0.0439\n",
      "Epoch 29/200\n",
      "Standard error at epoch 028: 0.8795/0.1738\n",
      "MC error at epoch 028: 0.8795/0.1738\n",
      "12s - loss: 0.0382 - mean_squared_error: 0.0381\n",
      "Epoch 30/200\n",
      "11s - loss: 0.0359 - mean_squared_error: 0.0358\n",
      "Epoch 31/200\n",
      "Standard error at epoch 030: 0.8448/0.1703\n",
      "MC error at epoch 030: 0.8448/0.1703\n",
      "12s - loss: 0.0382 - mean_squared_error: 0.0381\n",
      "Epoch 32/200\n",
      "11s - loss: 0.0369 - mean_squared_error: 0.0368\n",
      "Epoch 33/200\n",
      "Standard error at epoch 032: 0.8787/0.1737\n",
      "MC error at epoch 032: 0.8787/0.1737\n",
      "12s - loss: 0.0345 - mean_squared_error: 0.0344\n",
      "Epoch 34/200\n",
      "11s - loss: 0.0346 - mean_squared_error: 0.0345\n",
      "Epoch 35/200\n",
      "Standard error at epoch 034: 0.8262/0.1684\n",
      "MC error at epoch 034: 0.8262/0.1684\n",
      "12s - loss: 0.0354 - mean_squared_error: 0.0353\n",
      "Epoch 36/200\n",
      "11s - loss: 0.0353 - mean_squared_error: 0.0352\n",
      "Epoch 37/200\n",
      "Standard error at epoch 036: 0.8330/0.1691\n",
      "MC error at epoch 036: 0.8330/0.1691\n",
      "12s - loss: 0.0314 - mean_squared_error: 0.0313\n",
      "Epoch 38/200\n",
      "11s - loss: 0.0322 - mean_squared_error: 0.0321\n",
      "Epoch 39/200\n",
      "Standard error at epoch 038: 0.8392/0.1697\n",
      "MC error at epoch 038: 0.8392/0.1697\n",
      "12s - loss: 0.0308 - mean_squared_error: 0.0307\n",
      "Epoch 40/200\n",
      "11s - loss: 0.0309 - mean_squared_error: 0.0308\n",
      "Epoch 41/200\n",
      "Standard error at epoch 040: 0.8644/0.1723\n",
      "MC error at epoch 040: 0.8644/0.1723\n",
      "12s - loss: 0.0313 - mean_squared_error: 0.0312\n",
      "Epoch 42/200\n",
      "11s - loss: 0.0303 - mean_squared_error: 0.0302\n",
      "Epoch 43/200\n",
      "Standard error at epoch 042: 0.8516/0.1710\n",
      "MC error at epoch 042: 0.8516/0.1710\n",
      "12s - loss: 0.0302 - mean_squared_error: 0.0301\n",
      "Epoch 44/200\n",
      "11s - loss: 0.0281 - mean_squared_error: 0.0280\n",
      "Epoch 45/200\n",
      "Standard error at epoch 044: 0.8473/0.1706\n",
      "MC error at epoch 044: 0.8473/0.1706\n",
      "12s - loss: 0.0274 - mean_squared_error: 0.0273\n",
      "Epoch 46/200\n",
      "11s - loss: 0.1109 - mean_squared_error: 0.1109\n",
      "Epoch 47/200\n",
      "Standard error at epoch 046: 0.8013/0.1659\n",
      "MC error at epoch 046: 0.8013/0.1659\n",
      "12s - loss: 0.0567 - mean_squared_error: 0.0566\n",
      "Epoch 48/200\n",
      "12s - loss: 0.0396 - mean_squared_error: 0.0395\n",
      "Epoch 49/200\n",
      "Standard error at epoch 048: 0.8343/0.1692\n",
      "MC error at epoch 048: 0.8343/0.1692\n",
      "12s - loss: 0.0366 - mean_squared_error: 0.0365\n",
      "Epoch 50/200\n",
      "11s - loss: 0.0332 - mean_squared_error: 0.0331\n",
      "Epoch 51/200\n",
      "Standard error at epoch 050: 0.8666/0.1725\n",
      "MC error at epoch 050: 0.8666/0.1725\n",
      "12s - loss: 0.0319 - mean_squared_error: 0.0319\n",
      "Epoch 52/200\n",
      "11s - loss: 0.0301 - mean_squared_error: 0.0300\n",
      "Epoch 53/200\n",
      "Standard error at epoch 052: 0.8538/0.1712\n",
      "MC error at epoch 052: 0.8538/0.1712\n",
      "12s - loss: 0.0275 - mean_squared_error: 0.0274\n",
      "Epoch 54/200\n",
      "11s - loss: 0.0297 - mean_squared_error: 0.0296\n",
      "Epoch 55/200\n",
      "Standard error at epoch 054: 0.8525/0.1711\n",
      "MC error at epoch 054: 0.8525/0.1711\n",
      "12s - loss: 0.0307 - mean_squared_error: 0.0306\n",
      "Epoch 56/200\n",
      "11s - loss: 0.0302 - mean_squared_error: 0.0301\n",
      "Epoch 57/200\n",
      "Standard error at epoch 056: 0.8661/0.1724\n",
      "MC error at epoch 056: 0.8661/0.1724\n",
      "12s - loss: 0.0281 - mean_squared_error: 0.0280\n",
      "Epoch 58/200\n",
      "12s - loss: 0.0268 - mean_squared_error: 0.0267\n",
      "Epoch 59/200\n",
      "Standard error at epoch 058: 0.8613/0.1720\n",
      "MC error at epoch 058: 0.8613/0.1720\n",
      "12s - loss: 0.0280 - mean_squared_error: 0.0279\n",
      "Epoch 60/200\n",
      "11s - loss: 0.0265 - mean_squared_error: 0.0264\n",
      "Epoch 61/200\n",
      "Standard error at epoch 060: 0.8569/0.1715\n",
      "MC error at epoch 060: 0.8569/0.1715\n",
      "12s - loss: 0.0265 - mean_squared_error: 0.0264\n",
      "Epoch 62/200\n",
      "11s - loss: 0.0262 - mean_squared_error: 0.0261\n",
      "Epoch 63/200\n",
      "Standard error at epoch 062: 0.8613/0.1720\n",
      "MC error at epoch 062: 0.8613/0.1720\n",
      "12s - loss: 0.0271 - mean_squared_error: 0.0270\n",
      "Epoch 64/200\n",
      "11s - loss: 0.0268 - mean_squared_error: 0.0267\n",
      "Epoch 65/200\n",
      "Standard error at epoch 064: 0.8617/0.1720\n",
      "MC error at epoch 064: 0.8617/0.1720\n",
      "12s - loss: 0.0275 - mean_squared_error: 0.0274\n",
      "Epoch 66/200\n",
      "11s - loss: 0.0272 - mean_squared_error: 0.0271\n",
      "Epoch 67/200\n",
      "Standard error at epoch 066: 0.8399/0.1698\n",
      "MC error at epoch 066: 0.8399/0.1698\n",
      "12s - loss: 0.0275 - mean_squared_error: 0.0274\n",
      "Epoch 68/200\n",
      "11s - loss: 0.0277 - mean_squared_error: 0.0276\n",
      "Epoch 69/200\n",
      "Standard error at epoch 068: 0.8461/0.1704\n",
      "MC error at epoch 068: 0.8461/0.1704\n",
      "12s - loss: 0.0267 - mean_squared_error: 0.0266\n",
      "Epoch 70/200\n",
      "11s - loss: 0.0244 - mean_squared_error: 0.0243\n",
      "Epoch 71/200\n",
      "Standard error at epoch 070: 0.8371/0.1695\n",
      "MC error at epoch 070: 0.8371/0.1695\n",
      "12s - loss: 0.0260 - mean_squared_error: 0.0259\n",
      "Epoch 72/200\n",
      "11s - loss: 0.0254 - mean_squared_error: 0.0253\n",
      "Epoch 73/200\n",
      "Standard error at epoch 072: 0.8584/0.1717\n",
      "MC error at epoch 072: 0.8584/0.1717\n",
      "12s - loss: 0.0271 - mean_squared_error: 0.0270\n",
      "Epoch 74/200\n",
      "12s - loss: 0.0258 - mean_squared_error: 0.0257\n",
      "Epoch 75/200\n",
      "Standard error at epoch 074: 0.8379/0.1696\n",
      "MC error at epoch 074: 0.8379/0.1696\n",
      "12s - loss: 0.0253 - mean_squared_error: 0.0252\n",
      "Epoch 76/200\n",
      "12s - loss: 0.0272 - mean_squared_error: 0.0271\n",
      "Epoch 77/200\n",
      "Standard error at epoch 076: 0.8355/0.1694\n",
      "MC error at epoch 076: 0.8355/0.1694\n",
      "12s - loss: 0.0245 - mean_squared_error: 0.0245\n",
      "Epoch 78/200\n",
      "12s - loss: 0.0233 - mean_squared_error: 0.0233\n",
      "Epoch 79/200\n",
      "Standard error at epoch 078: 0.8522/0.1710\n",
      "MC error at epoch 078: 0.8522/0.1710\n",
      "12s - loss: 0.0231 - mean_squared_error: 0.0230\n",
      "Epoch 80/200\n",
      "12s - loss: 0.0225 - mean_squared_error: 0.0224\n",
      "Epoch 81/200\n",
      "Standard error at epoch 080: 0.8329/0.1691\n",
      "MC error at epoch 080: 0.8329/0.1691\n",
      "12s - loss: 0.0226 - mean_squared_error: 0.0225\n",
      "Epoch 82/200\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "12s - loss: 0.0219 - mean_squared_error: 0.0218\n",
      "Epoch 83/200\n",
      "Standard error at epoch 082: 0.8369/0.1695\n",
      "MC error at epoch 082: 0.8369/0.1695\n",
      "12s - loss: 0.0239 - mean_squared_error: 0.0238\n",
      "Epoch 84/200\n",
      "12s - loss: 0.0229 - mean_squared_error: 0.0228\n",
      "Epoch 85/200\n",
      "Standard error at epoch 084: 0.8380/0.1696\n",
      "MC error at epoch 084: 0.8380/0.1696\n",
      "12s - loss: 0.0222 - mean_squared_error: 0.0221\n",
      "Epoch 86/200\n",
      "12s - loss: 0.0233 - mean_squared_error: 0.0232\n",
      "Epoch 87/200\n",
      "Standard error at epoch 086: 0.8300/0.1688\n",
      "MC error at epoch 086: 0.8300/0.1688\n",
      "12s - loss: 0.0227 - mean_squared_error: 0.0227\n",
      "Epoch 88/200\n",
      "12s - loss: 0.0231 - mean_squared_error: 0.0231\n",
      "Epoch 89/200\n",
      "Standard error at epoch 088: 0.8452/0.1703\n",
      "MC error at epoch 088: 0.8452/0.1703\n",
      "12s - loss: 0.0254 - mean_squared_error: 0.0253\n",
      "Epoch 90/200\n",
      "12s - loss: 0.0244 - mean_squared_error: 0.0243\n",
      "Epoch 91/200\n",
      "Standard error at epoch 090: 0.8329/0.1691\n",
      "MC error at epoch 090: 0.8329/0.1691\n",
      "12s - loss: 0.0231 - mean_squared_error: 0.0230\n",
      "Epoch 92/200\n",
      "11s - loss: 0.0218 - mean_squared_error: 0.0217\n",
      "Epoch 93/200\n",
      "Standard error at epoch 092: 0.8433/0.1702\n",
      "MC error at epoch 092: 0.8433/0.1702\n",
      "12s - loss: 0.0216 - mean_squared_error: 0.0215\n",
      "Epoch 94/200\n",
      "11s - loss: 0.0208 - mean_squared_error: 0.0207\n",
      "Epoch 95/200\n",
      "Standard error at epoch 094: 0.8479/0.1706\n",
      "MC error at epoch 094: 0.8479/0.1706\n",
      "12s - loss: 0.0197 - mean_squared_error: 0.0196\n",
      "Epoch 96/200\n",
      "11s - loss: 0.0215 - mean_squared_error: 0.0214\n",
      "Epoch 97/200\n",
      "Standard error at epoch 096: 0.8315/0.1690\n",
      "MC error at epoch 096: 0.8315/0.1690\n",
      "12s - loss: 0.0207 - mean_squared_error: 0.0206\n",
      "Epoch 98/200\n",
      "12s - loss: 0.0208 - mean_squared_error: 0.0207\n",
      "Epoch 99/200\n",
      "Standard error at epoch 098: 0.8147/0.1672\n",
      "MC error at epoch 098: 0.8147/0.1672\n",
      "12s - loss: 0.0202 - mean_squared_error: 0.0202\n",
      "Epoch 100/200\n",
      "12s - loss: 0.0201 - mean_squared_error: 0.0200\n",
      "Epoch 101/200\n",
      "Standard error at epoch 100: 0.8236/0.1682\n",
      "MC error at epoch 100: 0.8236/0.1682\n",
      "12s - loss: 0.0207 - mean_squared_error: 0.0206\n",
      "Epoch 102/200\n",
      "12s - loss: 0.0231 - mean_squared_error: 0.0230\n",
      "Epoch 103/200\n",
      "Standard error at epoch 102: 0.8369/0.1695\n",
      "MC error at epoch 102: 0.8369/0.1695\n",
      "12s - loss: 0.0231 - mean_squared_error: 0.0230\n",
      "Epoch 104/200\n",
      "12s - loss: 0.0215 - mean_squared_error: 0.0215\n",
      "Epoch 105/200\n",
      "Standard error at epoch 104: 0.8153/0.1673\n",
      "MC error at epoch 104: 0.8153/0.1673\n",
      "12s - loss: 0.0236 - mean_squared_error: 0.0235\n",
      "Epoch 106/200\n",
      "11s - loss: 0.0252 - mean_squared_error: 0.0251\n",
      "Epoch 107/200\n",
      "Standard error at epoch 106: 0.8104/0.1668\n",
      "MC error at epoch 106: 0.8104/0.1668\n",
      "12s - loss: 0.0224 - mean_squared_error: 0.0224\n",
      "Epoch 108/200\n",
      "11s - loss: 0.0225 - mean_squared_error: 0.0224\n",
      "Epoch 109/200\n",
      "Standard error at epoch 108: 0.8072/0.1665\n",
      "MC error at epoch 108: 0.8072/0.1665\n",
      "12s - loss: 0.0208 - mean_squared_error: 0.0207\n",
      "Epoch 110/200\n",
      "11s - loss: 0.0193 - mean_squared_error: 0.0192\n",
      "Epoch 111/200\n",
      "Standard error at epoch 110: 0.8476/0.1706\n",
      "MC error at epoch 110: 0.8476/0.1706\n",
      "12s - loss: 0.0181 - mean_squared_error: 0.0180\n",
      "Epoch 112/200\n",
      "11s - loss: 0.0186 - mean_squared_error: 0.0186\n",
      "Epoch 113/200\n",
      "Standard error at epoch 112: 0.8264/0.1684\n",
      "MC error at epoch 112: 0.8264/0.1684\n",
      "12s - loss: 0.0195 - mean_squared_error: 0.0194\n",
      "Epoch 114/200\n",
      "12s - loss: 0.0189 - mean_squared_error: 0.0188\n",
      "Epoch 115/200\n",
      "Standard error at epoch 114: 0.8340/0.1692\n",
      "MC error at epoch 114: 0.8340/0.1692\n",
      "13s - loss: 0.0193 - mean_squared_error: 0.0192\n",
      "Epoch 116/200\n",
      "12s - loss: 0.0190 - mean_squared_error: 0.0189\n",
      "Epoch 117/200\n",
      "Standard error at epoch 116: 0.8327/0.1691\n",
      "MC error at epoch 116: 0.8327/0.1691\n",
      "12s - loss: 0.0199 - mean_squared_error: 0.0198\n",
      "Epoch 118/200\n",
      "12s - loss: 0.0191 - mean_squared_error: 0.0190\n",
      "Epoch 119/200\n",
      "Standard error at epoch 118: 0.8081/0.1666\n",
      "MC error at epoch 118: 0.8081/0.1666\n",
      "12s - loss: 0.0189 - mean_squared_error: 0.0188\n",
      "Epoch 120/200\n",
      "12s - loss: 0.0193 - mean_squared_error: 0.0192\n",
      "Epoch 121/200\n",
      "Standard error at epoch 120: 0.7964/0.1654\n",
      "MC error at epoch 120: 0.7964/0.1654\n",
      "12s - loss: 0.0192 - mean_squared_error: 0.0192\n",
      "Epoch 122/200\n",
      "12s - loss: 0.0197 - mean_squared_error: 0.0196\n",
      "Epoch 123/200\n",
      "Standard error at epoch 122: 0.8280/0.1686\n",
      "MC error at epoch 122: 0.8280/0.1686\n",
      "12s - loss: 0.0202 - mean_squared_error: 0.0201\n",
      "Epoch 124/200\n",
      "12s - loss: 0.0204 - mean_squared_error: 0.0203\n",
      "Epoch 125/200\n",
      "Standard error at epoch 124: 0.8018/0.1659\n",
      "MC error at epoch 124: 0.8018/0.1659\n",
      "12s - loss: 0.0215 - mean_squared_error: 0.0214\n",
      "Epoch 126/200\n",
      "12s - loss: 0.0195 - mean_squared_error: 0.0194\n",
      "Epoch 127/200\n",
      "Standard error at epoch 126: 0.8236/0.1682\n",
      "MC error at epoch 126: 0.8236/0.1682\n",
      "12s - loss: 0.0185 - mean_squared_error: 0.0184\n",
      "Epoch 128/200\n",
      "12s - loss: 0.0180 - mean_squared_error: 0.0179\n",
      "Epoch 129/200\n",
      "Standard error at epoch 128: 0.8328/0.1691\n",
      "MC error at epoch 128: 0.8328/0.1691\n",
      "12s - loss: 0.0176 - mean_squared_error: 0.0175\n",
      "Epoch 130/200\n",
      "11s - loss: 0.0177 - mean_squared_error: 0.0176\n",
      "Epoch 131/200\n",
      "Standard error at epoch 130: 0.7953/0.1652\n",
      "MC error at epoch 130: 0.7953/0.1652\n",
      "12s - loss: 0.0180 - mean_squared_error: 0.0180\n",
      "Epoch 132/200\n",
      "11s - loss: 0.0185 - mean_squared_error: 0.0184\n",
      "Epoch 133/200\n",
      "Standard error at epoch 132: 0.8266/0.1685\n",
      "MC error at epoch 132: 0.8266/0.1685\n",
      "12s - loss: 0.0181 - mean_squared_error: 0.0180\n",
      "Epoch 134/200\n",
      "11s - loss: 0.0491 - mean_squared_error: 0.0490\n",
      "Epoch 135/200\n",
      "Standard error at epoch 134: 0.8042/0.1662\n",
      "MC error at epoch 134: 0.8042/0.1662\n",
      "12s - loss: 0.0517 - mean_squared_error: 0.0516\n",
      "Epoch 136/200\n",
      "11s - loss: 0.0324 - mean_squared_error: 0.0323\n",
      "Epoch 137/200\n",
      "Standard error at epoch 136: 0.8223/0.1680\n",
      "MC error at epoch 136: 0.8223/0.1680\n",
      "12s - loss: 0.0286 - mean_squared_error: 0.0285\n",
      "Epoch 138/200\n",
      "11s - loss: 0.0235 - mean_squared_error: 0.0234\n",
      "Epoch 139/200\n",
      "Standard error at epoch 138: 0.8372/0.1695\n",
      "MC error at epoch 138: 0.8372/0.1695\n",
      "12s - loss: 0.0230 - mean_squared_error: 0.0229\n",
      "Epoch 140/200\n",
      "11s - loss: 0.0209 - mean_squared_error: 0.0208\n",
      "Epoch 141/200\n",
      "Standard error at epoch 140: 0.8386/0.1697\n",
      "MC error at epoch 140: 0.8386/0.1697\n",
      "12s - loss: 0.0188 - mean_squared_error: 0.0187\n",
      "Epoch 142/200\n",
      "11s - loss: 0.0276 - mean_squared_error: 0.0275\n",
      "Epoch 143/200\n",
      "Standard error at epoch 142: 0.8343/0.1692\n",
      "MC error at epoch 142: 0.8343/0.1692\n",
      "12s - loss: 0.0205 - mean_squared_error: 0.0204\n",
      "Epoch 144/200\n",
      "11s - loss: 0.0191 - mean_squared_error: 0.0190\n",
      "Epoch 145/200\n",
      "Standard error at epoch 144: 0.8383/0.1696\n",
      "MC error at epoch 144: 0.8383/0.1696\n",
      "12s - loss: 0.0206 - mean_squared_error: 0.0205\n",
      "Epoch 146/200\n",
      "11s - loss: 0.0199 - mean_squared_error: 0.0198\n",
      "Epoch 147/200\n",
      "Standard error at epoch 146: 0.8224/0.1680\n",
      "MC error at epoch 146: 0.8224/0.1680\n",
      "12s - loss: 0.0187 - mean_squared_error: 0.0186\n",
      "Epoch 148/200\n",
      "12s - loss: 0.0173 - mean_squared_error: 0.0172\n",
      "Epoch 149/200\n",
      "Standard error at epoch 148: 0.8126/0.1670\n",
      "MC error at epoch 148: 0.8126/0.1670\n",
      "12s - loss: 0.0181 - mean_squared_error: 0.0180\n",
      "Epoch 150/200\n",
      "11s - loss: 0.0186 - mean_squared_error: 0.0185\n",
      "Epoch 151/200\n",
      "Standard error at epoch 150: 0.8354/0.1693\n",
      "MC error at epoch 150: 0.8354/0.1693\n",
      "12s - loss: 0.0176 - mean_squared_error: 0.0175\n",
      "Epoch 152/200\n",
      "11s - loss: 0.0177 - mean_squared_error: 0.0176\n",
      "Epoch 153/200\n",
      "Standard error at epoch 152: 0.8235/0.1681\n",
      "MC error at epoch 152: 0.8235/0.1681\n",
      "12s - loss: 0.0167 - mean_squared_error: 0.0166\n",
      "Epoch 154/200\n",
      "11s - loss: 0.0165 - mean_squared_error: 0.0164\n",
      "Epoch 155/200\n",
      "Standard error at epoch 154: 0.8160/0.1674\n",
      "MC error at epoch 154: 0.8160/0.1674\n",
      "12s - loss: 0.0167 - mean_squared_error: 0.0166\n",
      "Epoch 156/200\n",
      "11s - loss: 0.0167 - mean_squared_error: 0.0166\n",
      "Epoch 157/200\n",
      "Standard error at epoch 156: 0.8257/0.1684\n",
      "MC error at epoch 156: 0.8257/0.1684\n",
      "12s - loss: 0.0166 - mean_squared_error: 0.0165\n",
      "Epoch 158/200\n",
      "12s - loss: 0.0168 - mean_squared_error: 0.0167\n",
      "Epoch 159/200\n",
      "Standard error at epoch 158: 0.8253/0.1683\n",
      "MC error at epoch 158: 0.8253/0.1683\n",
      "12s - loss: 0.0165 - mean_squared_error: 0.0164\n",
      "Epoch 160/200\n",
      "12s - loss: 0.0169 - mean_squared_error: 0.0169\n",
      "Epoch 161/200\n",
      "Standard error at epoch 160: 0.8351/0.1693\n",
      "MC error at epoch 160: 0.8351/0.1693\n",
      "12s - loss: 0.0176 - mean_squared_error: 0.0175\n",
      "Epoch 162/200\n",
      "11s - loss: 0.0177 - mean_squared_error: 0.0176\n",
      "Epoch 163/200\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Standard error at epoch 162: 0.8084/0.1666\n",
      "MC error at epoch 162: 0.8084/0.1666\n",
      "12s - loss: 0.0175 - mean_squared_error: 0.0174\n",
      "Epoch 164/200\n",
      "11s - loss: 0.0178 - mean_squared_error: 0.0177\n",
      "Epoch 165/200\n",
      "Standard error at epoch 164: 0.8343/0.1692\n",
      "MC error at epoch 164: 0.8343/0.1692\n",
      "12s - loss: 0.0171 - mean_squared_error: 0.0170\n",
      "Epoch 166/200\n",
      "12s - loss: 0.0174 - mean_squared_error: 0.0173\n",
      "Epoch 167/200\n",
      "Standard error at epoch 166: 0.8435/0.1702\n",
      "MC error at epoch 166: 0.8435/0.1702\n",
      "12s - loss: 0.0170 - mean_squared_error: 0.0169\n",
      "Epoch 168/200\n",
      "12s - loss: 0.0169 - mean_squared_error: 0.0168\n",
      "Epoch 169/200\n",
      "Standard error at epoch 168: 0.8110/0.1669\n",
      "MC error at epoch 168: 0.8110/0.1669\n",
      "12s - loss: 0.0176 - mean_squared_error: 0.0175\n",
      "Epoch 170/200\n",
      "11s - loss: 0.0171 - mean_squared_error: 0.0170\n",
      "Epoch 171/200\n",
      "Standard error at epoch 170: 0.8201/0.1678\n",
      "MC error at epoch 170: 0.8201/0.1678\n",
      "12s - loss: 0.0176 - mean_squared_error: 0.0175\n",
      "Epoch 172/200\n",
      "12s - loss: 0.0174 - mean_squared_error: 0.0173\n",
      "Epoch 173/200\n",
      "Standard error at epoch 172: 0.8099/0.1667\n",
      "MC error at epoch 172: 0.8099/0.1667\n",
      "12s - loss: 0.0172 - mean_squared_error: 0.0171\n",
      "Epoch 174/200\n",
      "11s - loss: 0.0172 - mean_squared_error: 0.0172\n",
      "Epoch 175/200\n",
      "Standard error at epoch 174: 0.8042/0.1662\n",
      "MC error at epoch 174: 0.8042/0.1662\n",
      "12s - loss: 0.0165 - mean_squared_error: 0.0164\n",
      "Epoch 176/200\n",
      "11s - loss: 0.0168 - mean_squared_error: 0.0167\n",
      "Epoch 177/200\n",
      "Standard error at epoch 176: 0.8285/0.1686\n",
      "MC error at epoch 176: 0.8285/0.1686\n",
      "12s - loss: 0.0167 - mean_squared_error: 0.0166\n",
      "Epoch 178/200\n",
      "12s - loss: 0.0167 - mean_squared_error: 0.0167\n",
      "Epoch 179/200\n",
      "Standard error at epoch 178: 0.8017/0.1659\n",
      "MC error at epoch 178: 0.8017/0.1659\n",
      "12s - loss: 0.0174 - mean_squared_error: 0.0174\n",
      "Epoch 180/200\n",
      "12s - loss: 0.0184 - mean_squared_error: 0.0184\n",
      "Epoch 181/200\n",
      "Standard error at epoch 180: 0.8091/0.1667\n",
      "MC error at epoch 180: 0.8091/0.1667\n",
      "12s - loss: 0.0189 - mean_squared_error: 0.0188\n",
      "Epoch 182/200\n",
      "12s - loss: 0.0172 - mean_squared_error: 0.0171\n",
      "Epoch 183/200\n",
      "Standard error at epoch 182: 0.8201/0.1678\n",
      "MC error at epoch 182: 0.8201/0.1678\n",
      "12s - loss: 0.0166 - mean_squared_error: 0.0165\n",
      "Epoch 184/200\n",
      "11s - loss: 0.0167 - mean_squared_error: 0.0166\n",
      "Epoch 185/200\n",
      "Standard error at epoch 184: 0.7956/0.1653\n",
      "MC error at epoch 184: 0.7956/0.1653\n",
      "12s - loss: 0.0157 - mean_squared_error: 0.0156\n",
      "Epoch 186/200\n",
      "11s - loss: 0.0156 - mean_squared_error: 0.0156\n",
      "Epoch 187/200\n",
      "Standard error at epoch 186: 0.8126/0.1670\n",
      "MC error at epoch 186: 0.8126/0.1670\n",
      "12s - loss: 0.0171 - mean_squared_error: 0.0170\n",
      "Epoch 188/200\n",
      "11s - loss: 0.0169 - mean_squared_error: 0.0168\n",
      "Epoch 189/200\n",
      "Standard error at epoch 188: 0.8111/0.1669\n",
      "MC error at epoch 188: 0.8111/0.1669\n",
      "12s - loss: 0.0168 - mean_squared_error: 0.0167\n",
      "Epoch 190/200\n",
      "11s - loss: 0.0161 - mean_squared_error: 0.0160\n",
      "Epoch 191/200\n",
      "Standard error at epoch 190: 0.8019/0.1659\n",
      "MC error at epoch 190: 0.8019/0.1659\n",
      "12s - loss: 0.0167 - mean_squared_error: 0.0166\n",
      "Epoch 192/200\n",
      "11s - loss: 0.0171 - mean_squared_error: 0.0170\n",
      "Epoch 193/200\n",
      "Standard error at epoch 192: 0.7957/0.1653\n",
      "MC error at epoch 192: 0.7957/0.1653\n",
      "12s - loss: 0.0161 - mean_squared_error: 0.0160\n",
      "Epoch 194/200\n",
      "11s - loss: 0.0159 - mean_squared_error: 0.0158\n",
      "Epoch 195/200\n",
      "Standard error at epoch 194: 0.8011/0.1658\n",
      "MC error at epoch 194: 0.8011/0.1658\n",
      "12s - loss: 0.0156 - mean_squared_error: 0.0155\n",
      "Epoch 196/200\n",
      "11s - loss: 0.0173 - mean_squared_error: 0.0172\n",
      "Epoch 197/200\n",
      "Standard error at epoch 196: 0.8159/0.1674\n",
      "MC error at epoch 196: 0.8159/0.1674\n",
      "12s - loss: 0.0173 - mean_squared_error: 0.0172\n",
      "Epoch 198/200\n",
      "11s - loss: 0.0162 - mean_squared_error: 0.0161\n",
      "Epoch 199/200\n",
      "Standard error at epoch 198: 0.7989/0.1656\n",
      "MC error at epoch 198: 0.7989/0.1656\n",
      "12s - loss: 0.0163 - mean_squared_error: 0.0162\n",
      "Epoch 200/200\n",
      "11s - loss: 0.0160 - mean_squared_error: 0.0160\n"
     ]
    }
   ],
   "source": [
    "history_2 = model.fit(\n",
    "    X_train, Y_train,\n",
    "    verbose=2,\n",
    "    shuffle=True,\n",
    "    # validation_data=[X_test, Y_test],\n",
    "    batch_size=BATCH_SIZE, epochs=200, callbacks=[modeltest_2]\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best RMSE: 0.1559 Best Epoch: 8\n"
     ]
    }
   ],
   "source": [
    "print(\"Best RMSE: {:.4f} Best Epoch: {}\".format(\n",
    "    np.min([x[1] ** 0.5 for x in modeltest_2.history]), \n",
    "    (np.argmin([x[1] ** 0.5 for x in modeltest_2.history]) + 1)*2\n",
    "))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## LSTM with Standard Dropout (different mask at differnt time steps)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Build model...\n"
     ]
    }
   ],
   "source": [
    "print('Build model...')\n",
    "model = get_model(edrop=0.3, rdrop=0, odrop=0.3, idrop=0, weight_decay=1e-4, lr=1e-3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "modeltest_3 = ModelTest(X_test, Yt=Y_test,\n",
    "                        test_every_X_epochs=2, verbose=0, T=10,\n",
    "                        mean_y_train=mean_y_train, std_y_train=std_y_train,\n",
    "                        loss='euclidean', batch_size=TEST_BATCH_SIZE)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/200\n",
      "Standard error at epoch 000: 1.0070/0.1859\n",
      "MC error at epoch 000: 1.0068/0.1859\n",
      "14s - loss: 76.7270 - mean_squared_error: 1.0347\n",
      "Epoch 2/200\n",
      "11s - loss: 59.6876 - mean_squared_error: 1.0039\n",
      "Epoch 3/200\n",
      "Standard error at epoch 002: 1.0091/0.1861\n",
      "MC error at epoch 002: 1.0086/0.1861\n",
      "14s - loss: 46.2750 - mean_squared_error: 0.9993\n",
      "Epoch 4/200\n",
      "11s - loss: 35.7048 - mean_squared_error: 0.9974\n",
      "Epoch 5/200\n",
      "Standard error at epoch 004: 1.0013/0.1854\n",
      "MC error at epoch 004: 1.0014/0.1854\n",
      "14s - loss: 27.4043 - mean_squared_error: 0.9963\n",
      "Epoch 6/200\n",
      "11s - loss: 20.9170 - mean_squared_error: 0.9887\n",
      "Epoch 7/200\n",
      "Standard error at epoch 006: 0.9632/0.1818\n",
      "MC error at epoch 006: 0.9662/0.1821\n",
      "14s - loss: 15.8819 - mean_squared_error: 0.9751\n",
      "Epoch 8/200\n",
      "11s - loss: 12.0042 - mean_squared_error: 0.9417\n",
      "Epoch 9/200\n",
      "Standard error at epoch 008: 0.7466/0.1601\n",
      "MC error at epoch 008: 0.7558/0.1611\n",
      "14s - loss: 9.0166 - mean_squared_error: 0.8642\n",
      "Epoch 10/200\n",
      "11s - loss: 6.6235 - mean_squared_error: 0.6075\n",
      "Epoch 11/200\n",
      "Standard error at epoch 010: 0.5916/0.1425\n",
      "MC error at epoch 010: 0.5916/0.1425\n",
      "14s - loss: 4.8836 - mean_squared_error: 0.4120\n",
      "Epoch 12/200\n",
      "11s - loss: 3.6519 - mean_squared_error: 0.3152\n",
      "Epoch 13/200\n",
      "Standard error at epoch 012: 0.6118/0.1449\n",
      "MC error at epoch 012: 0.6127/0.1450\n",
      "14s - loss: 2.7628 - mean_squared_error: 0.2575\n",
      "Epoch 14/200\n",
      "11s - loss: 2.1190 - mean_squared_error: 0.2230\n",
      "Epoch 15/200\n",
      "Standard error at epoch 014: 0.6487/0.1492\n",
      "MC error at epoch 014: 0.6416/0.1484\n",
      "14s - loss: 1.6486 - mean_squared_error: 0.1975\n",
      "Epoch 16/200\n",
      "11s - loss: 1.3107 - mean_squared_error: 0.1835\n",
      "Epoch 17/200\n",
      "Standard error at epoch 016: 0.6481/0.1492\n",
      "MC error at epoch 016: 0.6488/0.1492\n",
      "14s - loss: 1.0577 - mean_squared_error: 0.1664\n",
      "Epoch 18/200\n",
      "11s - loss: 0.8684 - mean_squared_error: 0.1509\n",
      "Epoch 19/200\n",
      "Standard error at epoch 018: 0.6555/0.1500\n",
      "MC error at epoch 018: 0.6533/0.1498\n",
      "14s - loss: 0.7333 - mean_squared_error: 0.1440\n",
      "Epoch 20/200\n",
      "11s - loss: 0.6325 - mean_squared_error: 0.1382\n",
      "Epoch 21/200\n",
      "Standard error at epoch 020: 0.6501/0.1494\n",
      "MC error at epoch 020: 0.6487/0.1492\n",
      "14s - loss: 0.5497 - mean_squared_error: 0.1275\n",
      "Epoch 22/200\n",
      "11s - loss: 0.4914 - mean_squared_error: 0.1240\n",
      "Epoch 23/200\n",
      "Standard error at epoch 022: 0.6696/0.1516\n",
      "MC error at epoch 022: 0.6661/0.1512\n",
      "14s - loss: 0.4498 - mean_squared_error: 0.1238\n",
      "Epoch 24/200\n",
      "11s - loss: 0.4075 - mean_squared_error: 0.1154\n",
      "Epoch 25/200\n",
      "Standard error at epoch 024: 0.6760/0.1523\n",
      "MC error at epoch 024: 0.6749/0.1522\n",
      "14s - loss: 0.3733 - mean_squared_error: 0.1088\n",
      "Epoch 26/200\n",
      "11s - loss: 0.3519 - mean_squared_error: 0.1096\n",
      "Epoch 27/200\n",
      "Standard error at epoch 026: 0.6610/0.1506\n",
      "MC error at epoch 026: 0.6610/0.1506\n",
      "14s - loss: 0.3348 - mean_squared_error: 0.1102\n",
      "Epoch 28/200\n",
      "11s - loss: 0.3179 - mean_squared_error: 0.1078\n",
      "Epoch 29/200\n",
      "Standard error at epoch 028: 0.6754/0.1523\n",
      "MC error at epoch 028: 0.6735/0.1521\n",
      "14s - loss: 0.2978 - mean_squared_error: 0.1012\n",
      "Epoch 30/200\n",
      "11s - loss: 0.2817 - mean_squared_error: 0.0978\n",
      "Epoch 31/200\n",
      "Standard error at epoch 030: 0.6808/0.1529\n",
      "MC error at epoch 030: 0.6774/0.1525\n",
      "14s - loss: 0.2713 - mean_squared_error: 0.0976\n",
      "Epoch 32/200\n",
      "12s - loss: 0.2605 - mean_squared_error: 0.0958\n",
      "Epoch 33/200\n",
      "Standard error at epoch 032: 0.6642/0.1510\n",
      "MC error at epoch 032: 0.6646/0.1510\n",
      "14s - loss: 0.2557 - mean_squared_error: 0.0981\n",
      "Epoch 34/200\n",
      "11s - loss: 0.2493 - mean_squared_error: 0.0977\n",
      "Epoch 35/200\n",
      "Standard error at epoch 034: 0.6767/0.1524\n",
      "MC error at epoch 034: 0.6738/0.1521\n",
      "14s - loss: 0.2436 - mean_squared_error: 0.0974\n",
      "Epoch 36/200\n",
      "11s - loss: 0.2337 - mean_squared_error: 0.0930\n",
      "Epoch 37/200\n",
      "Standard error at epoch 036: 0.6633/0.1509\n",
      "MC error at epoch 036: 0.6634/0.1509\n",
      "14s - loss: 0.2250 - mean_squared_error: 0.0901\n",
      "Epoch 38/200\n",
      "12s - loss: 0.2181 - mean_squared_error: 0.0881\n",
      "Epoch 39/200\n",
      "Standard error at epoch 038: 0.6764/0.1524\n",
      "MC error at epoch 038: 0.6768/0.1524\n",
      "14s - loss: 0.2166 - mean_squared_error: 0.0906\n",
      "Epoch 40/200\n",
      "12s - loss: 0.2171 - mean_squared_error: 0.0936\n",
      "Epoch 41/200\n",
      "Standard error at epoch 040: 0.6599/0.1505\n",
      "MC error at epoch 040: 0.6561/0.1501\n",
      "15s - loss: 0.2123 - mean_squared_error: 0.0915\n",
      "Epoch 42/200\n",
      "12s - loss: 0.2107 - mean_squared_error: 0.0919\n",
      "Epoch 43/200\n",
      "Standard error at epoch 042: 0.6745/0.1522\n",
      "MC error at epoch 042: 0.6695/0.1516\n",
      "14s - loss: 0.2023 - mean_squared_error: 0.0868\n",
      "Epoch 44/200\n",
      "11s - loss: 0.1980 - mean_squared_error: 0.0856\n",
      "Epoch 45/200\n",
      "Standard error at epoch 044: 0.6702/0.1517\n",
      "MC error at epoch 044: 0.6676/0.1514\n",
      "14s - loss: 0.1935 - mean_squared_error: 0.0842\n",
      "Epoch 46/200\n",
      "11s - loss: 0.1917 - mean_squared_error: 0.0847\n",
      "Epoch 47/200\n",
      "Standard error at epoch 046: 0.6747/0.1522\n",
      "MC error at epoch 046: 0.6748/0.1522\n",
      "14s - loss: 0.1912 - mean_squared_error: 0.0862\n",
      "Epoch 48/200\n",
      "11s - loss: 0.1941 - mean_squared_error: 0.0898\n",
      "Epoch 49/200\n",
      "Standard error at epoch 048: 0.6598/0.1505\n",
      "MC error at epoch 048: 0.6598/0.1505\n",
      "14s - loss: 0.1889 - mean_squared_error: 0.0856\n",
      "Epoch 50/200\n",
      "11s - loss: 0.1814 - mean_squared_error: 0.0807\n",
      "Epoch 51/200\n",
      "Standard error at epoch 050: 0.6663/0.1512\n",
      "MC error at epoch 050: 0.6657/0.1512\n",
      "14s - loss: 0.1834 - mean_squared_error: 0.0842\n",
      "Epoch 52/200\n",
      "11s - loss: 0.1812 - mean_squared_error: 0.0831\n",
      "Epoch 53/200\n",
      "Standard error at epoch 052: 0.6778/0.1525\n",
      "MC error at epoch 052: 0.6748/0.1522\n",
      "14s - loss: 0.1794 - mean_squared_error: 0.0826\n",
      "Epoch 54/200\n",
      "11s - loss: 0.1779 - mean_squared_error: 0.0826\n",
      "Epoch 55/200\n",
      "Standard error at epoch 054: 0.6635/0.1509\n",
      "MC error at epoch 054: 0.6627/0.1508\n",
      "14s - loss: 0.1761 - mean_squared_error: 0.0819\n",
      "Epoch 56/200\n",
      "11s - loss: 0.1732 - mean_squared_error: 0.0803\n",
      "Epoch 57/200\n",
      "Standard error at epoch 056: 0.6601/0.1505\n",
      "MC error at epoch 056: 0.6590/0.1504\n",
      "14s - loss: 0.1737 - mean_squared_error: 0.0816\n",
      "Epoch 58/200\n",
      "11s - loss: 0.1731 - mean_squared_error: 0.0816\n",
      "Epoch 59/200\n",
      "Standard error at epoch 058: 0.6556/0.1500\n",
      "MC error at epoch 058: 0.6550/0.1500\n",
      "14s - loss: 0.1704 - mean_squared_error: 0.0797\n",
      "Epoch 60/200\n",
      "11s - loss: 0.1684 - mean_squared_error: 0.0789\n",
      "Epoch 61/200\n",
      "Standard error at epoch 060: 0.6680/0.1514\n",
      "MC error at epoch 060: 0.6662/0.1512\n",
      "14s - loss: 0.1686 - mean_squared_error: 0.0799\n",
      "Epoch 62/200\n",
      "11s - loss: 0.1701 - mean_squared_error: 0.0821\n",
      "Epoch 63/200\n",
      "Standard error at epoch 062: 0.6514/0.1495\n",
      "MC error at epoch 062: 0.6493/0.1493\n",
      "14s - loss: 0.1646 - mean_squared_error: 0.0773\n",
      "Epoch 64/200\n",
      "11s - loss: 0.1644 - mean_squared_error: 0.0783\n",
      "Epoch 65/200\n",
      "Standard error at epoch 064: 0.6600/0.1505\n",
      "MC error at epoch 064: 0.6565/0.1501\n",
      "14s - loss: 0.1630 - mean_squared_error: 0.0775\n",
      "Epoch 66/200\n",
      "11s - loss: 0.1625 - mean_squared_error: 0.0777\n",
      "Epoch 67/200\n",
      "Standard error at epoch 066: 0.6705/0.1517\n",
      "MC error at epoch 066: 0.6678/0.1514\n",
      "14s - loss: 0.1589 - mean_squared_error: 0.0753\n",
      "Epoch 68/200\n",
      "11s - loss: 0.1675 - mean_squared_error: 0.0832\n",
      "Epoch 69/200\n",
      "Standard error at epoch 068: 0.6718/0.1519\n",
      "MC error at epoch 068: 0.6699/0.1516\n",
      "14s - loss: 0.1610 - mean_squared_error: 0.0772\n",
      "Epoch 70/200\n",
      "11s - loss: 0.1582 - mean_squared_error: 0.0756\n",
      "Epoch 71/200\n",
      "Standard error at epoch 070: 0.6537/0.1498\n",
      "MC error at epoch 070: 0.6510/0.1495\n",
      "14s - loss: 0.1597 - mean_squared_error: 0.0775\n",
      "Epoch 72/200\n",
      "11s - loss: 0.1603 - mean_squared_error: 0.0784\n",
      "Epoch 73/200\n",
      "Standard error at epoch 072: 0.6709/0.1518\n",
      "MC error at epoch 072: 0.6687/0.1515\n",
      "14s - loss: 0.1612 - mean_squared_error: 0.0795\n",
      "Epoch 74/200\n",
      "11s - loss: 0.1572 - mean_squared_error: 0.0761\n",
      "Epoch 75/200\n",
      "Standard error at epoch 074: 0.6566/0.1501\n",
      "MC error at epoch 074: 0.6538/0.1498\n",
      "14s - loss: 0.1539 - mean_squared_error: 0.0736\n",
      "Epoch 76/200\n",
      "11s - loss: 0.1577 - mean_squared_error: 0.0780\n",
      "Epoch 77/200\n",
      "Standard error at epoch 076: 0.6474/0.1491\n",
      "MC error at epoch 076: 0.6467/0.1490\n",
      "14s - loss: 0.1611 - mean_squared_error: 0.0809\n",
      "Epoch 78/200\n",
      "11s - loss: 0.1537 - mean_squared_error: 0.0742\n",
      "Epoch 79/200\n",
      "Standard error at epoch 078: 0.6518/0.1496\n",
      "MC error at epoch 078: 0.6528/0.1497\n",
      "14s - loss: 0.1527 - mean_squared_error: 0.0741\n",
      "Epoch 80/200\n",
      "11s - loss: 0.1528 - mean_squared_error: 0.0745\n",
      "Epoch 81/200\n",
      "Standard error at epoch 080: 0.6707/0.1517\n",
      "MC error at epoch 080: 0.6684/0.1515\n",
      "14s - loss: 0.1518 - mean_squared_error: 0.0741\n",
      "Epoch 82/200\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "11s - loss: 0.1521 - mean_squared_error: 0.0747\n",
      "Epoch 83/200\n",
      "Standard error at epoch 082: 0.6655/0.1512\n",
      "MC error at epoch 082: 0.6632/0.1509\n",
      "14s - loss: 0.1501 - mean_squared_error: 0.0735\n",
      "Epoch 84/200\n",
      "11s - loss: 0.1520 - mean_squared_error: 0.0754\n",
      "Epoch 85/200\n",
      "Standard error at epoch 084: 0.6508/0.1495\n",
      "MC error at epoch 084: 0.6494/0.1493\n",
      "14s - loss: 0.1507 - mean_squared_error: 0.0745\n",
      "Epoch 86/200\n",
      "11s - loss: 0.1501 - mean_squared_error: 0.0739\n",
      "Epoch 87/200\n",
      "Standard error at epoch 086: 0.6470/0.1490\n",
      "MC error at epoch 086: 0.6455/0.1489\n",
      "14s - loss: 0.1508 - mean_squared_error: 0.0750\n",
      "Epoch 88/200\n",
      "11s - loss: 0.1514 - mean_squared_error: 0.0756\n",
      "Epoch 89/200\n",
      "Standard error at epoch 088: 0.6370/0.1479\n",
      "MC error at epoch 088: 0.6376/0.1479\n",
      "14s - loss: 0.1508 - mean_squared_error: 0.0751\n",
      "Epoch 90/200\n",
      "11s - loss: 0.1475 - mean_squared_error: 0.0723\n",
      "Epoch 91/200\n",
      "Standard error at epoch 090: 0.6579/0.1503\n",
      "MC error at epoch 090: 0.6580/0.1503\n",
      "14s - loss: 0.1472 - mean_squared_error: 0.0725\n",
      "Epoch 92/200\n",
      "11s - loss: 0.1486 - mean_squared_error: 0.0739\n",
      "Epoch 93/200\n",
      "Standard error at epoch 092: 0.6513/0.1495\n",
      "MC error at epoch 092: 0.6500/0.1494\n",
      "14s - loss: 0.1496 - mean_squared_error: 0.0751\n",
      "Epoch 94/200\n",
      "11s - loss: 0.1493 - mean_squared_error: 0.0749\n",
      "Epoch 95/200\n",
      "Standard error at epoch 094: 0.6542/0.1499\n",
      "MC error at epoch 094: 0.6519/0.1496\n",
      "14s - loss: 0.1492 - mean_squared_error: 0.0749\n",
      "Epoch 96/200\n",
      "12s - loss: 0.1482 - mean_squared_error: 0.0744\n",
      "Epoch 97/200\n",
      "Standard error at epoch 096: 0.6491/0.1493\n",
      "MC error at epoch 096: 0.6472/0.1491\n",
      "14s - loss: 0.1443 - mean_squared_error: 0.0711\n",
      "Epoch 98/200\n",
      "11s - loss: 0.1439 - mean_squared_error: 0.0712\n",
      "Epoch 99/200\n",
      "Standard error at epoch 098: 0.6560/0.1501\n",
      "MC error at epoch 098: 0.6534/0.1498\n",
      "14s - loss: 0.1458 - mean_squared_error: 0.0730\n",
      "Epoch 100/200\n",
      "11s - loss: 0.1482 - mean_squared_error: 0.0752\n",
      "Epoch 101/200\n",
      "Standard error at epoch 100: 0.6564/0.1501\n",
      "MC error at epoch 100: 0.6540/0.1498\n",
      "14s - loss: 0.1479 - mean_squared_error: 0.0747\n",
      "Epoch 102/200\n",
      "11s - loss: 0.1471 - mean_squared_error: 0.0738\n",
      "Epoch 103/200\n",
      "Standard error at epoch 102: 0.6421/0.1485\n",
      "MC error at epoch 102: 0.6386/0.1481\n",
      "14s - loss: 0.1448 - mean_squared_error: 0.0721\n",
      "Epoch 104/200\n",
      "11s - loss: 0.1434 - mean_squared_error: 0.0713\n",
      "Epoch 105/200\n",
      "Standard error at epoch 104: 0.6544/0.1499\n",
      "MC error at epoch 104: 0.6529/0.1497\n",
      "14s - loss: 0.1458 - mean_squared_error: 0.0739\n",
      "Epoch 106/200\n",
      "11s - loss: 0.1433 - mean_squared_error: 0.0713\n",
      "Epoch 107/200\n",
      "Standard error at epoch 106: 0.6570/0.1502\n",
      "MC error at epoch 106: 0.6542/0.1499\n",
      "14s - loss: 0.1400 - mean_squared_error: 0.0687\n",
      "Epoch 108/200\n",
      "11s - loss: 0.1430 - mean_squared_error: 0.0721\n",
      "Epoch 109/200\n",
      "Standard error at epoch 108: 0.6406/0.1483\n",
      "MC error at epoch 108: 0.6385/0.1480\n",
      "14s - loss: 0.1422 - mean_squared_error: 0.0713\n",
      "Epoch 110/200\n",
      "11s - loss: 0.1406 - mean_squared_error: 0.0703\n",
      "Epoch 111/200\n",
      "Standard error at epoch 110: 0.6705/0.1517\n",
      "MC error at epoch 110: 0.6672/0.1513\n",
      "14s - loss: 0.1430 - mean_squared_error: 0.0727\n",
      "Epoch 112/200\n",
      "11s - loss: 0.1436 - mean_squared_error: 0.0730\n",
      "Epoch 113/200\n",
      "Standard error at epoch 112: 0.6660/0.1512\n",
      "MC error at epoch 112: 0.6633/0.1509\n",
      "14s - loss: 0.1439 - mean_squared_error: 0.0734\n",
      "Epoch 114/200\n",
      "11s - loss: 0.1428 - mean_squared_error: 0.0720\n",
      "Epoch 115/200\n",
      "Standard error at epoch 114: 0.6424/0.1485\n",
      "MC error at epoch 114: 0.6411/0.1484\n",
      "14s - loss: 0.1388 - mean_squared_error: 0.0687\n",
      "Epoch 116/200\n",
      "11s - loss: 0.1368 - mean_squared_error: 0.0675\n",
      "Epoch 117/200\n",
      "Standard error at epoch 116: 0.6404/0.1483\n",
      "MC error at epoch 116: 0.6394/0.1482\n",
      "14s - loss: 0.1382 - mean_squared_error: 0.0692\n",
      "Epoch 118/200\n",
      "11s - loss: 0.1390 - mean_squared_error: 0.0705\n",
      "Epoch 119/200\n",
      "Standard error at epoch 118: 0.6579/0.1503\n",
      "MC error at epoch 118: 0.6574/0.1502\n",
      "14s - loss: 0.1405 - mean_squared_error: 0.0716\n",
      "Epoch 120/200\n",
      "11s - loss: 0.1424 - mean_squared_error: 0.0732\n",
      "Epoch 121/200\n",
      "Standard error at epoch 120: 0.6545/0.1499\n",
      "MC error at epoch 120: 0.6496/0.1493\n",
      "14s - loss: 0.1408 - mean_squared_error: 0.0716\n",
      "Epoch 122/200\n",
      "11s - loss: 0.1381 - mean_squared_error: 0.0692\n",
      "Epoch 123/200\n",
      "Standard error at epoch 122: 0.6603/0.1506\n",
      "MC error at epoch 122: 0.6571/0.1502\n",
      "14s - loss: 0.1367 - mean_squared_error: 0.0681\n",
      "Epoch 124/200\n",
      "11s - loss: 0.1380 - mean_squared_error: 0.0696\n",
      "Epoch 125/200\n",
      "Standard error at epoch 124: 0.6717/0.1518\n",
      "MC error at epoch 124: 0.6672/0.1513\n",
      "14s - loss: 0.1396 - mean_squared_error: 0.0711\n",
      "Epoch 126/200\n",
      "11s - loss: 0.1424 - mean_squared_error: 0.0736\n",
      "Epoch 127/200\n",
      "Standard error at epoch 126: 0.6622/0.1508\n",
      "MC error at epoch 126: 0.6579/0.1503\n",
      "14s - loss: 0.1380 - mean_squared_error: 0.0694\n",
      "Epoch 128/200\n",
      "11s - loss: 0.1368 - mean_squared_error: 0.0687\n",
      "Epoch 129/200\n",
      "Standard error at epoch 128: 0.6479/0.1491\n",
      "MC error at epoch 128: 0.6450/0.1488\n",
      "14s - loss: 0.1386 - mean_squared_error: 0.0707\n",
      "Epoch 130/200\n",
      "11s - loss: 0.1370 - mean_squared_error: 0.0693\n",
      "Epoch 131/200\n",
      "Standard error at epoch 130: 0.6542/0.1499\n",
      "MC error at epoch 130: 0.6510/0.1495\n",
      "14s - loss: 0.1387 - mean_squared_error: 0.0710\n",
      "Epoch 132/200\n",
      "11s - loss: 0.1402 - mean_squared_error: 0.0717\n",
      "Epoch 133/200\n",
      "Standard error at epoch 132: 0.6488/0.1492\n",
      "MC error at epoch 132: 0.6457/0.1489\n",
      "14s - loss: 0.1378 - mean_squared_error: 0.0694\n",
      "Epoch 134/200\n",
      "11s - loss: 0.1366 - mean_squared_error: 0.0687\n",
      "Epoch 135/200\n",
      "Standard error at epoch 134: 0.6354/0.1477\n",
      "MC error at epoch 134: 0.6337/0.1475\n",
      "14s - loss: 0.1382 - mean_squared_error: 0.0707\n",
      "Epoch 136/200\n",
      "12s - loss: 0.1386 - mean_squared_error: 0.0709\n",
      "Epoch 137/200\n",
      "Standard error at epoch 136: 0.6499/0.1494\n",
      "MC error at epoch 136: 0.6458/0.1489\n",
      "14s - loss: 0.1341 - mean_squared_error: 0.0666\n",
      "Epoch 138/200\n",
      "11s - loss: 0.1334 - mean_squared_error: 0.0666\n",
      "Epoch 139/200\n",
      "Standard error at epoch 138: 0.6375/0.1479\n",
      "MC error at epoch 138: 0.6357/0.1477\n",
      "14s - loss: 0.1354 - mean_squared_error: 0.0688\n",
      "Epoch 140/200\n",
      "11s - loss: 0.1341 - mean_squared_error: 0.0675\n",
      "Epoch 141/200\n",
      "Standard error at epoch 140: 0.6423/0.1485\n",
      "MC error at epoch 140: 0.6404/0.1483\n",
      "14s - loss: 0.1342 - mean_squared_error: 0.0680\n",
      "Epoch 142/200\n",
      "11s - loss: 0.1343 - mean_squared_error: 0.0683\n",
      "Epoch 143/200\n",
      "Standard error at epoch 142: 0.6468/0.1490\n",
      "MC error at epoch 142: 0.6448/0.1488\n",
      "14s - loss: 0.1387 - mean_squared_error: 0.0725\n",
      "Epoch 144/200\n",
      "11s - loss: 0.1391 - mean_squared_error: 0.0723\n",
      "Epoch 145/200\n",
      "Standard error at epoch 144: 0.6384/0.1480\n",
      "MC error at epoch 144: 0.6372/0.1479\n",
      "14s - loss: 0.1363 - mean_squared_error: 0.0695\n",
      "Epoch 146/200\n",
      "11s - loss: 0.1346 - mean_squared_error: 0.0682\n",
      "Epoch 147/200\n",
      "Standard error at epoch 146: 0.6392/0.1481\n",
      "MC error at epoch 146: 0.6379/0.1480\n",
      "14s - loss: 0.1319 - mean_squared_error: 0.0661\n",
      "Epoch 148/200\n",
      "11s - loss: 0.1347 - mean_squared_error: 0.0689\n",
      "Epoch 149/200\n",
      "Standard error at epoch 148: 0.6515/0.1495\n",
      "MC error at epoch 148: 0.6447/0.1488\n",
      "14s - loss: 0.1374 - mean_squared_error: 0.0712\n",
      "Epoch 150/200\n",
      "11s - loss: 0.1388 - mean_squared_error: 0.0722\n",
      "Epoch 151/200\n",
      "Standard error at epoch 150: 0.6377/0.1480\n",
      "MC error at epoch 150: 0.6332/0.1474\n",
      "14s - loss: 0.1355 - mean_squared_error: 0.0689\n",
      "Epoch 152/200\n",
      "11s - loss: 0.1315 - mean_squared_error: 0.0657\n",
      "Epoch 153/200\n",
      "Standard error at epoch 152: 0.6555/0.1500\n",
      "MC error at epoch 152: 0.6521/0.1496\n",
      "14s - loss: 0.1324 - mean_squared_error: 0.0670\n",
      "Epoch 154/200\n",
      "11s - loss: 0.1324 - mean_squared_error: 0.0672\n",
      "Epoch 155/200\n",
      "Standard error at epoch 154: 0.6522/0.1496\n",
      "MC error at epoch 154: 0.6454/0.1489\n",
      "14s - loss: 0.1339 - mean_squared_error: 0.0685\n",
      "Epoch 156/200\n",
      "11s - loss: 0.1350 - mean_squared_error: 0.0692\n",
      "Epoch 157/200\n",
      "Standard error at epoch 156: 0.6487/0.1492\n",
      "MC error at epoch 156: 0.6414/0.1484\n",
      "14s - loss: 0.1336 - mean_squared_error: 0.0680\n",
      "Epoch 158/200\n",
      "11s - loss: 0.1333 - mean_squared_error: 0.0680\n",
      "Epoch 159/200\n",
      "Standard error at epoch 158: 0.6321/0.1473\n",
      "MC error at epoch 158: 0.6289/0.1469\n",
      "14s - loss: 0.1343 - mean_squared_error: 0.0690\n",
      "Epoch 160/200\n",
      "11s - loss: 0.1355 - mean_squared_error: 0.0700\n",
      "Epoch 161/200\n",
      "Standard error at epoch 160: 0.6434/0.1486\n",
      "MC error at epoch 160: 0.6399/0.1482\n",
      "14s - loss: 0.1333 - mean_squared_error: 0.0680\n",
      "Epoch 162/200\n",
      "11s - loss: 0.1323 - mean_squared_error: 0.0674\n",
      "Epoch 163/200\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Standard error at epoch 162: 0.6458/0.1489\n",
      "MC error at epoch 162: 0.6421/0.1485\n",
      "14s - loss: 0.1326 - mean_squared_error: 0.0679\n",
      "Epoch 164/200\n",
      "11s - loss: 0.1319 - mean_squared_error: 0.0673\n",
      "Epoch 165/200\n",
      "Standard error at epoch 164: 0.6436/0.1486\n",
      "MC error at epoch 164: 0.6402/0.1483\n",
      "14s - loss: 0.1315 - mean_squared_error: 0.0668\n",
      "Epoch 166/200\n",
      "11s - loss: 0.1320 - mean_squared_error: 0.0673\n",
      "Epoch 167/200\n",
      "Standard error at epoch 166: 0.6358/0.1477\n",
      "MC error at epoch 166: 0.6324/0.1473\n",
      "14s - loss: 0.1350 - mean_squared_error: 0.0700\n",
      "Epoch 168/200\n",
      "11s - loss: 0.1329 - mean_squared_error: 0.0676\n",
      "Epoch 169/200\n",
      "Standard error at epoch 168: 0.6582/0.1503\n",
      "MC error at epoch 168: 0.6531/0.1497\n",
      "14s - loss: 0.1312 - mean_squared_error: 0.0665\n",
      "Epoch 170/200\n",
      "11s - loss: 0.1318 - mean_squared_error: 0.0672\n",
      "Epoch 171/200\n",
      "Standard error at epoch 170: 0.6527/0.1497\n",
      "MC error at epoch 170: 0.6497/0.1493\n",
      "14s - loss: 0.1355 - mean_squared_error: 0.0703\n",
      "Epoch 172/200\n",
      "11s - loss: 0.1339 - mean_squared_error: 0.0687\n",
      "Epoch 173/200\n",
      "Standard error at epoch 172: 0.6479/0.1491\n",
      "MC error at epoch 172: 0.6442/0.1487\n",
      "14s - loss: 0.1340 - mean_squared_error: 0.0689\n",
      "Epoch 174/200\n",
      "11s - loss: 0.1306 - mean_squared_error: 0.0656\n",
      "Epoch 175/200\n",
      "Standard error at epoch 174: 0.6496/0.1493\n",
      "MC error at epoch 174: 0.6453/0.1488\n",
      "14s - loss: 0.1310 - mean_squared_error: 0.0668\n",
      "Epoch 176/200\n",
      "11s - loss: 0.1298 - mean_squared_error: 0.0657\n",
      "Epoch 177/200\n",
      "Standard error at epoch 176: 0.6395/0.1482\n",
      "MC error at epoch 176: 0.6351/0.1477\n",
      "14s - loss: 0.1301 - mean_squared_error: 0.0664\n",
      "Epoch 178/200\n",
      "11s - loss: 0.1318 - mean_squared_error: 0.0681\n",
      "Epoch 179/200\n",
      "Standard error at epoch 178: 0.6602/0.1505\n",
      "MC error at epoch 178: 0.6529/0.1497\n",
      "14s - loss: 0.1310 - mean_squared_error: 0.0672\n",
      "Epoch 180/200\n",
      "11s - loss: 0.1325 - mean_squared_error: 0.0684\n",
      "Epoch 181/200\n",
      "Standard error at epoch 180: 0.6351/0.1477\n",
      "MC error at epoch 180: 0.6314/0.1472\n",
      "14s - loss: 0.1311 - mean_squared_error: 0.0671\n",
      "Epoch 182/200\n",
      "11s - loss: 0.1304 - mean_squared_error: 0.0662\n",
      "Epoch 183/200\n",
      "Standard error at epoch 182: 0.6365/0.1478\n",
      "MC error at epoch 182: 0.6323/0.1473\n",
      "14s - loss: 0.1307 - mean_squared_error: 0.0667\n",
      "Epoch 184/200\n",
      "11s - loss: 0.1308 - mean_squared_error: 0.0668\n",
      "Epoch 185/200\n",
      "Standard error at epoch 184: 0.6418/0.1484\n",
      "MC error at epoch 184: 0.6370/0.1479\n",
      "14s - loss: 0.1338 - mean_squared_error: 0.0694\n",
      "Epoch 186/200\n",
      "11s - loss: 0.1337 - mean_squared_error: 0.0690\n",
      "Epoch 187/200\n",
      "Standard error at epoch 186: 0.6301/0.1471\n",
      "MC error at epoch 186: 0.6268/0.1467\n",
      "14s - loss: 0.1314 - mean_squared_error: 0.0668\n",
      "Epoch 188/200\n",
      "11s - loss: 0.1311 - mean_squared_error: 0.0670\n",
      "Epoch 189/200\n",
      "Standard error at epoch 188: 0.6403/0.1483\n",
      "MC error at epoch 188: 0.6373/0.1479\n",
      "14s - loss: 0.1303 - mean_squared_error: 0.0665\n",
      "Epoch 190/200\n",
      "11s - loss: 0.1299 - mean_squared_error: 0.0665\n",
      "Epoch 191/200\n",
      "Standard error at epoch 190: 0.6438/0.1487\n",
      "MC error at epoch 190: 0.6367/0.1478\n",
      "14s - loss: 0.1313 - mean_squared_error: 0.0680\n",
      "Epoch 192/200\n",
      "11s - loss: 0.1297 - mean_squared_error: 0.0664\n",
      "Epoch 193/200\n",
      "Standard error at epoch 192: 0.6353/0.1477\n",
      "MC error at epoch 192: 0.6320/0.1473\n",
      "14s - loss: 0.1296 - mean_squared_error: 0.0660\n",
      "Epoch 194/200\n",
      "11s - loss: 0.1294 - mean_squared_error: 0.0661\n",
      "Epoch 195/200\n",
      "Standard error at epoch 194: 0.6360/0.1478\n",
      "MC error at epoch 194: 0.6348/0.1476\n",
      "14s - loss: 0.1291 - mean_squared_error: 0.0660\n",
      "Epoch 196/200\n",
      "11s - loss: 0.1309 - mean_squared_error: 0.0677\n",
      "Epoch 197/200\n",
      "Standard error at epoch 196: 0.6274/0.1468\n",
      "MC error at epoch 196: 0.6234/0.1463\n",
      "14s - loss: 0.1310 - mean_squared_error: 0.0673\n",
      "Epoch 198/200\n",
      "11s - loss: 0.1318 - mean_squared_error: 0.0681\n",
      "Epoch 199/200\n",
      "Standard error at epoch 198: 0.6453/0.1488\n",
      "MC error at epoch 198: 0.6423/0.1485\n",
      "14s - loss: 0.1296 - mean_squared_error: 0.0659\n",
      "Epoch 200/200\n",
      "11s - loss: 0.1321 - mean_squared_error: 0.0683\n"
     ]
    }
   ],
   "source": [
    "history_3 = model.fit(\n",
    "    X_train, Y_train,\n",
    "    verbose=2,\n",
    "    shuffle=True,\n",
    "    # validation_data=[X_test, Y_test],\n",
    "    batch_size=BATCH_SIZE, epochs=200, callbacks=[modeltest_3])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best RMSE: 0.1425 Best Epoch: 12\n"
     ]
    }
   ],
   "source": [
    "print(\"Best RMSE: {:.4f} Best Epoch: {}\".format(\n",
    "    np.min([x[1] ** 0.5 for x in modeltest_3.history]), \n",
    "    (np.argmin([x[1] ** 0.5 for x in modeltest_3.history]) + 1)*2\n",
    "))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "## Visualizations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.text.Text at 0x7f7b89d9bb70>"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAzkAAAE0CAYAAAAPE/3uAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQvQo0skRQgghhBBC9Cj/D7xQ9WhcNWb8AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b89e53be0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(12, 4))\n",
    "plt.subplot(1, 2, 1)\n",
    "plt.title(\"Raw MSE Comparison - Training Set\")\n",
    "plt.plot(np.arange(len(history_1.history[\"mean_squared_error\"])), \n",
    "         history_1.history[\"mean_squared_error\"], label=\"variational\")\n",
    "plt.plot(np.arange(len(history_2.history[\"mean_squared_error\"])), \n",
    "         history_2.history[\"mean_squared_error\"], \"g-\", label=\"no dropout\")\n",
    "plt.plot(np.arange(len(history_3.history[\"mean_squared_error\"])), \n",
    "         history_3.history[\"mean_squared_error\"], \"y-\", label=\"naive dropout\")\n",
    "plt.legend(loc='best')\n",
    "plt.xlabel(\"epochs\")\n",
    "plt.ylabel(\"Raw MSE\")\n",
    "plt.subplot(1, 2, 2)\n",
    "plt.title(\"(MC - Approx) Histogram\")\n",
    "plt.hist([x[1] ** 0.5 - x[0] ** 0.5 for x in modeltest_1.history], alpha=0.5, label=\"varational\")\n",
    "plt.hist([x[1] ** 0.5 - x[0] ** 0.5 for x in modeltest_3.history], alpha=0.5, label=\"navie dropout\")\n",
    "plt.legend(loc='best')\n",
    "plt.xlabel(\"Difference in Raw MSE\")\n",
    "plt.ylabel(\"Count\")\n",
    "plt.xticks(fontsize=8, rotation=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.text.Text at 0x7f7b89ed6240>"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAjwAAAFuCAYAAACFocL7AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b90082208>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.title(\"RMSE Comparison - Validation Set\")\n",
    "plt.plot(np.arange(len(modeltest_1.history)), [x[1] ** 0.5 for x in modeltest_1.history], \"b-\", label=\"variational(mc)\")\n",
    "plt.plot(np.arange(len(modeltest_2.history)), [x[0] ** 0.5 for x in modeltest_2.history], \"g-\", label=\"no dropout\")\n",
    "plt.plot(np.arange(len(modeltest_3.history)), [x[1] ** 0.5 for x in modeltest_3.history], \"y-\", label=\"naive dropout(mc)\")\n",
    "plt.legend(loc='best')\n",
    "plt.xlabel(\"epochs\")\n",
    "plt.ylabel(\"RMSE\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}